大致整理了一些。现在大多数还是基于机器学习的模型。
论文集合如下。
数据集整理如下。
Name | where | abbreviation | content | distortions | url |
Perceptual Evaluation of Light Field Image | 2018-ICIP | Win5-LID | 220 distorted LFIs with 6 distortion types | HEVC / JPEG2000 / LN / NN / two distortions from conv | |
Reconstruction Distortion Oriented Light Field Image Dataset for Visual Communication | 2019-ISNCC | NBU1.0 | 210 distorted LFIs with 5 distortion types | NN / bicubic interpolation / learning based reconstruction / disparity map based reconstruction / spatial super-resolution reconstruction(3 levels) | https://github.com/JianjunXiang/PVRI/blob/616b68ba13a959b42479f423b0425e2da6a57e8e/README.txt |
VALID: Visual quality Assessment for Light field Images Dataset | 2018-QoMEX | VALID | 8bits and 10bits LFIs | 5 original LFIs and 40 8bits distorted with VP9 and HEVC(4 levels) | VALID: Visual quality Assessment for Light field Images Dataset ‒ MMSPG ‐ EPFL |
A no-reference image quality assessment metric by multiple characteristics of light field images | 2019-Access | 240 distortion LFIs with 5 types | JPEG / JPEG2000 / Gaussian blur / motion blur / white noise(6 levels) | ||
Towards a New Quality Metric for 3-D Synthesized View Assessment | 2011-J-STSP | MPI-LFA | 336 distorted LFIs | HEVC / LINEAR / NN / optical flow estimation(OPT) / quantized depth maps(DQ) / gaussian blur(6 levels) | |
Towards the perceptual quality evaluation of compressed light field images | 2017-T-Broadcast | SMART | 256 distorted LFIs | HEVC / JPEG / JPEG2000 / SSDC | COMLAB - Telecommunication Lab (20230126更新了现在可以了 这是文章链接 里面有数据集链接) |
Towards a quality metric for dense light fields | 2017-CVPR | Dense Light Fields dataset | LINEAR / NN / OPT / DQ / GAUSS | ||
A Dataset and Evaluation Methodology for Depth Estimation on 4D Light Fields | 2016-ACCV | ||||
The (New) Stanford Light Field Archive |
(这里表格编辑的好累。。虽然很丑但是懒得改了。。