Study On Video Quality Assessment For No-reference PSNR Methods

Rapid development of technologies in multimedia and video compression has led to proliferation of video applications in our daily life. The digital video quality assessment is widely used in various video applications, especially the no-reference video quality assessment plays an important role in network communication application. Therefore, there is an acute need for study and researches on video quality assessment.In this dissertation, the factors which have influences on video quality, especially the quantization parameter are analyzed firstly. Then the existing video quality assessment methods are generally summarized, where the No-Reference video quality assessment methods are discussed in detail. The peak signal-to-noise ratio (PSNR) is wildly used for measuring video quality, which is restricted from application as it requires the original video information.A no-reference method for estimating PSNR is proposed for the video coding standard H.264 in this dissertation. The statistical characteristics of the discrete cosine transform (DCT) coefficients are firstly estimated by maximum likelihood criterion. The quantization distortion is then obtained by combining the quantization parameter information extracted from a video bit stream. This method does not require complete or partial video decoding, which can be used for monitoring the networked video. Experiment results show that the estimated PSNR is consistent with the actual PSNR, especially when the value of quantization parameter is low or medium.