No-reference PSNR Estimation Based on HEVC Coding

Abstract

With the continuous rising of video applications, the compression of video information has become necessary. As the newest video coding standard, HEVC achieves twice the compression ratio of H.264 under the same video quality. But meanwhile, it produces certain quality impairments, so it's necessary to do video quality evaluation on HEVC coding. This paper mainly refers to the no-reference algorithm based on the peak signal-to-noise ratio (PSNR) from ZhengYuan. The algorithm uses the Laplace distribution to build model for discrete cosine transform (DCT) coefficients. Moreover, since HEVC adopts four sizes of transform block, the algorithm uses the pixel block size weighted method to estimate quantization error to replace the pixel domain mean square error (MSE). Then compute the estimated peak signal to noise ratio (PSNR). Our experimental results shows that the estimated PSNR can reflect video quality well and has good accuracy, correlation and monotonicity with the actual PSNR.