Video Quality Assessment Combining Danmuku And Saliency Regions

Currently,video quality assessment methods play an important role in the field of video analysis and have attracted widespread attention.There are two methods for video quality assessment: subjective and objective quality assessment.Among them,objective quality assessment methods have significant advantages,especially in terms of processing speed and accuracy based on deep learning.However,the objective quality assessment method based on deep learning has the limitation of poor interpretability of experimental results,and lacks the participation of human thinking and experience.Therefore,this paper focuses on the interpretability and processing speed and accuracy of video quality assessment.Firstly,the interpretability of existing objective quality assessment methods is poor,and human thinking and emotions have not been considered in the evaluation method,which reduces the participation of subjective factors.Therefore,the danmaku,which is full of human thinking and emotion,is added to the video quality assessment.Danmaku is a kind of real-time comment text produced by the viewer during watching the video.This experiment obtains the number of danmaku in the video,analyzes the distribution trend of the number of danmaku,and then counts the release time of each danmaku.Secondly,in order to further speed up the speed of video quality assessment,considering that the danmaku itself contains thoughts and emotions,we can conduct sentiment analysis on the content of the danmaku,so as to select the time points in the video with prominent sentiment changes,and extract the corresponding video keyframes according to the corresponding time points.Finally,when studying video keyframes,human visual features are added to facilitate the extraction of visual saliency regions from keyframes and to evaluate the video quality of the obtained saliency regions.This article proposes a method for emotional analysis of danmaku content and selecting visual saliency regions of keyframes for video quality assessment.Experiments are designed based on the proposed method.And the effectiveness and accuracy of the proposed method were verified by comparing it with subjective quality assessment and other scholars’ methods.