Research On No-reference Video Service QoE Model Based On Natural Scenes Statistic

The explosion of mobile data traffic has been one of the most serious challenges faced by the current network environment,which is represented by the user demand to mobile video services.At the same time,intelligent mobile terminals and advanced network technology have also spawned growing user demands for higher quality video services.Therefore,how to ensure end-user's quality of experience becomes the key measure of the merits of network service,which urges improving the accuracy of video quality assessment(VQA)model.This thesis mainly conducts in-depth research on nature scence statistic based no reference video quality QoE model.The proposed model can accurately grab spatio-temporal statistics,especially the measurement of motion distortion in temporal domain.Simulation also proved the accuracy of its prediction ability.The thesis models a novel no-reference video QoE assessment method based on natural scenes statistic,which on the basis of the theory that human visual system(HVS)has formed adaptation to certain statistics during the long term evolution.The model makes use of the measurable deviations in spatio-temporal statistical regularities to describe the degree of video distortions,especially temporal statistic represented by motion information.Besides,this thesis further put forward another optimization model combined with human speed perception theory,where both the motion information contained in video content and the perceptual uncertainty as a inherent attribute of HVS channel have been taken into consideration.The quantitative values of these two indexes will be calculated to be the weighting factors of model based on NSS,which will make assessment more in line with user's subjective satisfaction score.Results show the that optimization model fully matchs to HVS's certain features and can run in high performance.