Study on rate-quantization models based on main DCT coefficient distribution models

Rate-distortion (R-D) model that includes rate quantization (R-Q) and distortion-quantization (D-Q) models is very important in rate control and mode decision for video coding algorithm optimization. R-D model can be derived by rigorous mathematic analysis based on DCT coefficient distribution. Thus, the statistical distribution of DCT coefficients is very important for R-D modeling. Laplacian, Cauchy, and Generalized Gaussian Distribution (GGD) are three main distribution models. Some R-D models based on these three models were proposed in the literature. However, there is no systematic and rigorous comparison on their accuracy currently. In this paper, we had examined three coefficient distribution models for accuracy comparison. Also, we had examined main R Q models based on these three models. This work is meaningful for more accurate R-Q modeling in the future for efficient video coding algorithm optimization.