PAD: a perceptual application-dependent metric for quality assessment of segmentation algorithms

Abstract

Extracting elements of interest from video frames is a necessary task in many applications, such as those that require replacing the original background. Quality assessment of foreground extraction algorithms is essential to find the best algorithm for a particular application. This paper presents an application-dependent objective metric capable of evaluating the quality of those algorithms by considering user perception. Our metric identifies types of errors that cause the greatest annoyance based on regions of the scene where users tend to keep their attention during videoconference sessions. We demonstrate the efficiency of our metric by evaluating bilayer segmentation algorithms. The results showed that metric is effective compared to others used to evaluate algorithms for videoconferencing systems.