Enhanced Linear R-Q Model Based Rate Control for H.264/AVC Using Context-Adaptive Parameter Estimation

We present an efficient and accurate macroblock layer rate control for H.264/AVC using a context-adaptive parameter estimation scheme. Our enhanced linear rate quantization (R-Q) model based rate control, which simultaneously controls texture and non-texture bits, is developed using an approximation of the theoretical rate distortion (R-D) function. A second key component is a novel context-adaptive prediction model for improving the estimation accuracy of both the mean absolute difference (MAD) of texture bits and the other model parameters. This prediction model maintains high spatial and temporal correlations among the neighboring macroblocks within the context using a distance metric and an improved sliding window. Compared to JM 12.2, the proposed algorithm achieves up to 0.98 dB improvements in PSNR and bit mismatch has also been significantly reduced.