Objective quality assessment of 3D stereoscopic video based on motion vectors and depth map features

Video content distortion and visual quality degradation during transmission has obstructed the improvement of 3D video visualization systems. In this context, efficient objective quality assessment of 3D video is a critical demand, in particular for real world applications. Recent objective 3D video quality assessment methods are based on availability of the original (reference) video as well as utilizing depth map information of video frames. In this paper, we study the efficacy of using robust features extracted from 3D stereoscopic video, to propose a no-reference (NR) video quality assessment (VQA) model. The extracted features included; motion vector lengths and depth map information of 3D stereoscopic video frames. The proposed quality metric, entitled MD-QA, was tested on EPFL 3D stereoscopic video database and compared to full-reference (FR) objective quality assessment methods. The performance evaluation of MD-QA was achieved by applying it over original and compressed videos with three types of compression techniques such as MPEG1, MPEG2 and H264/AVC. Based on the experimental results, MD-QA metric demonstrates an efficient and accurate quality measure of 3D stereoscopic videos under H264/AVC compression.