Enhanced Full-reference Quality Assessment Method for Image and Video

Evaluation and optimization of visual quality is of fundamental importance to numerous image and video processing applications including broadcasting, archiving, and restoration. As quality assessment using subjective tests is very costly and time-consuming, objective measures are required that can predict the perceptual judgment of human viewers. In this paper, a topdown full-reference quality assessment tool that mimics a selection of prominent human visual system properties is presented. Perceptually salient feature points are thereby used. The proposed approach is based on an enhanced structure detection method using spatial segmentation. Experimental evaluations are conducted on two test sets. It is shown that compared to the best quality measure by the Video Quality Experts Group, the proposed method yields the most robust prediction of subjective quality given a wide range of video content.