Multimedia Signal Processing Seminar
Crowdsourcing and Deep Learning for Visual Quality Assessment in the Wild
In this seminar we cover topics related to our research focus and that may be relevant to Bachelor, Master, and PhD projects in the Multimedia Signal Processing Group.
- Image and video quality assessment (IQA/VQA): Nowadays, multimedia data are ubiquitous. You take pictures and videos with your camera and camcorders everyday. Have you thought about how you can judge the perceptual quality of the pictures and videos you made and how good your camera/camcorder is? In this seminar, we will discuss algorithms for image and video quality assessment, especially for the purpose of assessing performance of digital cameras and camcorders. The following aspects of perceptual quality are of interests: blur/sharpness, noise, color fidelity, resolution, geometric distortions, blockiness, ringing, frame dropping and freezing, and so on. Topics will be related to our ongoing VQA Project in the SFB TRR 161.
- Recently deep learning has been applied the problem of IQA/VQA and we may cover several approaches in the seminar.
- Besides quality assessment for multimedia, algorithms for assessing similar perceptual properties like aesthetics and enhancement of media heve been porposed and will be a topic in the seminar.
- We also discuss crowdsourcing applications not only for multimedia quality assessment, but also for other purposes like social experiments and employing methods of psychometrics for the data analysis.
- Literature will be selected in the first week of the seminar.
Requirements for obtaining ECTS credits for this seminar are:
- regular attendance at the weekly seminar
- self-study of one or more scientific papers on selected topics
- oral presentation of 45 minutes including discussion
- written essay on the subject of the presentation (about 10-20 pages, typeset, electronic submission as one pdf)
Students with a keen interest in multimedia signal processing are welcome to participate in the seminar. It may be helpful if previously related courses were taken like Digital Signal Processing, Eye Tracking Theory and Practice, Image Processing, but this is not required.
Topics and Literature
Topic: Learning from rankings
- Liu, X., van de Weijer, J. and Bagdanov, A.D., 2017. Rank-IQA: Learning from rankings for no-reference image quality assessment. In Proceedings of the IEEE International Conference on Computer Vision (pp. 1040-1049). Available online. Supplementary materials available online. Paper: http://openaccess.thecvf.com/content_ICCV_2017/papers/Liu_RankIQA_ Learning_From_ICCV_2017_paper.pdf
Topic: Image augmentations for database enhancement
- Zhu, Y., Zhai, G., Zhu, W. and Zhou, J., 2018, May. An Image Augmentation Method for Quality Assessment Database. In 2018 IEEE International Symposium on Circuits and Systems (ISCAS). Abstract available online at IEEE. Will get the paper asap
Topic: Artifact detection in video
- Goodall, T.R., Bovik, A., Detecting and Mapping Video Impairments, IEEE Transactions on Image Processing 28(6), June 2019. Thesis available online: Goodall, T.R., 2018. Inspection and evaluation of artifacts in digital video sources (Doctoral dissertation). Also the software is available online and should be tried out for the seminar. Thesis: https://repositories.lib.utexas.edu/handle/2152/68147
Topic: Just noticeable differences I
- Katsigiannis, S., Scovell, J., Ramzan, N., Janowski, L., Corriveau, P., Saad, M.A. and Van Wallendael, G., 2018. Interpreting MOS scores, when can users see a difference? Understanding user experience differences for photo quality. Quality and User Experience, 3(1), p.6. Paper: https://link.springer.com/article/10.1007/s41233-018-0019-8
Topic: Just noticeable differences II
- Wu, J., Shi, G., Lin, W., Liu, A., Qi, F. (2013). Just noticeable difference estimation for images with free-energy principle. IEEE Transactions on Multimedia, 15(7), 1705-1710. Talk should also consider the three methods of Yang, Liu, Zhang. Paper: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.701.980&rep=rep1&type=pdf
Topic: Weighted mean square errors
- Hu, S., Jin, L., Wang, H., Zhang, Y., Kwong, S. and Kuo, C.C.J., 2017. Objective video quality assessment based on perceptually weighted mean squared error. IEEE Transactions on Circuits and Systems for Video Technology, 27(9), pp.1844-1855. Paper: http://codec.siat.ac.cn/downloads/publications/47.pdf
- Xue, W., Mou, X., Zhang, L. and Feng, X., 2013. Perceptual fidelity aware mean squared error. In Proceedings of the IEEE International Conference on Computer Vision (pp. 705-712). Paper: https://www.cv-foundation.org/openaccess/content_iccv_2013/papers/Xue_Perceptual_Fidelity_Aware_2013_ICCV_paper.pdf