Effective macroblock layer rate control algorithm for H.264/AVC

Rate control plays a critical role in the video encoder. In this paper, an effective macroblock (MB) layer rate control algorithm is proposed for H.264/AVC to improve video quality and reduce the computational complexity. By thoroughly analyzing the temporal–spatial correlation and object direction, the mean absolute difference (MAD) value of current MB is a weighted combination of temporal MAD predicted by previous frame and spatial MAD predicted by current frame, and then four directional patterns are chosen adaptively for data point selection in quadratic rate-quantization (R-Q) model parameters estimation. Finally, a more accurate header bits prediction model is developed to improve the accuracy of allocated bits which can be used for quantization parameter (QP) calculation. Extensive experiment results show that compared with JVT-G012, the proposed algorithm achieves higher average peak signal to noise ratio (PSNR) up to 0.51 dB improvements and more accurate target bit rates.

Figure optionsDownload as PowerPoint slideHighlights
► We propose a temporal and spatial weighted MAD prediction model.
► We develop a directional data point selection schemed for parameters updating.
► We present a header bits prediction model to obtain more accurate bits.
► The proposed algorithm achieves PSNR 0.51 dB improvements compared with JVT-G012.