Continuous stereoscopic video quality evaluation
Z. M. Parvez Sazzad, S. Yamanaka, Yuuko Horita · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2010
This research aims to develop an objective no-reference video quality evaluation method for MPEG-2 MP@ML coded (symmetric and asymmetric) stereoscopic videos. Our proposed method is based on segmented local features of spatial artifacts, disparity, and temporal activities of videos. Segmented local features information such as edge and non-edge areas of any stereoscopic pair frames (i.e., left and right views) have taken into consideration for blockiness and zero crossing. In this method, a temporal segmentation approach is considered and each temporal segment is evaluated for artifacts and disparity. Temporal features are calculated separately for left and right video sequences based on segmented local features and sub temporal segment. Different weighting factors are also applied to measure the spatial artifacts, disparity, and temporal features of the segment. In order to verify the performance, we conducted subjective experiment on different symmetric and asymmetric coded (Bit rates: 2, 3, 5, and 8 Mbps) stereo video pairs. An auto stereoscopic display was used for fifteen (15) reference stereo videos; each of the video was 15 seconds length and the total length of each test sequence was (15×15 sec = 3 min 45 sec). Seven video sequences were used to complete the experiment. The Single Stimulus Continuous Quality Evaluation (SSCQE) method was used to conduct our subjective experiment. The experiment result indicates that our proposed method has given sufficient prediction performance.