Skip to Main Content
Summary form only given. This paper presents a simple, fast and accurate rate control algorithm using conditional mean estimator (nonlinear regression) that plays a central role in estimation theory. Central to nonlinear estimation and stochastic control problems is the determination of the probability density function of the state conditioned on the available data. If this a posteriori density function is known, then an estimate of the state for any performance can be determined. The proposed algorithm measures this conditional mean by estimating a joint probability density function (PDF) using Parzen's window by extending it to multivariate case. We use this window function to estimate a joint PDF using long training data. The training data pick up the joint PDF between the quantization parameter (QP) and the bits spent for each macroblock depending on the sum of absolute differences (SAD) value from motion estimation. Since the SAD information is obtained as by-product of motion estimation, the additional complexity is minimal. We increase the accuracy of this joint PDF by clustering the training data depending on the QP values within admissible ranges. This localization helps understand image characteristics more accurately. Then we apply the adaptive vector quantization to simplify the conditional mean estimation of the rate given the SAD and QP values. This information is stored into three look-up tables for I, P and B pictures. They contain the localized R-D function on macroblock basis. We use these tables to find the optimal QP values in least-mean-square (LMS) sense for a given bit budget of the current frame. We compared our proposed algorithm with the MPEG-4-rate control algorithm (Q2). Simulation results show that the proposed algorithm outperforms the informative MPEG-4 rate control algorithm in terms of reproduced image quality and coding efficiency while requiring much less implementation complexity.