Video Quality Assessment Algorithm Of H.264Based On Region Of Interest

With the technology of online video and video telephony having been inwidespread, the automatic evaluation of video quality is particularly important inthe design and monitoring of video transmission systems. The result of subjectivevideo quality assessment is more reliable, but this method is time-consuming andwastes more material resources. Therefore, objective video quality assessment playsan irreplaceable role in the field of video. Compared with FR and RR model, NRmodel does not depend on any information of test video, and has the advantagessuch as well real-time and large range of application. Therefore methods ofno-reference video quality assessment are researched in this paperThis paper introduces subjective methods and objective methods of videoquality assessment, and studies the basics of human visual system and digital videocompression. Then the basics of Human Visual System and digital videocompression coding are studied. The steps are as follows: Firstly, from the globalperspective, with the research of the relation between the characteristics on singleframe image, the features of content distortion degree, motion degree and DMOS,and then these three features are extracted. Secondly, with the studies of the codingstreaming structure for H.264, and the analysis of the relationship between thestream information and visual attention, then the motion vector and the encodingmode are to extract the region of interest of the video. Thirdly, an objectivevideo quality assessment model is presented which based on features extracted atthe time-space domain and degree of movement from video. The model extracts thecharacteristics on single frame image, content distortion degree, motion degree andthe bits message of the macro block in region of interest. Finally, the simulationresults and experimental data for the model are showed, and they are compared withthe classic full reference video quality assessment method. The results show that themodel is low computational complexity, good real time, and have a betterassessment of performance.This method is based on region of interest, and it is more real-time because thefeatures of region of interest are extracted before video decoding. With simpleassessment method, a good evaluation of sensitivity, the model is applicable toreal-time multimedia communications, not only H.264videos, but also based onregion of interest encoded videos.