Optimal bit allocation and efficient rate control for H64/AVCsed on general rate-istortion model and enhanced coding complexity measure

摘要

A general rate-distortion (RD) model for block-based video coding is proposed in this study. Based on this model, an innovative coding characteristic called texture-complexity (TC), which is defined as a function of discrete cosine transform (DCT) coefficient distribution and distortion-quantisation (DQ) relationship as well as variance of residual block, is presented for rate control. Moreover, an adaptive linear TC forward-prediction model is established to estimate the characteristic of incoming image block, and a head bits prediction model as well as an optimal bits allocation method are also proposed to best allocate available bits. Experimental results show that the H.264 encoder (JM15.1), using the proposed rate control algorithm, achieves an objective quality improvement up to 0.434 dB, produces flatter bit-rate/PSNR curve and meets closer target bits budget than outcomes of JM15.1 rate control method (JVT-W042).