The Study Of Video Coding Quantization Optimization

The international standard H.264/AVC and national Audio Video coding Standard (AVS) provide high compression efficiency. They are widely applied in digital TV, multimedia communication, video surveillance and so on. Quantization is a key component in picture/video coding schemes to provide coding efficiency, but it also introduces quantization distortion which causes degradation of quality of reconstructed picture/video.In this thesis, we first review the basic principle of quantization and some classical quantizers in video coding. Then we introduce several optimized quantization algorithms which are widely used in video coding systems:frame-level bit-rate allocation and quantization step size adjustment, macroblock-level quantization step size adjustment, subband-level quantization step size adjustment, adaptive quantization and reconstructed level (adaptive rounding) and Rate-Distortion optimized transform-coefficient-level adaptive quantization and reconstructed level.A rate-distortion optimized transform-coefficient-level quantization technique, which includes a computation method of distortions for non-normalized transform coefficients of AVS directly in transform-domain, is proposed in this thesis. By adjusting dead zone and rounding parameter of quantizer for each coefficient, this technique can achieve better performance in terms of both compression efficiency and video quality; by calculating distortion in transform-domain, this technique can be implemented with lower complexity. Considering of the 2-dimentional variable length coding defined in AVS, the rate-distortion optimization of quantization can be applied for one coefficient at a time or several coefficients jointly. Additionally, this thesis also discusses a rate-distortion optimized quantization technique for AVS' context-and-logarithm based advanced entropy coding of transform coefficients.