No-reference Video Quality Assessment Based On Natural Scene Statistic

Video quality assessment has a very important role in video applications,such as video transmission quality monitoring,optimizing video processing algorithms and other parameters.The human eye is the main receiving terminal of the video,therefore,the most accurate video quality assessment method should be viewed by the people and the evaluation results are given.But the subjective evaluation method implementation process complexity,time-consuming and labor.Therefore,it is trying to assess video quality through objective algorithms.Current research on objective evaluation algorithms are mostly concentrated in full reference evaluation methods,despite the better performance of the whole reference quality assessment algorithm and the results of subjective evaluation method has a higher consistency,but because the calculation must obtain all the information of the original video,the use of Scene restricted.No reference video quality assessment algorithm due to its low cost,high real-time,portable and other characteristics attention.At present,some of the better performance of the no-reference video quality assessment algorithms,and others made by the Michele A.Saad of Video-BLIIND algorithm performance is better,this paper based on the algorithm of video quality evaluation methods were studied,the main work done Have:1.Based on statistical analysis of the natural time-varying image,perceived quality and extracted relevant spatiotemporal parameters of the model as an important feature of learning time domain.2.Extracted high-dimensional feature principal component analysis,while reducing the dimension characteristic parameters,on the other hand by the principal component analysis obtained the greatest impact on perceived quality characteristics and weighted.3.On the basis of Video-BLIIND on adding video distortion type judgment module,a two-step realization evaluation algorithm,the first step in determining the type of distortion,second step score prediction.With this design,the algorithm is convertible to known types of video quality distortion prediction algorithm not only improves the accuracy of the results,while improving the scalability and flexibility of the algorithm.4.Introduced video complexity characteristics.Considering the human visual masking phenomenon.