Research Of Frame Bit Allocation Via Improved Weibull-Distribution-Based Rate And Distortion Models

Over the past few decades, transform-based compression for image and video sources has gained widespread popularity for visual information management, processing, and communications. As a result, several video standards have been developed. H.264/AVC video standard has become most popular up to now.In H.264/AVC video standard, Laplacian distribution model is used to describe the distribution of DCT coefficients. But Laplacian distribution model can not be well to approach the tailing of DCT coefficients distribution which is based on the rate-distortion model and corresponding rate control algorithm have some defects. Therefore, a new mathematic model need to be found to approach DCT coefficients distribution better so that more clearly and smoothly quality video output could be gotten. The most approaching model is Cauchy distribution at present, but the model also can not be well to approaching the tailing of DCT coefficients distribution sometimes.The thesis deeply researches the key technologies of H.264/AVC video standard. Through the statistical analysis of typical video source, statistical characteristics of DCT coefficients and the disadvantages of rate-distortion model, Improved Weibull distribution is proposed and a new rate-distortion model is built on. The thesis uses Improved Weibull probability density function to approximate actual DCT coefficients distribution. With encoding entropy theory and distortion analysis, the model is expressed as two functions of the video coder quantization step size, which are rate-quantization model (R-Q) and distortion-quantization model (D-Q) that are based on the distribution, and the model which is based on Improved Weibull distribution is to estimate encoding rate and quantization distortion. Simulation experiments results show that the density distribution is closer to the statistical distribution of the DCT coefficients than Cauchy density distribution. The model which is based on Improved Weibull density distribution provides better estimates for the actual encoding rate and quantization distortion performance than those that are based on Cauchy density distribution.The rate-distortion model which is based on Improved Weibull distribution can describe the relationship between rate and quantization step size, as well as the relationship between distortion and quantization step size more accurately than that which is based on Cauchy distribution, because Improved Weibull distribution is closer to the statistical distribution of the DCT coefficients than Cauchy distribution. A new frame bit allocation algorithm is proposed for the H.264 based on Improved Weibull distribution. Simulation experiments results show that a 0.30-dB average peak signal-to-noise ratio (PSNR) improvement over the JVT-H017 rate control algorithm and more smooth video quality could be obtained. The results also show that lower actual output rate and more smoothly buffer status that avoids overflow and underflow conditions could be obtained.