Segmentation-Driven Perceptual Quality Metrics (WP-P9)

Author(s) : Andrea Cavallaro (Queen Mary University of London, UK) Stefan Winkler (Genista Corporation, Switzerland)

Abstract : We present a perceptual video quality metric that incorporates both low-level and high-level aspects of vision. Low-level aspects include colour perception as well as spatio-temporal contrast sensitivity and masking. High-level aspects take into account the cognitive behaviour of an observer when watching a video through semantic segmentation. The evaluation of the proposed segmentation-driven perceptual quality metric when the semantic segmentation addresses faces demonstrates an improvement of the prediction performance for a range of scenes.

Menu