AutoStitch: a new dimension in automatic image stitching

ref: http://matthewalunbrown.com/autostitch/autostitch.html

https://github.com/ipcvgroup21/panorama

一个(全局)单应性对齐+柱面或球面投影+光束平差+多频带融合为核心的老一代拼接算法以BROWN大神03'ICCV和07'IJCV的AutoStitch AutoStitch为里程碑，已经非常成熟，各路拼接软件和应用都纷纷落地，著名的如OpenCV的实现stitcher http://stitching. Images stitching，微软的ICE，Image Composite Editor (64 bit) 和Photoshop中的拼接工具，某段时间图像拼接被认为是个已经完美解决的问题。  

Brown M, Lowe D G. Recognising Panoramas [C]// ICCV. 2003. 

Brown M, Lowe D G. Automatic Panoramic Image Stitching using Invariant Features [J]. IJCV, 2007.

---

Windows |  Mac OSX |  Licensing |  FAQ |  Publications

AutoStitch: a new dimension in automatic image stitching

What if your camera could see everything you can see? The human visual system has a field of view of around 135 x 200 degrees, but a typical camera has a field of view of only 35 x 50 degrees. Panoramic image mosaicing works by taking lots of pictures from an ordinary camera, and stitching them together to form a composite image with a much larger field of view. AutoStitch takes a step forward in panoramic image stitching by automatically recognising matching images. This allows entire panoramas to be constructed with no user input whatsoever. AutoStitch is incredibly simple to use! Just select a set of photos, and AutoStitch does the rest: digital photos in, panoramas out. Mobile versions are no longer available, sorry! Please try the free demo instead

Download the free demo version Windows • Mac OSX

For 32-bit Windows users here is the older  Win32 demo version.

For more advanced stitching capabilities, you may wish to try:  Autopano pro,  Serif PanoramaPlus, or  Calico

The AutoStitch Process

AutoStitch works from unordered collections of images, automatically finding matches between images using the SIFTalgorithm. It then robustly aligns all images and uses advanced blending algorithms to form seamless panoramas (see below). For more details, see our research papers.

25 of 57 images aligned

All 57 images aligned

Final Result

Licensing

AutoStitch is available to license from the University of British Columbia. Licensing is handled by the UBC Industry Liason Office, please see this page for details and contact information.

AutoStitch is now available in the following commercial products:

• Autopano Pro www.autopano.net (Windows, Mac, Linux)
• Serif PanoramaPlus www.serif.com (Windows)
• Calico www.kekus.com (Mac)

The University of British Columbia has also granted a commercial license to Industrial Light & Magic (ILM) www.ilm.com, a Lucasfilm Ltd. company, to use AutoStitch software to produce panoramas for film production.

The version of AutoStitch on this website is a demo only. Individuals or companies are free to use images that they generate using the demo version of AutoStitch without restriction or royalties so long as they acknowledge the use of AutoStitch in such works. A commercial license to AutoStitch provides access to the patent, source code, technical support and updates

The conversion of AutoStitch to C++ was partially funded by Western Economic Diversification Canada 

Press

• Flickr Photos tagged with AutoStitch
• Panoguide Q & A - AutoStitch demo release
• New Scientist (print) The Whole Shooting Match. 18 October 2003 p.25
• Forbes Student Develops Software for Digital Panoramas
• VRLog A New Stitching System

FAQ

Q: How do I get AutoStitch to stitch the panorama fullsize?

A:Go to Settings or Edit->Options and set Output Size to 100%. Then select Play or Stitch->Start.

Q: What projection method does AutoStitch use? Does it support full view panoramas?

A:The demo version of AutoStitch uses spherical projection and is capable of stitching full view 360 x 180 degree panoramas (everything visible from a point). Cylindrical and planar projections are not supported in the demo version.

Q: Why doesn't AutoStitch recognise multiple panoramas?

A:This feature is disabled in the demo version

Q: Does AutoStitch support planar stitching, such as flatbed scans or aerial photographs?

A:The demo version of AutoStitch assumes that the camera is rotating about a point, so distortions will be visible when stitching multiple views of a planar surface.

Publications

• Automatic Panoramic Image Stitching using Invariant Features. M. Brown and D. Lowe. 
    International Journal of Computer Vision. 74(1), pages 59-73, 2007. (pdf | bib)
• Recognising Panoramas. M. Brown and D. G. Lowe. 
    International Conference on Computer Vision (ICCV2003). pages 1218-1225, Nice, France, 2003. (pdf | bib | ppt)

Windows |  Mac OSX |  Licensing |  FAQ |  Publications