综述-无参考视频质量评价（NR-VQA）
无参考视频质量评价（NR-VQA）和全参考基本共享同类的模型框架，它们要么基于frame-by-frame的NR-IQA+时域特征+temporal pooling，要么基于时空特征提取。frame-by-frame的NR-IQA一般采用NIQE，如,，但也有一些作者是基于自己之前的NR-IQA工作，做的自然而然的扩展，如,。时空特征描述的方式也很大程度上借鉴了NR-IQA中的一些工作,。如果非要用某种方法对无参考视频质量评价（NR-VQA）方法进行分类，一种常用分类方式应是以深度学习（准确的说，CNN）为界，分为：[*]传统方法：不基于学习，或者基于经典机器学习方法做回归的方法，包括但不仅限于SVR、RF、NN等;
[*]半深度：一般是基于CNN提取特征+手工特征，或者使用CNN做特征增强；
[*]深度学习：完全基于深度学习的框架，没有手工特征的引入等。
传统方法传统方法里，基于自然视频统计特性（Natural Video Statistics，NVS）占据了绝大多数。NVS是从NSS的自然而然的过渡。-、和均是基于GGD/AGGD的参数拟合。其中，基本都是对帧间差分的MSCN的NVS建模，或者在DCT域对系数分布建模（2D-DCT/3D-DCT），对于运动的刻画一般使用两帧之间的运动补偿（DCT系数的变化等）。尽管一些工作用NIQE做为spatial quality，但是基于笔者对于videoBLIINDS 的验证，有NIQE和没有NIQE的影响没有这么大，也就是说，NIQE直接用到NR-VQA中效果极为有限，直接将NR-IQA用于做NR-VQA的度量工作可能有待商榷。当然，目前公布代码的NR-VQA工作基本只有和，更多的验证工作也没有办法展开。其他的传统方法里，videoCORNIA 延续了CORNIA的方法，只是增加了temporal pooling；一些基于DCT能量分布的方法 取得了较为不错的成绩；同样地，基于光流的光滑假设，一些基于光流特征刻画的工作也取得了较好的效果，甚至于在立体视频质量评价领域，工作的时域质量估计效果要优于空域质量估计；一些利用时空特征的工作，比如。工作基于3D-DCT的统计特性，工作基于3D张量分解，工作基于LBP_TOP。半深度一部分半深度的工作是基于现有的CNN网络提取空域特征，再手动加入时域特征，如；工作将3D-shearlet的系数作为1D-CNN的输入，用1D-CNN来进行特征增强与回归；工作以3D-DCT的AC系数和作为输入，以CNN来提取特征（但基于作者的实验，CNN什么的都不重要，resample策略才是重中之重，然而resample策略很有局限）。深度学习目前笔者能够搜集到的文章，仅有称得上是完全基于深度学习的NR-VQA。拓展了其在NR-IQA中的方法MEON，将2D卷积改为了3D卷积，增加了层级的slow fusion来捕获更多的时空信息，利用ffmpeg和SSIMPlus构建了包含3k个H.264和HEVC编码视频的训练数据集，得到了第一个针对两类压缩噪声的端对端的DNN模型。另有一些针对FR-VQA的DNN工作，如，要么基于作者自己IQA工作的拓展，要么只是将DNN应用于该领域，且不属于NR范畴。 Yao J, Xie Y, Tan J, et al. No-reference video quality assessment using statistical features along temporal trajectory. Procedia Engineering, 2012, 29: 947-951. Manasa K, KVSNL M P, Channappayya S S. A perceptually motivated no-reference video quality assessment algorithm for packet loss artifacts//2014 Sixth International Workshop on Quality of Multimedia Experience (QoMEX). IEEE, 2014: 67-68. Saad M A, Bovik A C, Charrier C. Blind prediction of natural video quality. IEEE Transactions on Image Processing, 2014, 23(3): 1352-1365. Xia X, Lu Z, Wang L, et al. Blind video quality assessment using natural video spatio-temporal statistics//2014 IEEE International Conference on Multimedia and Expo Workshops (ICMEW). IEEE, 2014: 1-6. Xu J, Ye P, Liu Y, et al. No-reference video quality assessment via feature learning//2014 IEEE international conference on image processing (ICIP). IEEE, 2014: 491-495. Zhu K, Li C, Asari V, et al. No-reference video quality assessment based on artifact measurement and statistical analysis. IEEE Transactions on Circuits and Systems for Video Technology, 2014, 25(4): 533-546. Li X, Guo Q, Lu X. Spatiotemporal statistics for video quality assessment. IEEE Transactions on Image Processing, 2016, 25(7): 3329-3342. Li Y, Po L M, Cheung C H, et al. No-reference video quality assessment with 3D shearlet transform and convolutional neural networks. IEEE Transactions on Circuits and Systems for Video Technology, 2015, 26(6): 1044-1057. Manasa K, Channappayya S S. An optical flow-based no-reference video quality assessment algorithm//2016 IEEE International Conference on Image Processing (ICIP). IEEE, 2016: 2400-2404. Mittal A, Saad M A, Bovik A C. A completely blind video integrity oracle. IEEE Transactions on Image Processing, 2015, 25(1): 289-300. Ghadiyaram D, Chen C, Inguva S, et al. A no-reference video quality predictor for compression and scaling artifacts//2017 IEEE International Conference on Image Processing (ICIP). IEEE, 2017: 3445-3449. Wang C, Su L, Huang Q. CNN-MR for no reference video quality assessment//2017 4th International Conference on Information Science and Control Engineering (ICISCE). IEEE, 2017: 224-228. Yang J, Wang H, Lu W, et al. A no-reference optical flow-based quality evaluator for stereoscopic videos in curvelet domain. Information Sciences, 2017, 414: 133-146. Ahn S, Lee S. Deep blind video quality assessment based on temporal human perception//2018 25th IEEE International Conference on Image Processing (ICIP). IEEE, 2018: 619-623. Göring S, Skowronek J, Raake A. DeViQ–A deep no reference video quality model. Electronic Imaging, 2018, 2018(14): 1-6. Jiang G, Liu S, Yu M, et al. No reference stereo video quality assessment based on motion feature in tensor decomposition domain. Journal of Visual Communication and Image Representation, 2018, 50: 247-262. Liu W, Duanmu Z, Wang Z. End-to-End Blind Quality Assessment of Compressed Videos Using Deep Neural Networks//ACM Multimedia. 2018: 546-554. Zhang Y, Gao X, He L, et al. Blind video quality assessment with weakly supervised learning and resampling strategy. IEEE Transactions on Circuits and Systems for Video Technology, 2018. Singh R, Aggarwal N. A distortion-agnostic video quality metric based on multi-scale spatio-temporal structural information. Signal Processing: Image Communication, 2019, 74: 299-308. Wang Y, Shuai Y, Zhu Y, et al. Jointly learning perceptually heterogeneous features for blind 3D video quality assessment. Neurocomputing, 2019, 332: 298-304. Yang J, Zhu Y, Ma C, et al. Stereoscopic video quality assessment based on 3D convolutional neural networks. Neurocomputing, 2018, 309: 83-93. Kim W, Kim J, Ahn S, et al. Deep video quality assessor: From spatio-temporal visual sensitivity to a convolutional neural aggregation network//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 219-234. Zhang Y, Gao X, He L, et al. Objective Video Quality Assessment Combining Transfer Learning With CNN. IEEE transactions on neural networks and learning systems, 2019.
https://blog.csdn.net/u014150857/article/details/94909715
页:
[1]