Adaptive rate control for H.264

This paper presents a rate control scheme for H.264 by introducing the concept of basic unit and a linear prediction model. The basic unit can be a macroblock (MB), a slice, or a frame. The linear model is used to predict the mean absolute differences (MADs) of the remaining basic units in the current stored picture by those of the co-located basic units in the previous stored picture. The target bits for the current stored picture are computed by adopting a fluid flow traffic model and linear tracking theory, and are further bounded by two values that are derived by taking the hypothetical reference decoder (HRD) into consideration. The remaining bits are allocated to the remaining basic units in the current stored picture according to their predicted MADs. The corresponding quantization parameter is computed by using a quadratic rate-distortion model. The rate distortion optimization (RDO) is then performed for all MBs in the current basic unit by the quantization parameter. Both constant bit rate and variable bit rate cases are studied. The average PSNR is improved by up to 0.8 dB for an encoder using our scheme compared to an encoder using a fixed quantization parameter. With our scheme, an H.264 encoder can be adaptive to time varying channel bandwidth that is available for the coding process.