Cauchy based Rate-Distortion Optimization Model for H.264 Rate Control

Base on the observation that Cauchy distribution provide accurate estimates of rate and distortion characteristics of video sequences, in this paper, we propose a Cauchy based rate-distortion optimization model for the application of H.264 bit allocation. Lagrange optimization is used as a tool to derive rate and distortion models subject to the target bit rate constraint resulting in optimum choice of quantization step sizes. Linear regression analysis is then used to update model parameters online. The technique proposed has been implemented in H.264 video encoder. Experimental results showed that the proposed rate control algorithm achieves an improvement of average PSNR of up to 0.68 dB with less PSNR variation compared to H.264 JM 8.6 rate control