A Novel Two-pass Vbr Coding Algorithm for the H.264/avc Video Coder Based on a New Analytical R-d Model

In this paper, the generalized Gaussian distribution is employed to model the 16(4*4) integer transform coefficients of the residue image in H.264 videos. Then the distortion−rate function of generalized Gaussian model is analyzed as well as a flexible rate−distortion (R−D) model is developed to approximate the actual R−D function. Finally, an effective variable bit rate (VBR) algorithm for H.264/AVC is proposed, which adopts two−pass encoding to achieve a constant video quality. Experimental results show that under the same average bit rate, the proposed algorithm achieves 94% to 99% reduction in PSNR variation compared to the model of JM9.8.