OpenCV4.0 Change Logs

OpenCV Change Logs


4.0.0-alphaSeptember, 2018

4.0.0-betaOctober, 2018

We are moving towards OpenCV 4.0 gold. Here is what's new in OpenCV 4.0 alpha/beta:

  • OpenCV is now C++11 library and requires C++11-compliant compiler. Note that OpenCV 3.x can also be built as C++11 library by passing -DENABLE_CXX11=ON flag to CMake. Now this is the only option so the flag is not needed and is not available anymore.

    • Thanks to the extended C++11 standard library, we could get rid of hand-crafted cv::String and cv::Ptr. Now cv::String == std::string and cv::Ptr is a thin wrapper on top of std::shared_ptr. Also, on Linux/BSD for cv::parallel_for_ we now use std::thread's instead of pthreads.

  • DNN improvements

    • Added support for Mask-RCNN model. Follow a guide and use a python sample.

    • Integrated ONNX parser. We now support many popular classification networks. YOLO object detection network in partially supported as well (ONNX version of YOLO lacks some final layers that actually give you the list of rectangles). Thanks to Lubov Batanina for her first-time contribution to OpenCV!

    • Further improved performance of DNN module when it's built with Intel DLDT support by utilizing more layers from DLDT.

    • API changes: by default, blobFromImage methods family do not swap red and blue channels and do not crop the input image. Beware that this API change has also been propagated to OpenCV 3.4 branch.

    • Added shortcuts for the most popular deep learning networks supported by OpenCV. You may specify an alias name of model to skip pre-processing parameters and even paths to models! In example, instead of

      python --model opencv_face_detector.caffemodel --config opencv_face_detector.prototxt --mean 104 177 123 --width 300 --height 300

      you can type just

      python opencv_fd
    • Fixed OpenCL target for AMD and NVIDIA GPUs. Now you may enable DNN_TARGET_OPENCLfor your model without extra environment variables. Please note that DNN_TARGET_OPENCL_FP16 is tested on Intel GPUs only so it still requires additional flags.

  • Performance improvements

    • A few hundreds of basic kernels in OpenCV have been rewritten using so-called "wide universal intrinsics". Those intrinsics map to SSE2, SSE4, AVX2, NEON or VSX intrinsics, depending on the target platform and the compile flags. It should translate to noticeably better performance, even for some already optimized functions. For example, if you configure and compile OpenCV with CPU_BASELINE=AVX2 CMake flag, you can get extra 15-30% speed improvement for certain image processing operations. By OpenCV 4.0 gold we plan to translate many more kernels to such intrinsics and also employ our dynamic dispatching mechanism, so that on x64 platform AVX2-optimized kernels are always built-in and could be selected on-fly if the actual hardware supports such instructions (without having to change CPU_BASELINE). Big thanks to Sayed Adel for VSX backend and the initial (but rather complete!) AVX2 backend of the wide universal intrinsics! Big thanks to Vitaly Tuzov for conversion of many functions in OpenCV to wide universal intrinsics.

  • QR code detector and decoder have been added to opencv/objdetect module together with live sample. Currently, the decoder is built on top of QUirc library (which snapshot we put into opencv). Thanks to our intern Alexander Nesterov for this important contribution!

  • The popular Kinect Fusion algorithm has been implemented, optimized for CPU and GPU (OpenCL), and integrated into opencv_contrib/rgbd module. To make the live sample work, we updated our Kinect 2 support in opencv/videoio module. In 4.0 beta the code has been accelerated for iGPU, which resulted in 3x performance boost in the high-resolution case (512x512x512 integration volume). Thanks to Rostislav Vasilikhin for the excellent work!





August, 2018

OpenCV 3.4.3 has been released, with further extended dnn module, documentation improvements, some other new functionality and bug fixes.



opencv (35 contributors)

opencv_contrib (13 contributors)


July, 2018

OpenCV 3.4.2 has been released, with further extended dnn module, documentation improvements, some other new functionality and bug fixes.

OpenCV 3.4.x development is switched from "master" to "3.4" branch. "master" branch is used for development of upcoming OpenCV 4.x releases.

Bugfixes / optimizations / small improvemets should go into "3.4" branch. We will merge changes from "3.4" into "master" regularly (weekly/bi-weekly).

  • DNN improvements

    • Added a new computational target DNN_TARGET_OPENCL_FP16 for half-precision floating point arithmetic of deep learning networks using OpenCL. Just use net.setPreferableTarget(DNN_TARGET_OPENCL_FP16).
    • Extended support of Intel's Inference Engine backend to run models on GPU (OpenCL FP32/FP16) and VPU (Myriad 2, FP16) devices. See an installation guide for details.
    • Enabled import of Intel's OpenVINO pre-trained networks from intermediate representation (IR).
    • Introduced custom layers support which let you define unimplemented layers or override existing ones. Learn more in a corresponding tutorial.
    • Implemented a new deep learning sample inspired by EAST: An Efficient and Accurate Scene Text Detector.
    • Added a support of YOLOv3 and image classification models from Darknet framework.
    • Reduced top DNN's memory consumption and improvements in support of networks from TensorFlow and Keras.

  • Excellent tutorials improvements. Python / Java examples are merged into C++ tutorials. Big thanks to Catree.

  • Video I/O: improved support of Microsoft Media Foundation (MSMF)


opencv (83 contributors)

opencv_contrib (17 contributors)


February, 2018

OpenCV 3.4.1 has been just released, with further extended dnn module, multiple bug fixes and other small and big improvements.

  • Added support for quantized TensorFlow networks. We are now able to load 8-bit matrices of weights. The computations are still done in FP32 (with plans to add FP16 path), but even now it's possible to compress the networks so that they consume less disk space. For example, our face detector can be modified to read and use our 8-bit quantized ResNet-based face detection network (pass it and this description file into cv::dnn::readNetFromTensorFlow()). And the face detection network takes just 2.7Mb.

  • OpenCV is now able to use Intel DL inference engine as DNN acceleration backend. It gives quite noticeable performance boost on many models:

Model CPU, default backend CPU, Inference Engine backend, MKL-DNN plugin Model Optimizer + Inference Engine, MKL-DNN plugin (a standalone application)
AlexNet 14.44ms 12.09ms (x1.19) 12.05ms
GoogLeNet 15.26ms 8.92ms (x1.71) 8.75ms
ResNet-50 35.78ms 19.53ms (x1.83) 19.4ms
SqueezeNet v1.1 4.01ms 2.60ms (x1.54) 2.5ms
MobileNet-SSD from Caffe 21.62ms 8.89ms (x2.43)  
DenseNet-121 61.71ms 28.21ms (x2.18)  
OpenPose (COCO) @ 368x368 885.57ms 544.05ms (x1.62)  
OpenPose (MPI) @ 368x368 879.13ms 533.96ms (x1.64)  
OpenPose (MPI, 4 stages) @ 368x368 605.63ms 378.49ms (x1.60)  
OpenFace 3.84ms 2.59ms (x1.48)  
  • Added AVX-512 acceleration to the performance-critical kernels, such as convolution and fully-connected layers. Some networks, such as SSD object detection and ENet semantical segmentation, have been accelerated by ~20%.

  • SSD-based models trained and retrained in TensorFlow Object Detection API can be easier imported by a single invocation of python script making a text graph representation. Read a wiki page for details.

  • Performance of pthreads backend of cv::parallel_for_() (which is used by default on Linux/Android, unless you installed TBB or chose OpenMP) has been greatly improved on many core machines, in particular 10-core Core i9. That let us to increase performance of DNN inference quite significantly (up to 6x) on such machines.

  • OpenCL backend has been expanded to cover more layers. The layer fusion has also been improved to increase the speed even further. It shall be reminded that in order to enable OpenCL backend (if it's available on the host machine) one should call the method my_dnn_net.setPreferableTarget(cv::dnn::DNN_TARGET_OPENCL) before the inference, where my_dnn_net is the network loaded using cv::dnn::readNetFromCaffe()cv::dnn::readNetFromTensorFlow() etc.

  • Several bugs in various layers have been fixed; in particular, SSD priors are now computed slightly differently so that we can more accurate bounding boxes when running SSD on variable-size images.

  • On-disk caching of precompiled OpenCL kernels has been fixed to comply with OpenCL standard. Correspondingly, it now works well with the new Intel OpenCL (NEO) drivers.
  • Certain cases with UMat deadlock when copying UMats in different threads has been fixed.

  • Supported Android NDK16
  • Added build.gradle into OpenCV 4 Android SDK
  • Added initial support of Camera2 API via JavaCamera2View interface

  • C++11: added support of multi-dimentional cv::Mat creation via C++ initializers lists:

    auto K = Mat_<double>({3, 3}, {0, -1, 0, -1, 5, -1, 0, -1, 0});
  • C++17: OpenCV source code and tests comply C++17 standard

  • opencv_contrib: added GMS matching

  • opencv_contrib: added CSR-DCF tracker

  • opencv_contrib: several improvements in OVIS module (OGRE 3D based visualizer)

  • ~248 patches have been merged since OpenCV 3.4.0

  • 203 issues have been closed



$ git shortlog --no-merges -ns 3.4.0..3.4.1
    98  Alexander Alekhin
    27  Dmitry Kurtaev
    27  Li Peng
    27  Maksim Shabunin
    10  Fangjun Kuang
     6  Vitaly Tuzov
     6  luz.paz
     5  Namgoo Lee
     4  Tomoaki Teshima
     3  Arjan van de Ven
     3  Glue Crow
     3  LaurentBerger
     3  Sayed Adel
     3  Suleyman TURKMEN
     2  Alessandro de Oliveira Faria (A.K.A.CABELO)
     2  Arthur Williams
     2  Gregory Morse
     2  Hamdi Sahloul
     2  Pavel Rojtberg
     2  Rostislav Vasilikhin
     2  Steven
     2  Woody Chow
     2  berak
     2  csukuangfj
     2  oqtvs
     2  victor-ludorum
     2  wxzs5
     1  Aaron Bray
     1  Aditya Rohan
     1  Alexander Nesterov
     1  Ali Sentas
     1  Ce Zheng
     1  Chris Hall
     1  Coralie RACHEX
     1  Csaba Keszegh
     1  David Koller
     1  Elena Gvozdeva
     1  Eric Wang
     1  Erik Man
     1  Florian Echtler
     1  GlueCrow
     1  Harshal Nishar
     1  Jozef Mlich
     1  Jun Zhao
     1  Li, Peng
     1  Mark Harfouche
     1  Moshe
     1  PalkoNazar
     1  Patrik Huber
     1  Philipp Hasper
     1  Ryo Suzuki
     1  Rémi Ratajczak
     1  SarenT
     1  Stanislaw Halik
     1  Sui Libin
     1  Tom Becker
     1  Travers
     1  Vinay Sharma
     1  Ya-Chiu Wu
     1  catree
     1  denmatfoton
     1  elenagvo
     1  jasjuang
     1  mworchel
     1  razerhell
     1  sergeybe
     1  take1014
     1  tobycollins
     1  woody.chow
     1  yuki takehara


$ git shortlog --no-merges -ns 3.4.0..3.4.1
    10  Pavel Rojtberg
     4  Hamdi Sahloul
     3  Alexander Alekhin
     3  KUANG Fangjun
     2  Maksim Shabunin
     2  Vitaly Tuzov
     2  berak
     2  catree
     1  Adam
     1  Andrej Muhič
     1  Fangjun Kuang
     1  John Cant
     1  Kv Manohar
     1  Plamen Dimitrov
     1  Suleyman TURKMEN
     1  Yujun Shi
     1  adrianheron
     1  caseymcc
     1  klchang


December, 2017

OpenCV 3.4 (NY 2018 edition) has been released, with further extended dnn module, some other new functionality, improvements and bug fixes.

The results of yet another GSoC 2017 OpenCV project on implementing new background subtraction algorithms have been integrated. Please, check the script, the algorithms LSBP-* and GSOC-*.

opencv_dnn has been further improved and extended:

  • Added faster R-CNN support and the corresponding example.

  • Javascript bindings have been extended to cover DNN module. See the corresponding tutorial
  • Also, DNN has been further accelerated for iGPU using OpenCL. In particular, MobileNet-SSD networks now run ~7 times faster than in OpenCV 3.3.1. Big thanks to Peng Li and his team for the contribution.

  • On-disk caching of precompiled OpenCL kernels has been finally implemented. It noticeably reduces initialization time of applications that use a lot of kernels.
  • Also, it's now possible to load and run pre-compiled OpenCL kernels via T-API. It can be useful on embedded platforms without OpenCL JIT compiler available. For details, see the following guide on OpenCL acceleration in OpenCV.

  • Bit-exact 8-bit and 16-bit resize has been implemented (currently supported only bilinear interpolation). Use INTER_LINEAR_EXACT interpolation mode. In many places in the library we've switched to this new resize. Bit-exact means that on any platform with any compiler etc. you will get absolutely the same results for the same scale factor values, there will be no difference (even +/-1) in pixel values in the output image. The function complements a few other bit-exact algorithms added in OpenCV 3.3.1: cvtColor(RGB<=>Lab, RGB<=>Luv).

  • Now the reference guide includes function/method prototypes for the Python bindings, e.g. type "convexHull" in the search field and you will see the C++ declaration and the Python wrapper spec below. That was another GSoC 2017 project. Big thanks to Cartucho and Vincent Rabaud.
  • Excellent tutorial on how to compute and use homography matrix in OpenCV has been added. Big thanks to Catree.
  • The guide on how to use OpenCV with Intel Media SDK in order to get hardware-accelerated video decoding/encoding has been written by Maksim Shabunin.
  • Another guide has been added a few months ago by Alexander Alekhin, but was not mentioned before, on how to profile OpenCV-based applications using Intel Tracing Technology tool.
  • Quite comprehensive set of tutorials on how to use OpenCV within a browser (via javascript bindings) has been created during GSoC 2017

  • ~250 patches have been merged since OpenCV 3.3.1
  • over 200 issues have been closed



103	Alexander Alekhin
 25	Dmitry Kurtaev
 23	Maksim Shabunin
 17	Li Peng
 17	elenagvo
 14	Vitaly Tuzov
 10	Suleyman TURKMEN
 10	catree
  8	Tomoaki Teshima
  6	Sayed Adel
  5	LaurentBerger
  5	Vladislav Sovrasov
  5	Wu Zhiwen
  4	Pavel Rojtberg
  4	Rostislav Vasilikhin
  4	Vadim Pisarevsky
  3	tribta
  2	Bhanudutta
  2	Fakabbir Amin
  2	Mattia Rizzolo
  2	Ryan Fox
  2	Shinya Ishikawa
  2	berak
  2	dtmoodie
  1	Akhilesh Kumar
  1	Alexander Nesterov
  1	Alexander Rybnikov
  1	Amro
  1	Andrey Smorodov
  1	Arthur Pastel
  1	Cartucho
  1	Christof Kaufmann
  1	David Geldreich
  1	Elena Gvozdeva
  1	Florian Echtler
  1	Hamdi Sahloul
  1	Haritha
  1	Iago Suárez
  1	Igor Wodiany
  1	Ivan Pozdeev
  1	Jacob MacRitchie
  1	James Perkins
  1	Jcrist99
  1	Jiri Horner
  1	Jonathan Viney
  1	Juha Reunanen
  1	KUANG Fangjun
  1	Mikhail Paulyshka
  1	Muhammad Abdullah
  1	Nickola
  1	Pushkal Katara
  1	Riyuzakii
  1	Roman Cattaneo
  1	Shresth Verma
  1	Simon Guo
  1	Wei Hao
  1	Wu, Zhiwen
  1	alessandro faria
  1	gdkessler
  1	klchang
  1	woody.chow
  1	wxzs5
  1	zhijackchen
  1	zhongwuzw


20	sghoshcvc
18	Vladislav Sovrasov
 9	Alexander Alekhin
 6	Vitaly Tuzov
 6	berak
 5	Hamdi Sahloul
 5	Maksim Shabunin
 4	Pavel Rojtberg
 4	Suman Ghosh
 3	LaurentBerger
 3	Leonardo Brás
 2	Suleyman TURKMEN
 1	Adam Gradzki
 1	Anup Parikh
 1	Dmitry Kurtaev
 1	Egor Pugin
 1	Leonardo lontra
 1	Oleg Kalachev
 1	Vladislav Samsonov
 1	cDc
 1	fiammante
 1	klchang
 1	kurnianggoro
 1	kushalvyaskv
 1	sukhad-app


October, 2017

OpenCV 3.3.1 with extended dnn module, most of GSoC 2017 results integrated and other improvements has been released.

Results of several GSoC 2017 projects have been integrated:

  • multi-language (e.g. C++/Python/Java) tutorials by João Cartucho, mentored by Vincent Rabaud
  • AKAZE acceleration by Jiri Horner, mentored by Bence Magyar
  • End-to-end text detection and recognition by Suman Kumar Ghosh, mentored by Prasanna Krishnasamy

One of GSoC 2017 projects that deserves a dedicated section in the change log:

  • Javascript interface to OpenCV (via Emscripten technology) and interactive Web-based OpenCV tutorials by Gang Song and Congxiang Pan. This small yet powerful team was supervised by Sajjad Taheri, Ningxin Hu and Mohammad R Haghighat.

opencv_dnn has been further improved and extended; new samples have been added:

  • Face detection sample and the light-weight Resnet-10 + SSD based network have been added. See the example for details. The detector runs around 20-50FPS on a normal desktop/laptop, and the network is just 10MB (FP32) or even 5MB (FP16).

  • The partial Darknet parser, enough to parse YOLO models, as well as the layers to support a few variations of YOLO object detection networks have been integrated. See the corresponding sample.

  • Preliminary support for FP16 networks has been added. We do not do computations in FP16 yet, we convert FP16 coeffs to FP32 when loading the networks. In the case of Caffe we rely on the following fork, whereas in the case of TF we use the official version.

  • Several new layers have been added to support text detection, image colorization and some other networks.

  • OpenCV has been optimised for PPC64 (64-bit PowerPC) architecture by mapping the universal intrinsics to VSX. Big thanks to Sayed Adel for the patches.

  • OpenCL acceleration path of the bioinspired module has been restored. See the bioinspired-based HDR/Background segmentation example. On Iris Pro HD5200 we get ~5x acceleration over the CPU branch.

  • KCF tracker has been accelerated by ~40%.

  • Hardware-accelerated video encoding/decoding via MediaSDK is now available on Windows too.



87  Alexander Alekhin
32  Dmitry Kurtaev
20  Vladislav Sovrasov
19  Rostislav Vasilikhin
19  tribta
14  Suleyman TURKMEN
10  saskatchewancatch
 9  Tomoaki Teshima
 8  Maksim Shabunin
 6  Ningxin Hu
 5  LaurentBerger
 5  berak
 4  Pavel Rojtberg
 4  Steven Puttemans
 4  Vitaly Tuzov
 3  Pavel Vlasov
 3  KUANG Fangjun
 3  Boris Fomitchev
 3  Igor Wodiany
 2  Ryan Fox
 2  Nishant Arora
 2  Emanuele Ruffaldi
 2  Wu Zhiwen
 2  Tony0
 2  Adi Shavit
 2  Christof Kaufmann
 2  Aleksandr Rybnikov
 2  Sayed Adel
 2  stalin18
 2  Jiri Horner
 2  Jasper Shemilt
 2  Zoltán Mizsei
 2  blendin
 1  Ricardo Ribalda Delgado
 1  Sayan Sinha
 1  Sebastian Höffner
 1  Sergey Alexandrov
 1  Vivek Wisdom
 1  Vladislav Vinogradov
 1  Woody Chow
 1  Yaron Inger
 1  bradford barr
 1  catree
 1  jodrescher
 1  joncrall
 1  jrobble
 1  lzx1413
 1  mevatron
 1  pengli
 1  tz70s
 1  vipinanand4
 1  678098
 1  woody.chow
 1  Adam Rankin
 1  Alex Zatsman
 1  AlexeyAB
 1  Arvid Piehl Lauritsen Böttiger
 1  Balint Cristian
 1  Benoit Blanchon
 1  Bisaloo
 1  Camille Barneaud
 1  Congxiang Pan
 1  David Geldreich
 1  Deric Crago
 1  Faheel Ahmad
 1  Geoff McIver
 1  Gregory Morse
 1  Guillaume Jacob
 1  James Perkins
 1  Jeremy Maitin-Shepard
 1  John Hany
 1  KUANG, Fangjun
 1  Kerry Billingham
 1  Khem Raj
 1  Kim Lindberg Schwaner
 1  Lewis B
 1  Li Peng
 1  Maarten de Vries
 1  Martin Seyferth
 1  Mattia Rizzolo
 1  Patel, Nilaykumar K
 1  Peter Fischer
 1  Peter J. Stieber
 1  Philipp Hasper
 1  Pranit Bauva
 1  Raymond Kirk


19  Alexander Alekhin
19  Vladislav Sovrasov
13  Anguelos Nicolaou
 4  Suleyman TURKMEN
 3  Vitaly Tuzov
 3  acyen
 3  jaco
 2  Woody Chow
 2  Maksim Shabunin
 2  LaurentBerger
 2  Andrew Hundt
 2  Fernando Martin
 1  baudenri
 1  f3rm4rf3r
 1  jasjuang
 1  jrobble
 1  klchang
 1  Ahmed Yarub Hani
 1  woody.chow
 1  Albert
 1  Cristian Duguet
 1  Dmitry Kurtaev
 1  Felix Wehnert
 1  Francis Charette Migneault
 1  Gregor Kovalčík
 1  Jaykob
 1  Linus Ericsson
 1  Nuzhny007
 1  Piotr Majkrzak
 1  atinfinity


August, 2017

OpenCV 3.3 with accelerated dnn module, moved to the main repository, first-ever appearance of Halide in OpenCV, multiple optimizations and other improvements has been released.

opencv_dnn module has been moved from the contribution repository (opencv_contrib) to the main repository (opencv) and was significantly improved:

  • High-level API has been modified and is even more convenient now.

  • The regression tests have been expanded, some new tests have been added. Now, there are 46 of them.

  • Many bugs have been fixed in Torch and TF loaders, as well as in some processing layers. Now we check that on a certain set of networks the results from OpenCV DNN match or very close to the results from the original frameworks. We also check that the results claimed in the papers for such networks are achievable with OpenCV DNN.

  • Performance has been substantially improved. Layer fusion has been implemented and some performance-critical layers have been optimized using AVX, AVX2, SSE and NEON. An external BLAS (OpenBLAS, MKL, ATLAS) is not needed anymore.

  • New samples in C++ and Python have been added.

  • The optional Halide backend has been added. It can accelerate OpenCV DNN on GPU when the GPU is fast enough.

    See Deep Learning in OpenCV for details about the module.

  • Upgraded IPPICV from 2015.12 to 2017.2 version brought ~15% speed improvement into core and imgproc modules (measured as geometrical mean over the corresponding performance tests).

  • Dynamic dispatching of SSE4.2/AVX/AVX2 code has been implemented. Previously, OpenCV had to be built with SSE4.x/AVX/AVX2 turned on in order to use such optimizations and that made it incompatible with older hardware. Now the OpenCV binaries automatically adapt to the real hardware and make use of new instructions if they are available while retaining compatibility with older hardware. All the existing AVX/AVX2 optimizations in OpenCV have been refactored to use this technology. AVX acceleration of DNN also uses dynamic dispatching.

  • OpenCV can now be configured and built as C++ 11 library. Pass -DENABLE_CXX11=ON to CMake. On some modern Linux distributions, like the latest Fedora, it’s enabled by default.

  • New features for C++ 11 users have been added:

    // 1. use parallel_for_ with lambda to compute Mandelbrot fractal
    parallel_for_(Range(0, img.rows*img.cols), [&](const Range& r) {
        for (int r = r.start; r < r.end; r++) {
            int i = r/img.cols, j = r%img.cols, t, maxiter=500;
            complex<float> z0(j * 2.7f/img.cols – 2.1f,
                              i * 2.4f/img.rows – 1.2f);
            for (t = 0; t < maxiter; t++) {
                if (z.real()*z.real()+z.imag()*z.imag()> 4.f) break;
                z = z*z + z0;
  <uchar>(i, j)=uchar(sqrt((float)t/maxiter)*255);
    // 2. using C++ 11 initializers
    auto A = Mat_<double>({0, -1, 0, -1, 5, -1, 0, -1, 0}).reshape(1, 3);
    Vec_<int, 10> digits = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
    // 3. apply LUT to 8-bit image in C++ 11 style
    Mat_<Vec3b> img = imread("lena.jpg");
    for( auto& pixel: img ) {
        pixel[0] = gamma_lut[pixel[0]];
        pixel[1] = gamma_lut[pixel[1]];
        pixel[2] = gamma_lut[pixel[2]];

  • Support for hardware-accelerated video encoding/decoding using Intel GPUs through Intel Media SDK has been implemented for Linux (in the form of backends for cv::VideoCapture and cv::VideoWriter).



210	Alexander Alekhin
 36	Vladislav Sovrasov
 35	Tomoaki Teshima
 31	Vitaly Tuzov
 26	Maksim Shabunin
 21	Fangjun KUANG
 17	chacha21
 14	LaurentBerger
 12	berak
 11	Rostislav Vasilikhin
 11	catree
 10	dkurt
  8	Matthias Grundmann
  8	Pavel Rojtberg
  8	Pyotr Chekmaryov
  8	Vadim Pisarevsky
  8	Woody Chow
  7	Aleksandr Rybnikov
  7	atinfinity
  6	Hamdi Sahloul
  6	vskarlsruhe
  5	Jiri Horner
  5	KUANG, Fangjun
  5	Michele Cancilla
  5	StevenPuttemans
  4	Dal Rupnik
  4	Naba Kumar
  4	Pavel Vlasov
  4	chrizandr
  4	saskatchewancatch
  3	Claudio
  3	Jaime O. Rios
  3	Peter Würtz
  3	Suleyman TURKMEN
  3	Wenju He
  3	Yuriy Solovyov
  3	abratchik
  3	nnorwitz
  3	oulenz
  2	Alexander Broemmer
  2	Artem Lukoyanov
  2	Chaoshuai Lu
  2	Drew Hintz
  2	Kerry Billingham
  2	LukeZhu
  2	Matt Bennett
  2	Philipp Hasper
  2	Rink Springer
  2	Sergiu Deitsch
  2	Tetragramm
  2	Tong Ke
  2	Utkarsh Sinha
  2	jveitchmichaelis
  2	lewisjb
  2	mshabunin
  2	tribta
  2	utibenkei
  2	ville-v
  2	ziggy90127
  1	Adeel
  1	Adeel Ahmad
  1	Alessandro Gentilini
  1	Alexander Duda
  1	Amro
  1	André Mewes
  1	Anupam Sobti
  1	Arek
  1	Arkadiusz Raj
  1	Arnaud Brejeon
  1	BadrinathS
  1	Bragadeesh
  1	Brian Armstrong
  1	Bryce Evans
  1	Cartucho
  1	Christof Kaufmann
  1	David Carlier
  1	Dinar
  1	Dinar Ahmatnurov
  1	Dmitry Kurtaev
  1	Duncan Barber
  1	Egor Pugin
  1	Fernando Iglesias Garcia
  1	Frank de Brabander
  1	Gabriel B. Nunes
  1	Guilherme William Guedes
  1	Hans Gaiser
  1	Ilija Puaca
  1	Ismo Puustinen
  1	James Clarke
  1	Jejos
  1	Jeremy Maitin-Shepard
  1	Jinay Patel
  1	Jiří Sejkora
  1	Jon Eyolfson
  1	Jose Gómez
  1	Juha Reunanen
  1	Julian Exner
  1	Julian Tanke
  1	Kene Udeh
  1	Kuan-Yi Li
  1	Kyle Fleming
  1	Lorena García
  1	Louie Helm
  1	Mark Szumowski
  1	Martin Ganeff
  1	Neil Shipp
  1	Neio Zhou
  1	Nicholas Nadeau
  1	Ninad P.Chaudhari
  1	Paolo Perkovic
  1	Pascal Thomet
  1	Patrik Huber
  1	Piotr Semenov
  1	PkLab
  1	Raj Natarajan
  1	Robert Bragg
  1	Robin Chavignat
  1	Ryan Pavlik
  1	Ryuhei Mori
  1	Sadika Sumanapala
  1	Satoshi Tanaka
  1	Sergey
  1	Shuyu Liang
  1	Sixela David
  1	Steve
  1	Steven Puttemans
  1	Tim King
  1	Tony Lian
  1	Tony Theodore
  1	Vadzim Piatrou
  1	Werner Palfinger
  1	Yakun Wang
  1	Yuval Langer
  1	Zhiwei Su
  1	brian-armstrong-discord
  1	cDc
  1	cclauss
  1	daB0bby
  1	gylns
  1	hailong-wang
  1	huangqinjin
  1	insoow
  1	ionut
  1	jexner
  1	jqly
  1	krishraghuram
  1	kvaghel1
  1	likan999
  1	logic1988
  1	mrquorr
  1	mschoeneck
  1	msk-repo01
  1	n001519
  1	neok-m4700
  1	orisano
  1	puqeko
  1	q
  1	remi
  1	rmensing
  1	sovrasov
  1	umbraclet16
  1	vartenkov
  1	wykvictor
  1	wzw
  1	xuning0
  1	ylemkimon
  1	zongwave


44	Vladislav Sovrasov
24	Alexander Alekhin
14	Aleksandr Rybnikov
11	Hamdi Sahloul
11	dkurt
 9	Vadim Pisarevsky
 7	berak
 6	LaurentBerger
 5	Maksim Shabunin
 4	Balint Cristian
 4	arrybn
 3	Dmitry Kurtaev
 3	Li Peng
 3	Pavel Rojtberg
 3	Pavel Vlasanek
 3	Philipp Hasper
 3	Suleyman TURKMEN
 3	jexner
 2	Kumataro
 2	Steven Puttemans
 2	Vladislav Samsonov
 2	Woody Chow
 2	atinfinity
 2	lhelontra
 2	mshabunin
 2	np-csu
 2	olivierpascal
 1	Diego
 1	Edgar Riba
 1	Erik Sandrén
 1	Erwan Normand
 1	Haaris
 1	Jcrist99
 1	Jeremy Nicola
 1	Julian Exner
 1	Justin Hotchkiss Palermo
 1	Kiran Pradeep
 1	Leonardo lontra
 1	Lorena García
 1	Max-Wang
 1	Natsuki Kawai
 1	Rostislav Vasilikhin
 1	Sergey
 1	StevenPuttemans
 1	Tham
 1	Yorwba
 1	Yuriy Obukh
 1	abratchik
 1	catree
 1	goldstar616
 1	logic1988
 1	nyanp
 1	oqtvs
 1	saskatchewancatch
 1	szk1509
 1	utibenkei


December, 2016

Long-awaited update to OpenCV 3.x release series, with tons of improvements and bug fixes. Big thanks to everyone who particpated! If you contributed something but your name is missing, please, let us know.

  • Results from 11 GSoC 2016 projects have been submitted to the library, 9 of them have been integrated already, 2 still pending (the numbers below are the id’s of the Pull Requests in opencv or opencv_contrib repository):

    • Ambroise Moreau (Delia Passalacqua) – sinusoidal patterns for structured light and phase unwrapping module (711)
    • Alexander Bokov (Maksim Shabunin) – DIS optical flow (excellent dense optical flow algorithm that is both significantly better and significantly faster than Farneback’s algorithm – our baseline), and learning-based color constancy algorithms implementation (689, 708, 722, 736, 745, 747)
    • Tyan Vladimir (Antonella Cascitelli) – CNN based tracking algorithm (GOTURN) (718, 899)
    • Vladislav Samsonov (Ethan Rublee) – PCAFlow and Global Patch Collider algorithms implementation (710, 752)
    • João Cartucho (Vincent Rabaud) – Multi-language OpenCV Tutorials in Python, C++ and Java (7041)
    • Jiri Horner (Bo Li) – New camera model and parallel processing for stitching pipeline (6933)
    • Vitaliy Lyudvichenko (Anatoly Baksheev) – Optimizations and improvements of dnn module (707, 750)
    • Iric Wu (Vadim Pisarevsky) – Base64 and JSON support for file storage (6697, 6949, 7088). Use names like “myfilestorage.xml?base64” when writing file storage to store big chunks of numerical data in base64-encoded form.
    • Edgar Riba (Manuele Tamburrano, Stefano Fabri) – tiny_dnn improvements and integration (720: pending)
    • Yida Wang (Manuele Tamburrano, Stefano Fabri) – Quantization and semantic saliency detection with tiny_dnn
    • Anguelos Nicolaou (Lluis Gomez) – Word-spotting CNN based algorithm (761: pending)

    big thanks to all the participants!

  • There have been many contributions besides GSoC (big thanks to the authors, which names/nicknames are enclosed in parentheses):
    • DICOM support via GDCM library (look4pritam), PAM format support (DmtKats)
    • Support GigE cameras via Aravis SDK (thanks to ArkadiuszRaj)
    • Interactive camera calibration app, see, and QR matrix decomposition (Vlad Sovrasov)
    • OpenCL accelerations exposed to Python (Nikolay Polyarniy)
    • Instrumentation framework for performance testing (Pavel Vlasov)
    • Many improvements in tests (Maksim Shabunin, Alexander Alekhin and many other people)
    • Weighted median filter (Zhou Chao)
    • VGG descriptor, BoostDesc descriptor (Balint Cristian)
    • Selective search segmentation (Maximilien Cuony)
    • Paillou and Deriche Filter (Laurent Berger)
    • Bilateral texture filter (jhlee525)
    • BIF descriptor (Alexey Spizhevoy)
    • BM3D noise reduction (Bartek Pawlik)
    • Fine Grained Saliency (Sebastian Montabone)
    • Position-Color-Texture signatures extractor (Gregor Kovalčík)
    • Thinning algorithms (Steven Puttemans)
    • Fast Line Detector (Jin Han Lee)
    • Freetype2 support (@Kumataro)
    • etc.

  • There are also many updates to make sure that OpenCV compiles and runs on modern OSes with modern compilers:
    • Ubuntu 16.04 LTS with GCC 5.×. GCC 6.x is supported too.
    • OSX 10.11 – 10.12 with XCode 8.x; in particular, we now use the new AVFoundation backend on OSX. Thanks to yoffy, Matthew Self and Ryan Govostes.
    • Visual Studio 2015; VS 2017 will likely work too.
    • Better support for AArch64 on Linux and Android

  • Intel MKL can now be used to accelerate SVD, LU, camera calibration, as well as DNN module.
  • Added initial support for Intel Compiler (Intel Compiler 2017 has been tested on Windows and Linux)
  • Many SSE optimizations have been done; big thanks to K-Shinotsuka, Tomoaki Teshima, mschoeneck and many other people.
  • Several OpenCL kernels, specially optimized for Intel GPUs, have been contributed by Peng Li and others.


  • OpenVX 1.x implementations can now be used to accelerate OpenCV, mostly through the HAL mechanism. Build OpenCV with WITH_OPENVX flag turned on and specifty where to find OpenVX (see for details).
  • Convenient C++ wrappers for OpenVX that simplify joint use of OpenCV + OpenVX have been created by Andrey Pavlenko. See the examples at, the wrappers are in ivx.hpp header.
  • External BLAS+Lapack implementations (such as MKL, OpenBLAS, Atlas or Apple’s Accelerate framework) can now be used to accelerate SVD decomposition, LU decomposition etc.
  • HAL has been extended to support much more functions such as image warping, color conversion, filtering operations.

  • This release features a record number of PRs and bug fixed:
    • 778 + 191 PRs merged into opencv + opencv_contrib master since 3.1.0
    • 348 bugs and 69 features in opencv master
    • 47 bugs and 14 features in opencv_contrib



git shortlog --no-merges -ns 3.1.0..3.2.0
   244  Alexander Alekhin
    63  Vitaly Tuzov
    57  Tomoaki Teshima
    54  Maksim Shabunin
    51  Vladislav Sovrasov
    46  Ilya Lavrenov
    40  Rostislav Vasilikhin
    33  k-shinotsuka
    30  StevenPuttemans
    29  Pavel Rojtberg
    24  MYLS
    18  Marina Noskova
    17  catree
    15  Alexander Smorkalov
    14  Suleyman TURKMEN
    13  Vladislav Vinogradov
    13  apavlenko
    13  berak
    12  Alexandr Kondratev
    11  mshabunin
    10  Dan
    10  Pavel Vlasov
    10  Valeriy Van
    10  atinfinity
    10  matze
     9  Li Peng
     9  abratchik
     8  Roman Donchenko
     8  Vadim Pisarevsky
     7  Aleksandar Atanasov
     7  Philipp Hasper
     6  ArkadiuszRaj
     6  Francisco Facioni
     6  LaurentBerger
     6  sourin
     5  E Sommerlade
     5  Eric Zavesky
     5  Ishank gulati
     5  Matthew Self
     5  Philip
     5  Tetragramm
     4  Amro
     4  Andrey Pavlenko
     4  Arthur Cinader
     4  Christoph Spörk
     4  Dikay900
     4  Marian Zajko
     4  alcinos
     4  daB0bby
     4  filipsladek
     4  ohnozzy
     3  Alexander Karsakov
     3  Andreas Franek
     3  Arek
     3  Arkadiusz Raj
     3  Dimitrios Katsaros
     3  Eric Sommerlade
     3  Gilles Rochefort
     3  Juha Reunanen
     3  Marek Smigielski
     3  Mathieu Barnachon
     3  Matthew Skolaut
     3  Philippe FOUBERT
     3  Sergei Nosov
     3  Vitaliy Lyudvichenko
     3  Your Name
     3  a-andre
     3  anirudt
     3  aravind
     3  kurenai
     3  panxiaochun
     3  shruthikashyap
     3  thierry
     3  will hunt
     3  Прун Виктор
     2  Addison Elliott
     2  Alexander Shishkov
     2  Anatoly Orlov
     2  Asal Mirzaieva
     2  Bleach666
     2  Bob Paulin
     2  Cartucho
     2  Dan Moodie
     2  David Wolever
     2  Declan Moran
     2  Dmitry Trifonov
     2  DozyC
     2  Elad Joseph
     2  Elif Albuz
     2  Eric Liao
     2  Erich Keane
     2  Hajo Nils Krabbenhöft
     2  Hannes
     2  Jan Starzynski
     2  Joe Howse
     2  Jose Luis Guardiola
     2  Josh Lubawy
     2  Lucas Solomon
     2  Naba Kumar
     2  Patrick Snape
     2  Patrik Huber
     2  Peter Rekdal Sunde
     2  Piotr Semenov
     2  Quentin
     2  Randy J. Ray
     2  Robbert Klarenbeek
     2  Rok Mandeljc
     2  Sean McBride
     2  Sergiu Deitsch
     2  Steven Puttemans
     2  Susmit
     2  Teng Cao
     2  emiswelt
     2  jiakai
     2  lewgate
     2  mschoeneck
     2  shengyu
     2  takacsd
     1  Aaron Simmons
     1  Adam Rankin
     1  Alexander Fedorov
     1  AlexeyAB
     1  Ankit Yadav
     1  Anton V. Shokurov
     1  Anush Elangovan
     1  Artyom Lebedev
     1  Balint Cristian
     1  Bernd Kuhls
     1  BinMatrix
     1  Bleach
     1  Bram Ton
     1  CSBVision
     1  Christian Kapeller
     1  Colin Finck
     1  Dmitry Mozgin
     1  Dmitry-Me
     1  Dustin Spicuzza
     1  Eduardo Arnold
     1  Elena Gvozdeva
     1  Elena Shipunova
     1  Emanuel Jöbstl
     1  Eugene Khvedchenya
     1  Evan Heidtmann
     1  Feng Zhoutian
     1  Finfa811
     1  Fraser Harris
     1  Graham Fyffe
     1  Gregory Kramida
     1  Gregory Morse
     1  Guillaume Jacob
     1  Guy Sheffer
     1  HMM@T430s-x64
     1  Han Hu
     1  Ian MacLeod
     1  Ivar Clemens
     1  JR Heard
     1  Jake Kesinger
     1  Jason von Nieda
     1  Jcrist99
     1  Jeff McBride
     1  Jeroen Kromwijk
     1  Jiri Horner
     1  Josep Bosch
     1  Juan Pablo Aveggio
     1  Julien Dehos
     1  Julien Dubiel
     1  Julien Nabet
     1  Kai Hugo Hustoft Endresen
     1  Kerry Billingham
     1  Kevin Hughes
     1  Kevin, Hu
     1  KeySecure
     1  Kotrix
     1  Krishnaraj Bhat
     1  Leonardo Brás
     1  Leonid Volnitsky
     1  Louis Letourneau
     1  Luis Roberto Morales Iglesias
     1  LukeZhu
     1  Matt Leotta
     1  Matthew Cook
     1  Matthieu FT
     1  Matěj Šmíd
     1  Michael Shtutman
     1  Morrison
     1  Nikolay Polyarniy
     1  Noel Eck
     1  Oded Green
     1  Oliver Schreer
     1  Oskari Vinko
     1  Patrick Spettel
     1  Peter Entwistle
     1  Philipp Nordhus
     1  Philippe Renon
     1  Pierre Dulac
     1  Pranav Ganorkar
     1  Prayag Verma
     1  Péter Trombitás
     1  Rahul Chhabra
     1  S. Garrido
     1  Sean Li
     1  Sergey Alexandrov
     1  Simmo Saan
     1  Simon Hänisch
     1  Sovrasov
     1  Stephane Thiell
     1  Steven E Wright
     1  Tauranis
     1  Thierry Fauck
     1  Thomas Schenker
     1  Tomi
     1  Tsukasa Sugiura
     1  Vadzim Piatrou
     1  Vincent Rabaud
     1  WANG Guoqin
     1  Xiaolei Yu
     1  Zhenqing, Hu
     1  Zhigang Gong
     1  ahb
     1  arybnikov
     1  bedbad
     1  chacha21
     1  danil-tolkachev
     1  debjan
     1  dharezlak
     1  fegorsch
     1  ggargov
     1  guoxuedong
     1  jacobdang
     1  joao.faro
     1  logic1988
     1  look4pritam
     1  monoid911
     1  mpkh
     1  mrquorr
     1  mvukad
     1  parismita
     1  qiao
     1  renu285
     1  rishirajsurti
     1  robertxwu
     1  sarthak
     1  solosuper
     1  stbnps
     1  themightyoarfish
     1  viodetools
     1  yigitsoy
     1  yoffy
     1  Özgür Can


git shortlog --no-merges -ns 3.1.0..3.2.0
    43  Vitaliy Lyudvichenko
    30  Alexander Alekhin
    29  Anna Petrovicheva
    25  Vladislav Samsonov
    25  Vladislav Sovrasov
    21  Gary Bradski
    15  Pavel Rojtberg
    12  berak
    10  Balint Cristian
     5  Maksim Shabunin
     5  Suleyman TURKMEN
     5  apetrovichev
     5  lluisgomez
     4  Alexander Bokov
     4  Amro
     4  StevenPuttemans
     4  mshabunin
     3  Jinhan Lee
     3  Patrick Snape
     3  Rok Mandeljc
     3  Steve Nicholson
     3  folz
     3  vbystricky
     2  LaurentBerger
     2  Maximilien Cuony
     2  Prof. Dr. Rudolf Haussmann
     2  Vadim Pisarevsky
     2  Vitaly Tuzov
     2  Yuki Furuta
     2  catree
     2  sbokov
     1  Aleksandr Rybnikov
     1  Alexey Spizhevoy
     1  AmbroiseMoreau
     1  Amy Winston
     1  Andrea Merello
     1  Bartek Pawlik
     1  Bleach
     1  Bleach666
     1  Christian Richardt
     1  Daniel Ching
     1  Dikay900
     1  Eric Zavesky
     1  Eugene Mamin
     1  Geoff
     1  Gregor Kovalcik
     1  Gregory Kramida
     1  Gunshi Gupta
     1  Han Hu
     1  Konstantin Volokh
     1  Kotrix
     1  Kumataro
     1  Lorena García
     1  MambaWong
     1  Marek Smigielski
     1  Martin Nowak
     1  Minh Ngo
     1  Nicu Stiurca
     1  Pavel Vlasanek
     1  Rostislav Vasilikhin
     1  Sarthak Ahuja
     1  Sunglok Choi
     1  Thomas Schulz
     1  VBystricky
     1  Vasilikhin
     1  Vladimir
     1  Wu Jiang
     1  Zhou Chao
     1  abratchik
     1  alessiodore
     1  andrewk-mccann
     1  aravind
     1  baisheng lai
     1  comdiv
     1  damir
     1  edgarriba
     1  jasjuang
     1  jhlee525
     1  lukaszi89
     1  lych77
     1  samontab
     1  sht3ch
     1  tharunkalwa


December, 2015

This is the first stabilization update in 3.x series. It shall be reminded that since OpenCV 3.0 we’ve changed the version enumeration scheme, so that 3.1 is the same sort of update to 3.0 as 2.4.1 was to 2.4.0.

  • There have been many successful projects this year (student and mentor(s) are listed after the project name), and the results are available as a part of OpenCV 3.1 (partially in the main repository, but mostly opencv_contrib):
    • Omnidirectional Cameras Calibration and Stereo 3D Reconstruction – opencv_contrib/ccalib module (Baisheng Lai, Bo Li)
    • Structure From Motion – opencv_contrib/sfm module (Edgar Riba, Vincent Rabaud)
    • Improved Deformable Part-based Models – opencv_contrib/dpm module (Jiaolong Xu, Bence Magyar)
    • Real-time Multi-object Tracking using Kernelized Correlation Filter – opencv_contrib/tracking module (Laksono Kurnianggoro, Fernando J. Iglesias Garcia)
    • Improved and expanded Scene Text Detection – opencv_contrib/text module (Lluis Gomez, Vadim Pisarevsky)
    • Stereo correspondence improvements – opencv_contrib/stereo module (Mircea Paul Muresan, Sergei Nosov)
    • Structured-Light System Calibration – opencv_contrib/structured_light (Roberta Ravanelli, Delia Passalacqua, Stefano Fabri, Claudia Rapuano)
    • Chessboard+ArUco for camera calibration – opencv_contrib/aruco (Sergio Garrido, Prasanna, Gary Bradski)
    • Implementation of universal interface for deep neural network frameworks – opencv_contrib/dnn module (Vitaliy Lyudvichenko, Anatoly Baksheev)
    • Recent advances in edge-aware filtering, improved SGBM stereo algorithm – opencv/calib3d and opencv_contrib/ximgproc (Alexander Bokov, Maksim Shabunin)
    • Improved ICF detector, waldboost implementation – opencv_contrib/xobjdetect (Vlad Shakhuro, Alexander Bovyrin)
    • Multi-target TLD tracking – opencv_contrib/tracking module (Vladimir Tyan, Antonella Cascitelli)
    • 3D pose estimation using CNNs – opencv_contrib/cnn_3dobj (Yida Wang, Manuele Tamburrano, Stefano Fabri)

  • Besides GSoC, we have multiple contributions from the community; below is the partial list of contributed functionality:
    • plot module – Nuno Moutinho
    • ni-black thresholding algorithm – ximgproc, Samyak Datta
    • Superpixel segmentation using Linear Spectral Clustering, SLIC superpixels – ximgproc, Balint Cristian
    • HDF (HDF5) support module – Balint Cristian
    • Depth to external RGB camera registration – rgbd, Pat O’Keefe
    • Computing normals for a point cloud – rgbd, Félix Martel-Denis
    • Fuzzy image processing module – Pavel Vlasanek
    • Rolling guidance filter – ximgproc, Zhou Chao
    • 3x faster SimpleFlow – optflow, Francisco Facioni
    • Code and docs for CVPR 15’ paper “DNNs are easily fooled” – Anh Nguyen
    • Efficient Graph-based image segmentation algorithm – ximgproc, Maximilien Cuony
    • Sparse-to-dense optical flow – optflow, Sergey Bokov
    • Unscented Kalman Filter (UKF) and Augmented UKF – tracking, Svetlana Filicheva
    • Fast Hough Transform – ximgproc, xolodilnik
    • Improved performance of haartraining – Teng Cao
    • Python samples made compatible with Python 3 – bastelflp

  • The iOS framework (opencv2.framework) can now be configured to include both opencv and opencv_contrib functionality. Previously, there were 2 separate frameworks, but because of the include directory names (“opencv2/…”, regardless of whether some module is from opencv or from opencv_contrib), the produced opencv_contrib was unusable.

  • more efficient camera support on Android 5+
  • faster round() on ARM (it’s also applicable to iOS); big thanks to Manuele Tamburano and Stefano Fabri for this!


  • OpenCV 3.1 supports fresh OSes from Microsoft and Apple, as well as the newest development tools (VS2015 and Xcode 7, respectively).

  • IPPICV (a specially-for-opencv free-of-charge subset of IPP ( that has been kindly provided by Intel Corporation) is now based on IPP 9.0.1, which should make OpenCV even faster on modern Intel chips.

  • There are multiple new features in the OpenCL layer, resulted from our collaboration with Intel Corporation
    • Improved/extended interoperability with DirectX 9, 10, 11. That includes conversions between DX Surface and UMat, with number of copying operations reduced where possible. See opencv/samples/directx.
    • Interoperability with Video Acceleration API on Linux. This feature requires Intel Media Server Studio ( and uses Intel’s cl_intel_va_api_media_sharing extension to OpenCL. See opencv/samples/va_intel
    • OpenCL<=>OpenGL interoperability via cl_khr_gl_sharing extension. It’s possible now to convert OpenGL textures and OpenGL buffers to/from UMat. See opencv/samples/opengl.
    • OpenCL<=>OpenCL interoperability. It means, people can now use OpenCL-accelerated OpenCV together with their custom OpenCL code and/or other OpenCL libraries. See opencv/samples/opencl for details.
    • Support for Camera2 API in Android 5+. See opencv/samples/android/tutorial-4-opencl, the tutorial that demonstrates 3 ways of processing data from camera:
      • purely CPU pipeline (3-4 fps on our test device)
      • OpenCV-based using UMat and transparent API (10-13 fps on the same device)
      • direct use of OpenCL (28-30fps). This is the fastest method, as you can see, because it uses OpenCL images, whereas more universal T-API has to use OpenCL buffers.

  • HAL (Hardware Acceleration Layer), as a separate OpenCV module, opencv_hal, lived a short yet a bright life; we decided to split it by parts and move the parts into the respective regular OpenCV modules (core, imgproc etc.). It will help us to keep the dependency graph clean and sane.
  • Added 2 examples (see opencv/samples/hal) on how to provide external accelerated versions of the HAL functions; note that an external implementation does not have to implement all of the HAL. The supplied README explains how to build the custom HAL accelerator, and how to build OpenCV with such a custom HAL replacement:
  • Added some more few functions to the HAL (arithmetic operations, split, merge); much more to come in OpenCV 3.2

  • Over 420 pull requests to opencv and over 100 pull requests to opencv_contrib have been merged in since 3.0.0. Big thanks to all the people who submitted bug fixes, new functionality and other small and big improvements! Special thanks goes to all the GSoC 2015 students and mentors, who created a lot of great functionality!



git shortlog --no-merges -ns 3.0.0..3.1.0
   146  Vladislav Vinogradov
   109  Alexander Alekhin
   103  Alexander Smorkalov
    73  Maksim Shabunin
    68  Ilya Lavrenov
    48  Roman Donchenko
    47  StevenPuttemans
    45  Pavel Rojtberg
    28  Andrey Pavlenko
    25  Seon-Wook Park
    20  Pavel Vlasov
    20  Suleyman TURKMEN
    18  Viet Dinh
    17  Alexey Ershov
    15  berak
    14  Vladimir Dudnik
    13  Artur Wieczorek
    12  Dan
    12  Lorena García
    10  Dikay900
    10  ippei ito
     9  Mansour Moufid
     8  E Braun
     8  Grigory Serebryakov
     7  Alexander Karsakov
     7  Alexander Nitsch
     7  Vadim Pisarevsky
     7  Vijay Pradeep
     6  LaurentBerger
     6  Marian Zajko
     6  Nguyen Nguyen
     6  Philipp Hasper
     6  flp
     6  sbokov
     5  Andreas Stührk
     5  PhilLab
     5  Rok Mandeljc
     5  Ruslan Baratov
     5  askourik
     5  songyuncen
     4  Abe Friesen
     4  Aman Verma
     4  Dan Moodie
     4  Dinar Ahmatnurov
     4  Dominik Kleiser
     4  Elad Joseph
     4  Gleb Gladilov
     4  Ishant Mrinal Haloi
     4  Mike Maraya
     4  Tom Stellard
     4  Vitaly Tuzov
     4  edgarriba
     4  grundman
     4  hoangviet1985
     4  themightyoarfish
     3  Adam Gibson
     3  Andrey Kamaev
     3  Camille
     3  Cédric Rousseau
     3  Elena Shipunova
     3  Gregory Morse
     3  Hajo Nils Krabbenhöft
     3  Igor Kuzmin
     3  Maxim Kostin
     3  Renato Florentino Garcia
     3  Richard Yoo
     3  Samuel Martin
     3  Seunghoon Park
     3  Teng Cao
     3  Yan Wang
     3  a-andre
     3  jiakai
     3  unknown
     2  Aleksandr Petrikov
     2  Alex Torres
     2  Alexander Duda
     2  Anatoly Baksheev
     2  Andrew Senin
     2  Anton V. Shokurov
     2  Ariel Bernal
     2  Balint Cristian
     2  Ben Hagen
     2  Bruno Goncalves
     2  Daniel Stonier
     2  Dmitry Melnichenko
     2  Enes Ateş
     2  Evgeny Talanin
     2  GregoryMorse
     2  Jan S. (Milania1)
     2  John Stowers
     2  Leonid Beynenson
     2  Luis Díaz Más
     2  Nils Plath
     2  Nisarg Thakkar
     2  Philippe FOUBERT
     2  Pierre Letessier
     2  Prof. Dr. Rudolf Haussmann
     2  Rodrigo Benenson
     2  Sergey Alexandrov
     2  Sergiu Dotenco
     2  Tsukasa Sugiura
     2  Vicente Olivert Riera
     2  Vincent Rabaud
     2  Vladimir Kolesnikov
     2  WilhelmHannemann
     2  howtobeahacker
     2  jayceelock
     2  jisli
     2  kalistratovag
     2  lujia
     2  qubick
     2  robertxwu
     2  theodore
     1  Aaron Simmons
     1  Abhijit Kundu
     1  Adam
     1  Adam Geitgey
     1  Adi Shavit
     1  Adrian Clark
     1  AfterEight
     1  Alankar Kotwal
     1  Alexander Petrikov
     1  Alexander Stohr
     1  AlexanderStohr
     1  Amro
     1  Andreas Franek
     1  Antonella Cascitelli
     1  Anush Elangovan
     1  Ashley Stacey
     1  Ashod Nakashian
     1  Belev
     1  Benoit Seguin
     1  Bert
     1  Boris Daskalov
     1  Bovaz
     1  Brian Park
     1  Chris Kitching
     1  Chris Mayo
     1  Christian Richardt
     1  ComFreek
     1  Craig Reynolds
     1  Deanna Hood
     1  Dmitry Budnikov
     1  Dmitry Gribanov
     1  Drew Jetter
     1  Ehren Metcalfe
     1  Eric Sommerlade
     1  Evgeny Agafonchikov
     1  Firat Kalaycilar
     1  Florent
     1  GabrieleDalmazzone
     1  HelenWong
     1  Hernan Badino
     1  Ignas Daukšas
     1  Ishank gulati
     1  Jaime Fernandez
     1  Jan Starzynski
     1  Jay Bosamiya
     1  Jeremy Huddleston Sequoia
     1  Jie Yang
     1  Joe Howse
     1  Jonas Vautherin
     1  Josh Klontz
     1  Jüri Aedla
     1  Kauê de Moraes Vestena
     1  Kazuki Matsuda
     1  Keita Kita
     1  Kevin Hughes
     1  Lectem
     1  LeonidBeynenson
     1  Leszek Swirski
     1  Manfred Georg
     1  ManuelFreudenreich
     1  Martijn Liem
     1  Martin Ueding
     1  Mat Moore
     1  Matteo Piovanelli
     1  Maximilien Cuony
     1  Miguel Algaba
     1  Milania1
     1  Nghia Ho
     1  Nickolay Polyarniy
     1  Niklas Frisk
     1  Nikolay Polyarniy
     1  Oliver Lumby
     1  Pastafarianist
     1  Patrik Huber
     1  Philip Salvaggio
     1  Pierre Moulon
     1  Pierre-Emmanuel Viel
     1  Raaj
     1  RafaRuiz
     1  Rafael Sadowski
     1  René
     1  Robbert Klarenbeek
     1  Rohit Rawat
     1  S. Garrido
     1  Samyak Datta
     1  Sethur
     1  Simon Heinen
     1  Stefan Cornelius
     1  Stephen Mell
     1  Steven Puttemans
     1  Thomas Dunker
     1  Thomas Gabrowski
     1  Thomas Weisbach
     1  Tian Zhi
     1  Tim D. Smith
     1  Valentin Tsatskin
     1  ValeryTyumen
     1  Vipul Sharma
     1  Vitaliy Lyudvichenko
     1  Wangyida
     1  Wenju He
     1  Yang Fan
     1  Ying Xiong
     1  Youssef Kashef
     1  aletheios
     1  alexander barakin (aka sash-kan)
     1  art-programmer
     1  ausentso
     1  bhack
     1  boatx
     1  brossetti
     1  comdiv
     1  doltac
     1  emiswelt
     1  greensea
     1  jormansa
     1  manuele
     1  mattip
     1  mbarann
     1  micalan
     1  michaelveth
     1  niederb
     1  quentin
     1  rajithr
     1  scorneli
     1  spmallick
     1  teng88
     1  thebucc
     1  ttagu99
     1  ultrafro
     1  侯骥


git shortlog --no-merges -ns 3.0.0..3.1.0
   169  Vladimir
    48  Kurnianggoro
    29  Wangyida
    19  Muresan Mircea Paul
    18  lluis
    17  baisheng lai
    16  Maksim Shabunin
    14  Anh Nguyen
    13  Lluis Gomez-Bigorda
     9  Seon-Wook Park
     9  Vlad Shakhuro
     9  nmoutinho
     8  Oded Green
     8  Patrick Snape
     8  Zhou Chao
     8  berak
     6  xolodilnik
     5  Baisheng Lai
     5  Marc-Stefan Cassola
     5  Pavel Rojtberg
     5  felix
     5  sbokov
     4  Alexander Alekhin
     4  Alexander Stohr
     4  StevenPuttemans
     4  edgarriba
     3  Balint Cristian
     3  Francisco Facioni
     3  Maximilien Cuony
     3  jiaolong_x220
     2  Aakash Chopra
     2  Jan S. (Milania1)
     2  RobertaRavanelli
     2  S. Garrido
     2  Samyak Datta
     2  Vadim Pisarevsky
     2  Yang Fan
     1  103yiran
     1  Alex Tomala
     1  Clint Pells
     1  Dikay900
     1  DirtyMaster
     1  Hyunjun Kim
     1  Jesus Briales
     1  Jesús Briales
     1  MMp131316
     1  Pat O'Keefe
     1  Philipp Hasper
     1  Rok Mandeljc
     1  Suleyman TURKMEN
     1  Sunglok Choi
     1  SvetlanaFilicheva
     1  Vitaliy Lyudvichenko
     1  Yan Wang
     1  baudenri
     1  comdiv
     1  grundman
     1  howtobeahacker
     1  lar-lab-unibo
     1  m4nh
     1  manuele
     1  previ
     1  tucna


June, 2015

These are changes since 3.0 rc1.

  • Over 200 bugs reported at have been resolved/closed. It includes many fixes in docs, build scripts, python wrappers, core, imgproc, photo, features2d, objdetect, contrib modules, as well as some performance improvements etc.

  • Added pthreads-based backend for cv::parallel_for_. It means that all the parallel processing should be available out-of-the-box on any POSIX-compatible OS, including QNX and such. You are welcome to try it out.

  • OpenCV Manager on Android has been completely rewritten in Java and now supports both OpenCV 2.4 and 3.0. After a bit more testing we will put it to Google Play.
  • The other, a bit sad but inevitable news – we had to drop support for nativecamera on Android. It used some undocumented API and never worked stable.

  • Many excellent contributions in opencv_contrib/xfeatures2d:
    • Christian Balint contributed long-awaited DAISY features
    • He also contributed improved version of rotation-invariant BRISK descriptor by Gil Levi
    • Gil Levi contributed very nice LATCH binary descriptor that outperforms all presently available in OpenCV binary descriptors (ORB, BRIEF, AKAZE etc.) and is comparable with heavier and slower SURF.

  • Over 80 pull requests have been merged in since 3.0 rc. Big thanks to all the people who submitted bug fixes, new functionality and other small and big improvements! This time special thanks goes to Philip (Dikay900) who did (and still does) excellent job on porting various bug fixes and other small improvements from 2.4 to master/3.0. Thank you, Philip!



$ git shortlog —no-merges -ns 3.0.0-rc1..origin/master
60  Vadim Pisarevsky
31  laurentBerger
17  Maksim Shabunin
10  cbalint13
7  Maxim Kostin
6  Alexander Smorkalov
6  Ilya Lavrenov
6  StevenPuttemans
6  Vladimir Dudnik
5  Andrey Pavlenko
5  GilLevi
3  Dikay900
3  Philipp Hasper
3  Ruslan Baratov
3  rajithr
3  sgjava
2  Andreas Stührk
2  Christian Richardt
2  Deanna Hood
2  Joe Minichino
2  Luis Díaz Más
2  Mansour Moufid
2  Marek Nogacki
2  Martin Ueding
2  Vladislav Vinogradov
2  Zhigang Gong
2  kalistratovag
1  AfterEight
1  Aleksandr Petrikov
1  Alexander Nitsch
1  Andrew Senin
1  Ashod Nakashian
1  Belev
1  Boris Daskalov
1  Camille
1  Craig Reynolds
1  David Hirvonen
1  Ehren Metcalfe
1  Evgeny Agafonchikov
1  Jüri Aedla
1  Leonid Beynenson
1  Martijn Liem
1  Marvin Smith
1  Michaël Sdika
1  Mike Maraya
1  Milania1
1  Niklas Frisk
1  Owen Healy
1  Patrik Huber
1  Piotr Dobrowolski
1  Roman Donchenko
1  S. Garrido
1  Samed Ozdemir
1  Samuel Martin
1  Samyak Datta
1  Thomas Gabrowski
1  Vicente Olivert Riera
1  Vincent Rabaud
1  Vladimir Kolesnikov
1  aletheios
1  berak
1  jormansa
1  lujia
1  msandler
1  yanlend
1  zenev


$ git shortlog —no-merges -ns 3.0.0-rc1..origin/master
19  cbalint13
11  GilLevi
6  Alexandre Benoit
4  Vadim Pisarevsky
2  laurentBerger
1  berak
1  root

version:3.0 rc

April, 2015

These are changes since 3.0 beta.

  • Improved compatibility with OpenCV 2.4:
    • 2.4.11 now includes “opencv2/core.hpp” and other such headers in addition to standard “opencv2/core/core.hpp”.
    • smart pointers (Ptr<>) can now be created in both 2.4 and 3.0 style (new ClassName(params) vs makePtr(params))
    • trained and stored stat models from opencv_ml 2.4 can now be loaded and used by opencv_ml 3.0 as-is.
    • the 2.4=>3.0 transition guide has been sketched:

  • We did not put OpenCV 3 to Google Play yet, but we prepared installable OpenCV 3.0 Manager for Android, which can be installed and tried on your device. It can co-exist with OpenCV 2.4 Manager on the same device.

  • There are multiple improvements and bug-fixes for WinRT port (as well as Windows 8.x port) of OpenCV by Microsoft guys (big thanks to Max Kostin!)
    • In particular, parallel_for is enabled on WinRT, so the code should run much faster on multi-core devices
    • Also, the WMF video capturing backend has been greatly improved.

  • Standalone motion jpeg codec has been added to opencv_videoio. It does not need ffmpeg or any other 3rd-party lib. According to our measurements, it’s also much faster than ffmpeg, especially on ARM. For the decoder you should have JPEG support enabled (through built-in or external libjpeg). How to use it? To encode motion jpeg video, use .avi file extension and CV_FOURCC(‘M’, ‘J’, ‘P’, ‘G’). The decoder part has been verified on such streams (avi files with index, where each frame is encoded using baseline jpeg) and few random motion jpeg clips from net, but we have not tested it thoroughly.

  • Preliminary version of OpenCV HAL, low-level acceleration API beneath OpenCV, has been introduced. Currently it includes just a few math functions, but will grow soon. It also includes so-called “universal intrinsics”, inspired by NEON=>SSE conversion header by Victoria Zhislina from Intel: The idea is that one can use a single SIMD code branch that will compile to either SSE or NEON instructions depending on the target platform. For example,

    // a, b and c are floating-point arrays
    for( int i = 0; i < n; i+=4 )
        v_store(c + i, v_load(a+i) + v_load(b+i));

    will be expanded to either

    for( int i = 0; i < n; i+=4 )
        _mm_storeu_ps(c + i, mm_add_ps(_mm_loadups(a+i), mm_loadu_ps(b+i));


    for( int i = 0; i < n; i+=4 )
        vst1q_f32(c + i, vaddq_f32(vld1q_f32(a+i), vld1qf32(b+i));

    Using such intrinsics one can write accelerated code, debug it on desktop and then run it without any changes on ARM and get reasonable performance.

  • Over 500 pull requests have been merged in since 3.0 beta. Big thanks to all the people who submitted bug fixes, new functionality and other small and big improvements! OpenCV becomes more and more a community-driven project and we are pleased to see that!



$ git shortlog —no-merges -ns 3.0.0-beta..origin/master
95  Vladislav Vinogradov
88  Maksim Shabunin
74  Ilya Lavrenov
40  Erik Karlsson
33  Olexa Bilaniuk
28  Vadim Pisarevsky
24  Dmitry-Me
24  orestis
18  StevenPuttemans
17  Maxim Kostin
15  Yan Wang
14  Alexander Alekhin
11  Roman Donchenko
8  Ashod Nakashian
7  Joe Howse
7  Nisarg Thakkar
6  Andrey Kamaev
6  Dikay900
5  Alexander Nitsch
5  Evgeny Agafonchikov
5  TobyWanKenobi
5  theodore
5  Сергей Нужный
4  Andrew Senin
4  Artur Wieczorek
4  Hamid Bazargani
4  Pavel Rojtberg
4  berak
3  Alexander Smorkalov
3  Ben Hagen
3  Deanna Hood
3  Marvin Smith
3  Nicu Știurcă
3  Philipp Hasper
3  Simon Heinen
3  Takahiro Poly Horikawa
3  Ying Xiong
2  Ahmatnurov Dinar
2  Alexander Duda
2  Ana Huaman Quispe
2  Eric Sommerlade
2  Naer Chang
2  Nils Plath
2  Sync-my-L2P
2  VBystricky
2  guopei
2  ippei ito
2  kalistratovag
2  karelknoest
2  sanuj
2  vbystricky
1  Adam Borowski
1  Adam Greig
1  Alexander Bohn / FI$H2k
1  Alexander Karsakov
1  Andrey Pavlenko
1  Ari Suwendi
1  Basheer Subei
1  Boaz Stolk
1  Chris Mayo
1  Chuanbo Weng
1  Claudio Caraffi
1  Daeyun Shin
1  Daniel Angelov
1  Daniel Sjölie
1  Dmitriy Anisimov
1  Florian Verdet
1  Guo Yejun
1  HelenWong
1  Jay Bosamiya
1  Jiri Drbalek
1  Jonathan Viney
1  Josep Bosch
1  Julien Nabet
1  Kai Wolf
1  Kevin Squire
1  LorenaGdL
1  Max Kostin
1  Michae Gharbi
1  Mike Izbicki
1  Nicolas Riebesel
1  Owen Healy
1  Pat O’Keefe
1  Pavel Vlasov
1  Prof. Dr. Rudolf Haussmann
1  Quentin Minster
1  RafaRuiz
1  Rok Mandeljc
1  S. Garrido
1  Samyak Datta
1  Sancho McCann
1  Scott Graybill
1  Shiquan Wang
1  Tim D. Smith
1  Timur Sattarov
1  Tzafrir
1  UQ Times
1  Yang Fan
1  burun
1  eltermann
1  exeltior
1  fhsm
1  greensea
1  hahne
1  mariadeanton
1  potpath
1  sergei
1  soyer
1  tcavallari
1  tt
1  ttagu99
1  vincentweb


$ git shortlog —no-merges -ns 3.0.0-beta..origin/master
13  Maksim Shabunin
5  cbalint13
3  Auron-X
3  Dikay900
3  Str3iber
2  Dmitriy Anisimov
2  Jay Bosamiya
2  Rok Mandeljc
2  Vadim Pisarevsky
2  berak
1  Alfonso Sanchez-Beato
1  Antonella Cascitelli
1  Beat Küng
1  Dinar Ahmatnurov
1  Nisarg Thakkar
1  Sh0
1  Shuda Li
1  StevenPuttemans
1  Vladimir
1  Yan Wang
1  jiaolong_cvc228
1  pmkalshetti
1  yifita

version:3.0 beta

November, 2014

These are changes since 3.0 alpha.

  • Performance of OpenCL-accelerated code on Intel Iris Graphics and Intel Iris Pro Graphics has been improved by 10%-230%. That includes faster Canny edge detector, Laplacian, StereoBM, BruteForceMatcher, BackgroundSubtractorMOG2, StitchingWarpers etc. New OpenCL kernels were implemented for several functions: HoughLines, HoughLinesP, cvtColor (selected modes), Gemm, DFT. Thanks Alexander Alekhin, Alexander Karsakov, Elena Gvozdeva, Vladimir Bystritsky, Sergey Sivolgin and other guys who did that!

  • About 40 commonly used image processing functions have been optimized using vector NEON instructions, so OpenCV 3.0 beta should run noticeably faster on modern ARM chips. Big thanks to Ilya Lavrenov for the amazing job!

  • Java bindings have been substantially improved. We now have bindings for the updated features2d, background subtractor classes, a part of opencv_contrib etc. Thanks to Maksim Shabunin for the improvements!

  • On x86 and x64 platforms OpenCV binaries include and use a subset of Intel® Integrated Performance Primitives (Intel® IPP) by default. OpenCV 3.0 beta includes a subset of Intel® IPP 8.2.1 with additional optimization for AVX2.

  • Over 200 pull requests have been merged in since 3.0 alpha. Big thanks to all the people who submitted bug fixes, new functionality and other small and big improvements! OpenCV becomes more and more a community-driven project and we are pleased to see that!



$ git shortlog —no-merges -ns 3.0.0-alpha..origin/master
129  edgarriba
71  Ilya Lavrenov
45  Alexander Karsakov
41  Adrien BAK
39  Rahul Kavi
37  Vadim Pisarevsky
27  Maksim Shabunin
20  Ernest Galbrun
15  Dmitriy Anisimov
14  Artur Wieczorek
13  Alexander Mordvintsev
12  Alexander Alekhin
9  vbystricky
9  Alexander Smorkalov
9  Rok Mandeljc
8  Elena Gvozdeva
8  Marc Rollins
7  Philipp Hasper
5  berak
5  ElenaGvozdeva
4  PhilLab
3  Juan Carlos Niebles
3  Alexander Duda
3  Alexander Shishkov
3  xantares
3  jaco
3  Brian Park
3  Chuanbo Weng
3  Igor Kuzmin
3  Michele Adduci
2  Andrey Pavlenko
2  Bence Magyar
2  Miroslav Benes
2  Shubhra Pandit
2  StevenPuttemans
2  abidrahmank
2  f-morozov
2  micha137
1  Jose-Luis Blanco-Claraco
1  Pat O’Keefe
1  Pavel Vlasov
1  John Kua
1  Huu Nguyen
1  Pierrick Koch
1  Poly Takahiro Horikawa
1  GregoryMorse
1  Felix Endres
1  Benoit Seguin
1  jeremy
1  Tomasz Posłuszny
1  U-KruchininD-ПК\KruchininD
1  VBystricky
1  Dmitry Gribanov
1  Yan Wang
1  Yusuke Suzuki
1  Anatoly Baksheev
1  bartaandras
1  Daniel Stonier
1  dkanafeev
1  Boaz Stolk
1  eltermann
1  Ling Wang@z77x
1  Manfred Georg
1  Leszek Swirski
1  Matthew Conway
1  Maximilien Cuony
1  Ariel Elkin


$ git shortlog —no-merges -ns e2fde8..origin/master
69  biagio montesano
55  Vlad Shakhuro
46  Alex Leontiev
36  Bellaktris
28  Vadim Pisarevsky
24  lluis
19  dmitriy.anisimov
18  vludv
11  Dmitriy Anisimov
10  Bence Magyar
10  Daniel Angelov
10  Alexander Shishkov
9  ahb
6  tpietruszka
5  Maksim Shabunin
5  Beat Küng
4  manuele
4  StevenPuttemans
4  Alexander Mordvintsev
4  vbystricky
3  Philipp Hasper
3  berak
3  Sancho McCann
2  Alexander Alekhin
2  Ilya Lavrenov
2  Alexandre Benoit
1  tolgabirdal
1  Shuda Li
1  Rok Mandeljc
1  VBystricky
1  Poly Takahiro Horikawa
1  Michele Adduci
1  Ernest Galbrun

version:3.0 alpha

August, 2014

This is brief version of the change log. There is also a separate page with detailed review of OpenCV_3.0

  • OpenCV 3.0 brings more GPU accelerated functions and makes it in much more convenient form than OpenCV 2.4.
    • The new technology is nick-named “Transparent API” and, in brief, is extension of classical OpenCV functions, such as cv::resize(), to use OpenCL underneath. See more details about it here: T-API.
    • Along with OpenCL code refactoring and Transparent API implementation OpenCL kernels were optimized for mainstream platforms, most notably for modern Intel chips (including Iris and Iris Pro) and AMD chips (such as Kaveri). More detailed results are to be provided later.

  • On x86 and x64 platforms OpenCV binaries include and use a subset of Intel® Integrated Performance Primitives (Intel® IPP) by default. Intel Corporation granted OpenCV Foundation and all our users the right to use those functions free of charge for both non-commercial and commercial use.

  • We now have a separate repository for community-contributed code It will help us to combine well-known maturity and stability of classical OpenCV functionality together with innovative algorithms and opens new opportunities for community to share computer vision code. For all the contributed code we provide the same QA infrastructure as for the main OpenCV repository. How to use it?
  • Grab a copy of contribution repository along with a copy of OpenCV 3.0 alpha or a fresh master snapshot.
  • Pass OPENCV_EXTRA_MODULES_PATH=<path_to_contrib_directory>/modules to CMake.
  • Documentation for the contributed modules is integrated into OpenCV doc tree:

  • We completely revised our QA infrastructure (big thanks to Alexander Alekhin), which now covers not only the main repository, but also the contrib repository as well:


  • OpenCV 3.0 and its contribution repository feature a lot of new functionality created by our GSoC students during the past 2 years, namely:
    • Text detection and recognition by Lluis Gomez and Stefano Fabri
    • HDR by Fedor Morozov
    • KAZE/A-KAZE by Eugene Khvedchenya, the algorithm author Pablo Alcantarilla and some improvements by F. Morozov.
    • Smart segmentation and edge-aware filters by Vitaly Lyudvichenko, Yuri Gitman, Alexander Shishkov and Alexander Mordvintsev
    • Car detection using Waldboost, ACF by Vlad Shakhuro and Nikita Manovich
    • TLD tracker and several common-use optimization algorithms by Alex Leontiev
    • Matlab bindings by Hilton Bristow, with support from Mathworks.
    • Greatly extended Python bindings, including Python 3 support, and several OpenCV+Python tutorials by Alexander Mordvintsev, Abid Rahman and others.
    • 3D Visualization using VTK by Ozan Tonkal and Anatoly Baksheev.
    • RGBD module by Vincent Rabaud
    • Line Segment Detector by Daniel Angelov
    • Many useful Computational Photography algorithms by Siddharth Kherada
    • Shape descriptors, matching and morphing shapes (shape module) by Juan Manuel Perez Rua and Ilya Lysenkov
    • Long-term tracking + saliency-based improvements (tracking module) by Antonella Cascitelli and Francesco Puja
    • Another good pose estimation algorithm and the tutorial on pose estimation by Edgar Riba and Alexander Shishkov
    • Line descriptors and matchers by Biagio Montesano and Manuele Tamburrano

  • There are many other contributions, besides GSoC programs. Here is incomplete list (PLEASE, mail to admin at if your name is missing here):
    • Myriads of improvements in various parts of the library by Steven Puttemans; thank you a lot, Steven!
    • Several NEON optimizations by Adrian Stratulat, Cody Rigney, Alexander Petrikov, Yury Gorbachev and others.
    • Fast foreach loop over cv::Mat by Kazuki Matsuda
    • Image alignment (ECC algorithm) by Georgios Evangelidis
    • GDAL image support by Marvin Smith
    • RGBD module by Vincent Rabaud
    • Fisheye camera model by Ilya Krylov
    • OSX framework build script by Eugene Khvedchenya
    • multiple FLANN improvements by Pierre-Emmanuel Viel
    • Improved WinRT support by Gregory Morse
    • Latent SVM Cascade by Evgeniy Kozhinov and NNSU team (awaiting integration)
    • Logistic regression by Rahul Kavi
    • Five-point pose estimation algorithm by Bo Li

Also, big thanks to all the Itseez guys who reviewed patches, watched bug reports and helped OpenCV in many other ways: Alexander Shishkov (who also maintains site), Andrey Pavlenko, Alexander Alekhin, Alexander Smorkalov, Roman Donchenko, Kirill Kornyakov, Andrey Kamaev, Sergey Sivolgin, Vladimir Bystritsky, Sergey Nosov, Nikita Manovich, Evgeniy Talanin, Elena Gvozdeva, Alexander Karsakov, Konstantin Matskevich, Ilya Lavrenov, Anna Kogan, Ivan Korolev, Dinar Ahmatnurov, Andrey Senin, Vlad Vinogradov, Alexey Spizhevoy, Anatoly Baksheev, Marina Kolpakova, Daniil Osokin, Leonid Beynenson, Dmitry Retinsky, Maria Dimashova, Ilya Lysenkov, Andrey Morozov, Victor Eruhimov, Alexander Bovyrin, Sergey Molinov, Gary Bradski, Vincent Rabaud, Harris Gasparakis, Stefano Fabri and many others.

  • Although OpenCV 3 can be viewed as refined OpenCV 2 and is similar to the latter, this new version is not completely backward-compatible with OpenCV 2. Here are some major changes:
    • Some functionality has been moved to opencv_contrib repository or just disappeared. After revision, we may put a part of disappeared functionality back to opencv (or opencv_contrib).

    • Note that among the moved functionality is Face Recognition (now a separate module opencv_contrib/face); SIFT, SURF, FREAK, Star feature detectors and descriptors (now a part of opencv_contrib/xfeatures2d); text detector (now a part of module opencv_contrib/text); matlab bindings (opencv_contrib/matlab); motion templates & simple flow (opencv_contrib/optflow).

    • In OpenCV 3 we write

      #include “opencv2/core.hpp”

      instead of

      #include “opencv2/core/core.hpp”

      The old method should also work.

    • ML module has been completely revised (and at the moment the functionality is not 100% match for OpenCV 2.4.x), you will need to update your code.

    • Similarly to ML, many high-level algorithms have been refactored to follow “open interface-hidden implementation” concept and this refactoring work will continue after alpha.

    • To lower the footprint, highgui module has been split into 3 parts: imgcodecs, videoio and highgui itself (thanks to Vladimir). You may need to update your code.

    • After massive functionality migration to opencv_contrib and other major changes some Python bindings may be missing. We will fix it by OpenCV 3.0 beta.


April, 2014

  • Several improvements in OpenCL optimizations (ocl::sumocl::countNonZeroocl::minMax, bitwise operationss, Haar face detector, etc)

  • Multiple fixes in Naitve Camera (NativeCameraView, cv::VideoCapture);
  • Improved CUDA support for all CUDA-enabled SoCs.

  • New VTK-based 3D visualization module viz stabilized and back-ported to 2.4 branch. The module provides a very convenient way to display and position clouds, meshes, cameras and trajectories, and simple widgets (cube, line, circle, etc.). Full demo video can be found at Itseez Youtube channel

  • Numerous bugfixes in code and docs from community

  • 156 pull requests have been merged since version#20

  • 55 reported bugs have been closed since version#20


$ git checkout 2.4.9
$ git shortlog --no-merges -n  -s 2.4 --not 2.4.8
36  Alexander Smorkalov
35  Ilya Lavrenov
33  Vladislav Vinogradov
23  Anatoly Baksheev
16  Andrey Pavlenko
9  Maks Naumov
8  Roman Donchenko
6  Pierre-Emmanuel Viel
6  yash
5  Daniil Osokin
4  Alexander Karsakov
4  Alexander Shishkov
4  Matti Picus
4  Nghia Ho
4  Stuart Cunningham
4  Vadim Pisarevsky
3  Alexander Alekhin
3  Kang Liu
3  Seunghoon Park
3  Vincent Rabaud
2  Firat Kalaycilar
2  Martin Jul
2  RJ2
2  berak
1  Alessandro Trebbi
1  Cody Rigney
1  ComFreek
1  Dave Hughes
1  David Rolland
1  Drew Jetter
1  Evgeniy
1  Ian Taylor
1  JinoBetti
1  Kazuki Matsuda
1  Kirill Kornyakov
1  Konstantin Matskevich
1  Linquize
1  Marijan Vukcevich
1  Martin Chodakowski
1  Robbert Klarenbeek
1  Vladimir Kolesnikov
1  Zhigang Gong
1  ahb
1  kuroda sho
1  kurodash


December, 2013

  • User provided OpenCL context can be used by OpenCV ( ocl::initializeContext )
  • A separate OpenCL command queue is created for every CPU thread (allows concurrent kernels execution)
  • Some new OpenCL optimizations and bug-fixes

  • NVidia CUDA support on CUDA capable SoCs;
  • Android 4.4 support, including native camera;
  • Java wrappers for GPU-detection functions from core module were added;
  • New sample with CUDA on Android was added;
  • OpenCV Manager and apps hanging were fixed on Samsung devices with Android 4.3 (#3368, #3372, #3403, #3414, #3436)

  • Static linkage support for native C++ libraries;

  • 139 pull requests have been merged since version:2.4.7!

  • 32 reported bugs have been closed since version:2.4.7


$ git checkout 2.4.8
$ git shortlog --no-merges -n  -s 2.4 --not 2.4.7
42  Alexander Smorkalov
35  Ilya Lavrenov
15  Roman Donchenko
15  Vladislav Vinogradov
14  Alexander Alekhin
11  krodyush
10  Andrey Pavlenko
9  Vladimir Bystricky
8  StevenPuttemans
7  Baichuan Su
7  Konstantin Matskevich
7  perping
6  Adrien BAK
6  Mimmmo Cosenza
6  Vincent Rabaud
4  Peter Andreas Entschev
4  Philippe FOUBERT
3  Anatoly Baksheev
3  berak
2  Alexander Karsakov
2  Bo Hu
2  GregoryMorse
2  Piotr Miecielica
2  moodoki
1  Alex Willisson
1  Alexander Shishkov
1  Alfonso (the fonz) de la Osa
1  Andrey Kamaev
1  Eric Sommerlade
1  Josep Bosch
1  Kiran Pradeep
1  Matthias Bady
1  Peng Xiao
1  Pierre-Emmanuel Viel
1  Séverin Lemaignan
1  Zhigang Gong
1  dpen2000
1  huangs
1  tim36272


November, 2013

  • Now ‘ocl’ module can be built without installing OpenCL SDK (Khronos headers in OpenCV tree);
  • Dynamic dependency on OpenCL runtime (allows run-time branching between OCL and non-OCL implementation);
  • Changing default OpenCL device via OPENCV_OPENCL_DEVICE environment variable (without app re-build);
  • Refactoring/extending/bug-fixing of existing OpenCL optimizations, updated documentation;
  • New OpenCL optimizations of SVM, MOG/MOG2, KalmanFilter and more;

  • New optimization for histograms, TV-L1 optical flow and resize;
  • Updated multi gpu sample for stereo matching;
  • Fixed BGR<→YUV color conversion and bitwize operations;
  • Fixed several build issues;

  • Android NDK-r9 (x86, x86_64) support;
  • Android 4.3 support: hardware detector (Bugs #3124, #3265, #3270) and native camera (Bug #3185);
  • MediaRecorder hint enabled for all Android devices with API level 14 and above;
  • Fixed JavaCameraView slowdown (Bugs #3033, #3238);

  • Fixed MS Certification test issues for all algorithmical modules and highgui, except OpenEXR and Media Foundation code for camera;
  • Implemented XAML-based sample for video processing using OpenCV;
  • Fixed issue in Media Foundation back-end for VideoCapture (#3189);

  • 382 pull requests have been merged since version:2.4.6!


$ git checkout 2.4.7
$ git shortlog --no-merges -n  -s 2.4 --not 2.4.6
111  Ilya Lavrenov
83  Alexander Alekhin
79  Roman Donchenko
41  Alexander Smorkalov
36  peng xiao
34  Jin Ma
32  Vladislav Vinogradov
27  StevenPuttemans
19  yao
17  Andrey Pavlenko
13  kdrobnyh
12  ilya-lavrenov
9  Alexander Shishkov
8  Vadim Pisarevsky
7  Greg Hale
7  Jan Starzynski
6  Bahram Dahi
5  Dmitry Retinskiy
5  Nghia Ho
5  konstantin
5  perping
3  Alexander Mordvintsev
3  Frederic Devernay
3  Kirill Kornyakov
3  Peng Xiao
3  Sebastian KrA¤mer
3  WonderRico
3  kamjagin
2  Alexander Pacha
2  Daniil Osokin
2  Gabe Schwartz
2  Hanusz Leszek
2  Nikita Manovich
2  Philippe FOUBERT
2  Zac Stewart
1  Alexander Karsakov
1  Andrew Senin
1  Baris Evrim Demiroz
1  Csaba Kertesz
1  George Andrew Brindeiro
1  Han Qiang
1  Harris Gasparakis
1  Heinz Hofbauer
1  Kevin
1  MarkBelmont
1  Michael Hanselmann
1  Miroslav Kobetski
1  Samuel Wejeus
1  Sebastian Kramer
1  abidrahmank
1  kocheganovvm
1  pengxiao


July, 2013

  • Windows RT: added video file i/o and sample application using camera, enabled parallelization with TBB or MS Concurrency

  • CUDA 5.5: added support for desktop and ARM

  • added Qt 5 support

  • Binary compatiblility with both OpenCL 1.1/1.2 platforms. Now the binaries compiled with any of AMD/Intel/Nvidia’s SDK can run on all other platforms.
  • New functions ported, CLAHE, GoodFeaturesToTrack, TVL1 optical flow and more.
  • Performance optimizations, HOG and more.
  • More kernel binary cache options though setBinaryDiskCache interface.
  • OpenCL binaries are now included into the superpack for Windows (for VS2010 and VS2012 only)

  • switched all the remaining parallel loops from TBB-only tbb::parallel_for() to universalcv::parallel_for_() with many possible backends (MS Concurrency, Apple’s GDC, OpenMP, Intel TBB etc.)

  • iOS build scripts (together with Android ones) moved to opencv/platforms directory
  • fixed bug with incorrect saved video from camera through CvVideoCamera
  • added rotateVideo flag to the CvVideoCamera class to control camera preview rotation on device rotation
  • added functions to convert between UIImage and cv::Mat (just include opencv2/highgui/ios.h)


~/Work/OpenCV/opencv> git checkout 2.4.6
~/Work/OpenCV/opencv> git shortlog --no-merges -n -s 2.4 --not 2.4.5
55  Peng Xiao
50  yao
41  Roman Donchenko
40  Alexander Smorkalov
21  Vladislav Vinogradov
10  Kirill Kornyakov
9  Andrey Pavlenko
8  Alexander Shishkov
8  Andrey Kamaev
5  alex77git
5  Dominik Rose
4  Ivan Korolev
3  Sergei Nosov
3  abidrahmank
3  poiuytrez
2  Leszek Swirski
1  Bo Li
1  Jan Machacek
1  ShengyinWu
1  Siva Prasad Varma
1  Gurpinder Singh Sandhu
1  Andrew Senin
1  Peter Minin
1  caorong
1  dave
1  ograycode
1  Dustin Spicuzza
1  Luis Díaz Más
1  Egbert van der Wal


April, 2013

  • Experimental WinRT support (build for WindowsRT guide)

  • the new video super-resolution module has been added that implements the following papers:

    • S. Farsiu, D. Robinson, M. Elad, P. Milanfar. Fast and robust Super-Resolution. Proc 2003 IEEE Int Conf on Image Process, pp. 291–294, 2003.
    • D. Mitzel, T. Pock, T. Schoenemann, D. Cremers. Video super resolution using duality based TV-L1 optical flow. DAGM, 2009.

  • CLAHE (adaptive histogram equalization) algorithm has been implemented, both CPU and GPU-accelerated versions (in imgproc and gpu modules, respectively)

  • there are further improvements and extensions in ocl module:\
    • 2 stereo correspondence algorithms: stereobm (block matching) and stereobp (belief propagation) have been added\
    • many bugs fixed, including some crashes on Intel HD4000

  • 78 pull requests have been merged. Big thanks to everybody who contributed!


March, 2013

  • This is the biggest news in 2.4.4 – we’ve got full-featured OpenCV Java bindings on a desktop, not only Android! In fact you can use any JVM language, including functional Java or handy Groovy. Big thanks to Eric Christiansen for the contribution! Check the tutorial for details and code samples.

  • Android application framework, samples, tutorials, OpenCV Manager are updated, see Android_Release_Notes for details.

  • Numerous improvements in gpu module and the following new functionality & optimizations:
    • Optimizations for the NVIDIA Kepler architecture
    • NVIDIA CARMA platform support
    • HoughLinesP for line segments detection
    • Lab/Luv <→ RGB conversions

  • Let us be more verbose here. The openCL-based hardware acceleration (ocl) module is now mature, and, with numerous bug fixes, it is largely bug-free. Correct operation has been verified on all tested platforms, including discrete GPUs (tested on NVIDIA and AMD boards), as well as integrated GPUs (AMD APUs as well as Intel Ivy Bridge iGPUs). On the host side, there has been exhaustive testing on 32/64 bit, Windows/Linux systems, making the ocl module a very serious and robust cross-platform GPU hardware acceleration solution. While we currently do not test on other devices that implement OpenCL (e.g. FPGA, ARM or other processors), it is expected that the ocl module will work well on such devices as well (provided the minimum requirements explained in the user guide are met).
    Here are specific highlights of the 2.4.4 release:
    • The ocl::Mat can now use “special” memory (e.g. pinned memory, host-local or device-local).
    • The ocl module can detect if the underlying hardware supports “integrated memory,” and if so use “device-local” memory by default for all operations.
    • New arithmetic operations for ocl::Mat, providing significant ease of use for simple numerical manipulations.
    • Interop with OpenCL enables very easy integration of OpenCV in existing OpenCL applications, and vice versa.
    • New algorithms include Hough circles, more color conversions (including YUV, YCrCb), and Hu Moments.
    • Numerous bug fixes, and optimizations, including in: blendLinear, square samples, erode/dilate, Canny, convolution fixes with AMD FFT library, mean shift filtering, Stereo BM.
    • Platform specific bug fixes: PyrLK, bruteForceMatcher, faceDetect now works also on Intel Ivy Bridge chips (as well as on AMD APUs/GPUs and NVIDIA GPUs); erode/dilate also works on NVIDIA GPUs (as well as AMD APUs/GPUs and Intel iGPUs).

  • Many people contributed their code in the form of pull requests. Here are some of the most interesting contributions, that were included into 2.4 branch:


November, 2012

  • Improved OpenCV Manager, new Java samples framework, better camera support on Android, see Android_Release_Notes for details.

  • opencv2.framework is now iOS6- and iPhone5- (armv7s) compatible. Thanks to the new threading mechanism several important OpenCV algorithms (e.g. face detection, bilateral filter, etc.) now run faster on A5 or newer dual-core chips. We also fixed bug in the framework build script and now the framework is built with “-O3” optimization instead of “-O0” in OpenCV 2.4.2. Finally, our GSoC students, Eduard Feicho and Charu Hans, have written detailed tutorials on how to add OpenCV to your iOS app, please, check

  • Numerous improvements and new functionality in GPU module:\
    • device layer opened for users; now users write CUDA device code using OpenCV.\
    • LBP cascade support in cascade classifier\
    • fast non local means image denoising\
    • faster integral image calculation on Kepler\
    • Hough circles, Hough lines and generalized Hough transform implementation\
    • bilateral filter\
    • generalized Flood Fill connected component labeling\
    • background/foreground segmentation algorithms: Mixture of Gaussian, ViBe, GMG (A. Godbehere, A. Matsukawa, K. Goldberg) .\
    • added confidence calculation into HOG

  • Technology-preview version of ocl – OpenCL-accelerated computer vision algorithms, contributed by the Chinese Academy of Science. It includes:\
    • arithmetical operations\
    • filtering\
    • geometrical image transformations (resize, remap etc.)\
    • cascade classifier (i.e. face detector)\
    • optical flow
      currently, it only runs on GPUs.

  • ~130 reported problems have been resolved since 2.4.2

  • Since 2.4.3rc we fixed several build problems (OpenCV-based applications on Xcode 4.5 & iOS6, OpenCV+TBB on Windows etc.) and closed a few issues, reported at


July, 2012

  • Android package introduces a new service-based distribution model (see Android_Release_Notes for details).

  • New keypoint descriptor FREAK has been contributed by EPFL group: Kirell Benzi, Raphael Ortiz, Alexandre Alahi and Pierre Vandergheynst. It’s claimed to be superior to ORB and SURF descriptors, yet it’s very fast (comparable to ORB). Please, see source://trunk/opencv/samples/cpp/freak_demo.cpp.

  • Improved face recognizer and excellent tutorial on using it has been added by Philipp Wagner. Check the face recognition tutorial.

  • opencv2.framework for iOS has been created. You can either download the binary from SourceForge or build it yourself using the simple guide:\#ios-installation. Also, you may be interested to look at some OpenCV on iOS samples, created by our GSoC 2012 students Eduard and Charu: gsoc2012:source:/ios/trunk and read the slides of the tutorial: gsoc2012:source:/ios/trunk/doc/CVPR2012_OpenCV4IOS_Tutorial.pdf

  • GPU module:\
    • reimplemented CUDA accelerated gpu::PyrLKOpticalFlow for dense and sparse cases. New implementation up 1.5 – 2 times faster then previous GPU optimized. Updated optical flow samples\
    • implemented resize with area interpolation. CUDA optimized version for integer matrix types up 30 – 35 faster then not optimized OpenCV implementation and up 7 in worst case of 3 channel floating point matrix.


June, 2012

  • The GPU module now supports CUDA 4.1 and CUDA 4.2 and can be compiled with CUDA 5.0 preview.

Added API for storing OpenCV data structures in text string and reading them back:\

//==== storing data ====
FileStorage fs(".xml", FileStorage::WRITE + FileStorage::MEMORY);
fs << "date" << date_string << "mymatrix" << mymatrix;
string buf = fs.releaseAndGetString();

//== reading it back ==

FileStorage fs(buf, FileStorage::READ + FileStorage::MEMORY);

fs[“date”] >> date_string;

fs[“mymatrix”] >> mymatrix;
  • Function signatures in documentation are made consistent with source code.
  • Restored python wrappers for SURF and MSER.


May, 2012

The major changes since 2.4 beta

  • OpenCV now provides pretty complete build information via cv::getBuildInformation().
  • reading/writing video via ffmpeg finally works and is now available on MacOS X too.
    note 1: we now demand reasonably fresh versions of ffmpeg/libav with libswscale included.
    note 2: if possible, try to avoid reading or writing more than one video simultaneously (even within a single thread) with ffmpeg 0.7.x or earlier, since they seem to use some global structures that are destroyed by the codecs executed synchronously. Either build and install a newer ffmpeg (0.10.x is recommended), or serialize your video i/o, or use parallel processes instead of threads.
  • MOG2 background subtraction by Zoran Zivkovic was optimized using TBB.
  • The reference manual has been updated to match OpenCV 2.4.0.
  • Asus Xtion is now properly supported for HighGUI. For now, you have to manually specify this device by using VideoCapture(CV_CAP_OPENNI_ASUS) instead of VideoCapture(CV_CAP_OPENNI).

2.4 beta

April, 2012

As usual, we created 2.4 branch in our repository (, where we will further stabilize the code. You can check this branch for changes periodically, before as well as after 2.4 release.

Common changes

  • At the age of 12, OpenCV got its own home! is now the primary site for OpenCV development and will be the official OpenCV user site.

  • Some of the old functionality from the modules imgprocvideocalib3dfeatures2dobjdetect has been moved to legacy.

  • CMake scripts have been substantially modified. Now it’s very easy to add new modules – just put the directory with include, src, doc and test sub-directories to the modules directory, create a very simple CMakeLists.txt and your module will be built as a part of OpenCV. Also, it’s possible to exclude certain modules from build (the CMake variables “BUILD_opencv_<modulename>” control that).

New functionality

  • A new essential class cv::Algorithm has been introduced. It’s planned to be the fundamental part of all of the “non-trivial” OpenCV functionality. All Algorithm-based classes have the following features:

    • “virtual constructor”, i.e. an algorithm instance can be created by name;
    • there is a list of available algorithms;
    • one can retrieve and set algorithm parameters by name;
    • one can save algorithm parameters to XML/YAML file and then load them.
  • A new ffmpeg wrapper has been created that features multi-threaded decoding, more robust video positioning etc. It’s used with ffmpeg starting with 0.7.x versions.

  • features2d API has been cleaned up. There are no more numerous classes with duplicated functionality. The base classes FeatureDetector and DescriptorExtractor are now derivatives of cv::Algorithm. There is also the base Feature2D, using which you can detect keypoints and compute the descriptors in a single call. This is also more efficient.

  • SIFT and SURF have been moved to a separate module named nonfree to indicate possible legal issues of using those algorithms in user applications. Also, SIFT performance has been substantially improved (by factor of 3-4x).

  • The current state-of-art textureless detection algorithm, Line-Mod by S. Hinterstoisser, has been contributed by Patrick Mihelich. See objdetect/objdetect.hpp, class Detector.

  • 3 face recognition algorithms have been contributed by Philipp Wagner. Please, check opencv/contrib/contrib.hppFaceRecognizer class, and opencv/samples/cpp/facerec_demo.cpp.

  • 2 algorithms for solving PnP problem have been added. Please, check flags parameter in solvePnP and solvePnPRansac functions.

  • Enhanced LogPolar implementation (that uses Blind-Spot model) has been contributed by Fabio Solari and Manuela Chessa, see opencv/contrib/contrib.hppLogPolar_* classes andopencv/samples/cpp/logpolar_bsm.cpp sample.

  • A stub module photo has been created to support a quickly growing “computational photography” area. Currently, it only contains inpainting algorithm, moved from imgproc, but it’s planned to add much more functionality.

  • Another module videostab (beta version) has been added that solves a specific yet very important task of video stabilization. The module is under active development. Please, checkopencv/samples/cpp/videostab.cpp sample.

  • findContours can now find contours on a 32-bit integer image of labels (not only on a black-and-white 8-bit image). This is a step towards more convenient connected component analysis.

  • Canny edge detector can now be run on color images, which results in better edge maps

  • Python bindings can now be used within python threads, so one can write multi-threaded computer vision applications in Python.

OpenCV on GPU

  • Different Optical Flow algorithms have been added:

    • Brox (contributed by NVidia)
    • PyrLK – both Dense and Sparse variations
    • Farneback
  • New feature detectors and descriptors:

    • GoodFeaturesToTrack
    • FAST/ORB which is patent free replacement of SURF.
  • Overall GPU module enhancements:

    • The module now requires CUDA 4.1 or later;
    • Improved similarity of results between CPU and GPU;
    • Added border extrapolation support for many functions;
    • Improved performance.
  • pyrUp/pyrDown implementations.

  • Matrix multiplication on GPU (wrapper for the CUBLAS library). This is optional, user need to compile OpenCV with CUBLAS support.

  • OpenGL back-end has been implemented for highgui module, that allows to display GpuMatdirectly without downloading them to CPU.


See the [Android_Release_Notes].


  • A few OpenCV functions, like color conversion, morphology, data type conversions, brute-force feature matcher have been optimized using TBB and/or SSE intrinisics.

  • Along with regression tests, now many OpenCV functions have got performance tests. Now for most modules one can build opencv_perf_<modulename> executables that run various functions from the particular module and produce a XML file. Note that if you want to run those tests, as well as the normal regression tests, you will need to get (a rather big) directory and set environment variable OPENCV_TEST_DATA_PATH to <your_copy_of_opencv_extra>/testdata.

Bug fixes

Known issues

  • When OpenCV is built statically, dynamically created classes (via Algorithm::create) can fail because linker excludes the “unused” object files. To avoid this problem, create classes explicitly, e.g

    Ptr<DescriptorExtractor> d = new BriefDescriptorExtractor;


August, 2011


OpenCV Java bindings for Android platform are released in ‘’Beta 2’’ quality. A lot of work is done to make them more stable and easier to use. Currently Java API has about 700 different OpenCV functions and covers 8 OpenCV modules including full port of features2d.

Other New Functionality and Features

  • Retina module has been contributed by Alexandre Benoit (in opencv_contrib module). See the new retina sample and

  • Support for Ximea cameras ( in highgui has been contributed by Ximea developers.

  • Planar subdivisions construction (Delaunay triangulation and Voronoi tessellation) have been ported to C++. See the new delaunay2.cpp sample.

  • Several new Python samples have been added.

  • FLANN in OpenCV has been upgraded to v1.6. Also, added Python bindings for FLANN.

  • We now support the latest FFMPEG (0.8.x) that features multi-threaded decoding. Reading videos in OpenCV has never been that fast.



  • Performance of the sparse Lucas-Kanade optical flow has been greatly improved. On 4-core machine it is now 9x faster than the previous version.

Bug Fixes

Known issues

  • TBB debug binaries are missed in the Windows installer. Here is a workaround:
    • Download from the TBB website.
    • Unzip and copy the tbb*_debug.dll files from bin// to the corresponding folder in the installed OpenCV location in /build/common/tbb//


July, 2011

Modifications and Improvements since 2.3rc

  • A few more bugs reported in the OpenCV bug tracker have been fixed.

  • Documentation has been improved a lot! The new reference manual combines information for C++ and C interfaces, the OpenCV 1.x-style Python bindings and the new C++-style Python bindings. It has also been thoroughly checked for grammar, style and integrity.

Besides, there are new and updated tutorials.

The up-to-date online documentation is available at

Unfortunately, it’s not a full-scale installation package, but simply a self-extracting archive with a readme.txt supplied.
The installation package is probably to come in the next version.

  • VS2005 should build OpenCV 2.3 out of the box, including DirectShow support.

  • ffmpeg bindings are now available for all Windows users via compiler- and configuration- and
    version-independent opencv_ffmpeg.dll (for 32-bit compilers) and opencv_ffmpeg_64.dll (for 64-bit compilers).

2.3 beta

June, 2011

General Modifications and Improvements

  • Buildbot-based Continuous Integration system is now continuously testing OpenCV snapshots. The current status is available at

  • OpenCV switched to Google Test ( engine for regression and correctness tests. Each module now has “test” sub-directory which includes the corresponding tests.

New Functionality, Features

  • Many functions and methods now take InputArray/OutputArray instead of cv::Matreferences. It retains compatibility with the existing code and yet brings more natural support for STL vectors and potentially other “foreign” data structures to OpenCV. See for details.

  • core

    • LAPACK is not used by OpenCV anymore. The change decreased the library footprint and the compile time. We now use our own implementation of Jacobi SVD. SVD performance on small matrices (2×2 to 10×10) has been considerably improved; on larger matrices it is still pretty good. SVD accuracy on poorly-conditioned matrices has also been polished.
    • Arithmetic operations now support mixed-type operands and arbitrary number of channels.
  • features2d

    • Completely new patent-free BRIEF and ORB feature descriptors have been added.
    • Very fast LSH matcher for BRIEF and ORB descriptors will be added in 2.3.1.
  • calib3d

    • A new calibration pattern, circles grid, has been added. See findCirclesGrid() function and the updated calibration.cpp sample. With the new pattern calibration accuracy is usually much higher.
  • highgui

    • Windows videoInput is now a part of highgui. If there are any problems with compiling highgui, set WITH_VIDEOINPUT=OFF in CMake.
  • stitching

    • opencv_stitching is a beta version of new application that makes a panorama out of a set of photos taken from the same point.
  • python
    • Now there are 2 extension modules: cv and cv2cv2 includes wrappers for OpenCV 2.x functionality. opencv/samples/python2 contain a few samples demonstrating cv2 in use.
  • contrib
    • A new experimental variational stereo correspondence algorithm StereoVar has been added.
  • gpu
    • the module now requires CUDA 4.0 or later; Many improvements and bug fixes have been made.

Android port

  • With support from NVIDIA, OpenCV Android port (which is actually not a separate branch of OpenCV, it’s the same code tree with additional build scripts) has been greatly improved, a few demos developed. Camera support has been added as well. See [Android_Release_Notes#231-beta1] for details.


  • OpenCV documentation is now written in ReStructured Text and built using Sphinx(

  • It’s not a single reference manual now, it’s 4 reference manuals (OpenCV 2.x C++ API, OpenCV 2.x Python API, OpenCV 1.x C API, OpenCV 1.x Python API), the emerging user guide and a set of tutorials for beginners.

  • Style and grammar of the main reference manual (OpenCV 2.x C++ API) have been thoroughly checked and fixed.

  • Online up-to-date version of the manual is available at



  • Several ML algorithms have been threaded using TBB.

Bug Fixes

Known Problems/Limitations

  • Documentation (especially on the new Python bindings) is still being updated. Watch for updates.

  • Android port does not provide Java interface for OpenCV. It is going to be added to 2.3 branchin a few weeks.


December, 2010

General Modifications and Improvements

  • The library has been reorganized. Instead of cxcorecvcvauxhighgui and ml we now have several smaller modules:
    • opencv_core – core functionality (basic structures, arithmetics and linear algebra, dft, XML and YAML I/O …).
    • opencv_imgproc – image processing (filterGaussianBlurerodedilateresizeremapcvtColorcalcHist etc.)
    • opencv_highgui – GUI and image & video I/O
    • opencv_ml – statistical machine learning models (SVM, Decision Trees, Boosting etc.)
    • opencv_features2d – 2D feature detectors and descriptors (SURF, FAST etc., including the new feature detectors-descriptor-matcher framework)
    • opencv_video – motion analysis and object tracking (optical flow, motion templates, background subtraction)
    • opencv_objdetect – object detection in images (Haar & LBP face detectors, HOG people detector etc.)
    • opencv_calib3d – camera calibration, stereo correspondence and elements of 3D data processing
    • opencv_flann – the Fast Library for Approximate Nearest Neighbors (FLANN 1.5) and the OpenCV wrappers
    • opencv_contrib – contributed code that is not mature enough
    • opencv_legacy – obsolete code, preserved for backward compatibility
    • opencv_gpu – acceleration of some OpenCV functionality using CUDA (relatively unstable, yet very actively developed part of OpenCV)

If you detected OpenCV and configured your make scripts using CMake or pkg-config tool, your code will likely build fine without any changes. Otherwise, you will need to modify linker parameters (change the library names) and update the include paths.

It is still possible to use #include <cv.h> etc. but the recommended notation is:

#include "opencv2/imgproc/imgproc.hpp"

Please, check the new C and C++ samples (, which now include the new-style headers.

  • The new-style wrappers now cover much more of OpenCV 2.x API. The documentation and samples are to be added later. You will need numpy in order to use the extra functionality.
    SWIG-based Python wrappers are not included anymore.

  • OpenCV can now be built for Android (GSoC 2010 project), thanks to Ethan Rublee; and there are some samples too. Please, check [OpenCV4Android]

  • The completely new opencv_gpu acceleration module has been created with support by NVIDIA. See below for details.

New Functionality, Features

  • core

    • The new cv::Matx<T, m, n> type for fixed-type fixed-size matrices has been added. Vec<T, n> is now derived from Matx<T, n, 1>. The class can be used for very small matrices, where cv::Mat use implies too much overhead. The operators to convert Matxto Mat and backwards are available.
    • cv::Mat and cv::MatND are made the same type: typedef cv::Mat cv::MatND.
      Note that many functions do not check the matrix dimensionality yet, so be careful when processing 3-, 4- … dimensional matrices using OpenCV.
    • Experimental support for Eigen 2.x/3.x is added (WITH_EIGEN2 option in CMake). Again, there are convertors from Eigen2 matrices to cv::Mat and backwards. Seemodules/core/include/opencv2/core/eigen.hpp.
    • cv::Mat can now be print with “<<” operator. See opencv/samples/cpp/cout_mat.cpp.
    • cv::exp and cv::log are now much faster thanks to SSE2 optimization.
  • imgproc

    • color conversion functions have been rewritten;
    • RGB→Lab & RGB→Luv performance has been noticeably improved. Now the functions assume sRGB input color space (e.g. gamma=2.2). If you want the original linear RGB→L** conversion (i.e. with gamma=1), use CV_LBGR2LAB etc.
    • VNG algorithm for Bayer→RGB conversion has been added. It’s much slower than the simple interpolation algorithm, but returns significantly more detailed images
    • The new flavors of RGB→HSV/HLS conversion functions have been added for 8-bit images. They use the whole 0..255 range for the H channel instead of 0..179. The conversion codes are CV_RGB2HSV_FULL etc.
    • special variant of initUndistortRectifyMap for wide-angle cameras has been added: initWideAngleProjMap()
  • features2d

    • the unified framework for keypoint extraction, computing the descriptors and matching them has been introduced. The previously available and some new detectors and descriptors, like SURFFASTStarDetector etc. have been wrapped to be used through the framework. The key advantage of the new framework (besides the uniform API for different detectors and descriptors) is that it also provides high-level tools for image matching and textured object detection. Please, see documentation\ and the C++ samples:
      • descriptor_extractor_matcher.cpp – finding object in a scene using keypoints and their descriptors.
      • generic_descriptor_matcher.cpp – variation of the above sample where the descriptors do not have to be computed explicitly.
      • bagofwords_classification.cpp – example of extending the framework and using it to process data from the VOC databases:
    • the newest super-fast keypoint descriptor BRIEF by Michael Calonder has been integrated by Ethan Rublee. See the sample opencv/samples/cpp/video_homography.cpp
    • SURF keypoint detector has been parallelized using TBB (the patch is by imahon and yvo2m)
  • objdetect

    • LatentSVM object detector, implementing P. Felzenszwalb algorithm, has been contributed by Nizhniy Novgorod State University (NNSU) team. Seeopencv/samples/c/latentsvmdetect.cpp
  • calib3d

    • The new rational distortion model has been introduced. It is useful for calibration of cameras with wide-angle lenses. Because of the increased number of parameters to optimize you need to supply more data to robustly estimate all of them. Or, simply initialize the distortion vectors with zeros and pass CV_CALIB_RATIONAL_MODEL to enable the new model CV_CALIB_FIX_K3 + CV_CALIB_FIX_K4 + CV_CALIB_FIX_K5 or other such combinations to selectively enable or disable certain coefficients.

    • rectification of trinocular camera setup, where all 3 heads are on the same line, is added. see samples/cpp/3calibration.cpp

  • ml

    • Gradient boosting trees model has been contributed by NNSU team.
  • highgui

    • Experimental Qt backend for OpenCV has been added as a result of GSoC 2010 project, completed by Yannick Verdie. The backend has a few extra features, not present in the other backends, like text rendering using TTF fonts, separate “control panel” with sliders, push-buttons, checkboxes and radio buttons, interactive zooming, panning of the images displayed in highgui windows, “save as” etc. Please, check the youtube videos where Yannick demonstrates the new features:
    • The new API is described here: To make use of the new API, you need to have Qt SDK (or libqt4 with development packages) installed on your machine, and build OpenCV with Qt support (pass -DWITH_QT=ON to CMake; watch the output, make sure Qt is used as GUI backend)
    • 16-bit and LZW-compressed TIFFs are now supported.
    • You can now set the mode for IEEE1394 cameras on Linux.
  • contrib

    • Chamfer matching algorithm has been contributed by Marius Muja, Antonella Cascitelli, Marco Di Stefano and Stefano Fabri. See samples/cpp/chamfer.cpp
  • gpu This is completely new part of OpenCV, created with the support by NVIDIA. Note that the package is at alpha, probably early beta state, so use it with care and check OpenCV SVN for updates.

In order to use it, you need to have the latest NVidia CUDA SDK installed, and build OpenCV with CUDA support (-DWITH_CUDA=ON CMake flag).

All the functionality is put to cv::gpu namespace. The full list of functions and classes can be found at opencv/modules/gpu/include/opencv2/gpu/gpu.hpp, and here are some major components of the API:

  • image arithmetics, filtering operations, morphology, geometrical transformations, histograms

  • 3 stereo correspondence algorithms: Block Matching, Belief Propagation and Constant-Space Belief Propagation.

  • HOG-based object detector. It runs more than order of magnitude faster than the CPU version! See opencv/samples/gpu

  • python bindings. A lot more of OpenCV 2.x functionality is now covered by Python bindings.

    These new wrappers require numpy to be installed
    (see for details).

    Likewise the C++ API, in the new Python bindings you do not need to allocate output arrays.
    They will be automatically created by the functions.

    Here is a micro example:

    import cv
    b=cv.canny(a, 50, 100, apertureSize=3)

    In the sample a and b are normal numpy arrays, so the whole power of numpy and scipycan now be combined with OpenCV functionality.

Documentation, Samples

  • Links to wiki pages have been added to each function description, see

  • All the samples have been documented with default output ‘’(0 or incomplete number of parameters)’’ set to print out “howto” run instructions; most samples have been converted to C++ to use the new OpenCV API.

Bug Fixes

Known Problems/Limitations

  • Installation package for Windows is still 32-bit only and does not include TBB support. You can build parallel or 64-bit version of OpenCV from the source code.


  • 广告
  • 抄袭
  • 版权
  • 政治
  • 色情
  • 无意义
  • 其他