Quality is in The Eye of the Beholder
Date: Wednesday Morning, October 15, 08:30 - 09:30
Location: San Diego Convention Center, Upper Level, Room 6D/E
Presented by
Al Bovik, University of Texas at Austin
Abstract
In this talk I draw analogies between the perception of visual quality and the classical theory of information communication. However, the picture is clouded by the fact that, unlike most engineering problems, the receiver and transmitter are difficult to access, model or even define.
I will describe my philosophy regarding these models and our recent efforts on quality assessment of visual signals. I’ll begin with our successful still image algorithms, the Structural SIMilarity (SSIM) Index and the Visual Information Fidelity (VIF) Index. I will then describe recent efforts on the more complex problem of Video Quality Assessment (VQA). This includes extending the ideas of SSIM and VIF to detect quality along motion trajectories. We have developed a new and effective VQA index as a result, which we call the MOtion-based Video Integrity Evaluation index, or MOVIE index. I will describe the rationale behind MOVIE, including relevant perceptual models for multiband video decomposition, motion estimation, spatial and temporal masking, and spatial and temporal quality evaluation.
Databases for comparing VQA algorithms are difficult to design and construct, since it requires gathering large volumes of video data, creating libraries of distorted videos, and conducting extensive human studies. I’ll describe our recent efforts in this direction: the soon-to-be-publicly-available LIVE Video Quality Assessment Database, which includes an interesting diversity of distortion types and perceptual annoyance levels, and the outcome of a large-scale human subjective study resulting in thousands of human judgments.
Speaker Biography
Al Bovik is the Curry/Cullen Trust Endowed Chair Professor at The University of Texas at Austin. He is well known for his fundamental work in nonlinear image processing, perceptual image and video processing and computational modeling of visual perception. Al is particularly noted for his pioneering work on robust image processing, multiresolution image processing, image and video quality assessment, and computational modeling of visual perception. He is also noted for innovations in engineering education, including his popular books and the widely-used Signal, Image and Video Audiovisual (SIVA) Demonstration Gallery. Al has also been a major contributor to engineering service, including innovating and creating the IEEE International Conference on Image Processing, first held in Austin, Texas, in November, 1994, and playing an instrumental role in proposing and creating the IEEE Transactions on Image Processing, for which he served as Editor-in-Chief for six years. Al has received a number of major awards from the IEEE Signal Processing Society, including: the Education Award (2008); the Technical Achievement Award (2005), the Distinguished Lecturer Award (2000); and the Meritorious Service Award (1998). He is a Fellow of the IEEE, a Fellow of the Optical Society of America (OSA), and a Fellow of the Society of Photo-Optical and Instrumentation Engineers (SPIE).