Judith-Alice Redi - Post Doc Multimedia Communications
Date: - Location: Eurecom
A major challenge in the design of modern imaging technologies is quality control and enhancement. Effective quality control should stem from the application of appropriate post-processing techniques, driven by the automated estimation of the signal visual quality. The talk will introduce the problem of visual quality assessment, both from a subjective and an objective point of view. Real-time estimation issues will be discussed and then addressed in a framework designed to allow in-service quality estimation of images. Three elements concur to the effectiveness of the framework. First, it is calibrated on overall image quality, to ensure its predictions to be independent on particular impairments. Second, Computational Intelligence methods enter the prediction paradigm to mimic perceptual mechanisms. Their well-known capabilities in modeling non-linearities are exploited so that the complexity of the objective quality metric can be minimized and both real-time and accuracy constraints can be fulfilled. Finally, the interaction between quality perception and visual attention is investigated and modeled, increasing the reliability of the estimates.
