Vision models and visual quality

One grand challenge for the engineering of multimedia quality has been to develop algorithms that can convert physical measurements - of displays, of images, of graphics, of video sequences, or of complete imaging systems - into metrics that have perceptual meaning. These metrics should enable automated detection of artifacts, quantification of system performance, and optimization of positive multimedia attributes. In the last decade, we and others have made some progress towards this goal. The progress has been achieved by joining display and image measurements to simplified models of processing in the human visual system. In this talk I will describe several of the key concepts and components of these models, and will show how the models can be applied to key problems in display design, image and video processing, and computer graphics.