No-reference model for video quality assessment based on SVM

This paper focuses on the video distortion which is caused by the packet loss. Considering the relationship between the human visual perception which is caused by the packet loss and the visual characteristic of the video content, we present a no-reference model for video quality assessment based on Support Vector Machine. The feature vector of the SVM contain temporal complexity, spatial complexity, the average number of bits per frame and the packet loss rate. Temporal complexity, spatial complexity and the average number of bits per frame represent the visual characteristic of the video content. The value of the packet loss rate means the distortion which is caused by the packet loss intuitively. Experimental results show that this model has a good consistency with the subjective.