关键词: 视频编码失真, H.264/AVC, 高斯分布, 直流系数, 拉普拉斯分布

Abstract:

A novel video coding distortion estimation algorithm is proposed for H.264/AVC. First, a new linear method is used to compute the variance of the direct current (DC) coefficient. Then, with the assumption that the DC coefficient follows a Gaussian distribution, the distortion of the Gaussian source under the quantization of the dead-zone plus-uniform threshold quantization (DZ-UTQ) with the uniform reconstruction quantizer (URQ) is derived, and according to the distortion of the Gaussian source and the variance of the DC coefficient, the coding distortion of the DC coefficient called DC distortion is calculated. Thirdly, a new function is employed to compute the ratio of video coding distortion to DC distortion. Finally, video coding distortion is estimated using the ratio and DC distortion. Compared with the algorithm with the best performance in the existing algorithms, the proposed algorithm decreases the averaged estimation error by up to 36％. Experimental results show that the proposed algorithm can provide the estimation of video coding distortion more accurately than the existing algorithms.

Key words: video coding distortion, H.264/AVC, Gaussian distribution, direct current coefficient, Laplacian distribution