O.I. Sheluhin, A.S. Filinova, A.A. Tenjakshev

It was shown that various subject groups have different influence perceived quality of a video content. It was proposed to use calculation of temporal and spatial features for prognostication of video-s quality according to plot-s dynamics(i.e. spatial complication and temporal activity of the image on the basis of the analysis of impairment of index MOS. For data analysis and its classification by plots was used statistical method. With that end in view unique statistical movement features and color parameters were appointed in correspondence for each of content classes. It was showed that empirical distribution functions obtained for each videoclip belonging to definite plot have considerable differences and defined by appropriated distribution functions. As a result of researches it was showed that content classification for mobile streamed video is defined by five most occurred classes with different influence on user-s perception. Video stream-s bitrate increase mostly affect on static subject group-s video sequence-s quality, while there-s no marked gain for videos with moving elements. It was shown that proposed algorithm of automatic classification allows to estimating video-s quality within one class of a content.