【作者】 陈庆；

【导师】 张明；

【作者基本信息】 浙江大学 ， 信息与通信工程， 2008， 硕士

【摘要】 在先进的H.264/AVC视频编码标准中,去块滤波技术能有效提高编码效率和图像的主客观质量,但是滤波处理性能的进一步提高却受制于其高度的计算复杂度。在面向未来的多视点视频编码MVC中,沿用了H.264/AVC的去块滤波边界强度判决算法,但随着光照补偿技术的引入,该算法呈现出了不足之处。本文对去块滤波进行了深入的分析和研究,为了进一步提高滤波处理性能,提出了H.264/AVC去块滤波优化的ASIC设计;针对MVC去块滤波已有的判决算法的不足,提出了更有效的滤波判决算法。一方面,本文在H.264/AVC去块滤波的ASIC设计中,通过研究滤波边界的特点,提出了优化的行列分组处理的滤波顺序;为了减少级间参数缓存,提出了在熵编/解码级就计算出滤波控制参数的方法;为了实现并行滤波和提高处理速度,设计了五级流水线边界滤波器;此外,通过优化转置单元的移位逻辑,复用流水寄存器组以及优化四舍五入运算的实现,有效地减小了硬件开销。实验结果表明,与已有的设计相比,提出的去块滤波硬件结构具有更好的综合性能。另一方面,本文研究了MVC的光照补偿技术,通过对残差变换系数编码过程的深入分析和公式推导得出,光照补偿的实质在于对残差直流系数量化编码之前额外进行了预测编码。针对光照补偿过程产生块效应的原因,提出了优化的滤波边界强度判决算法。该算法抓住了光照补偿的本质,简化了判决流程,能与H.264/AVC滤波判决算法保持一致。实验结果表明,新的判决算法较好地提高了图像主客观质量,与已有的MVC判决算法相比,不但算法复杂度较低,而且进一步节省了比特率。更多还原

【Abstract】 In the advanced video coding standard H.264/AVC, deblocking filter is utilized to effectively improve coding efficiency and enhance both subjective and objective picture quality. However, the further enhancement of the filtering performance will be restricted by its high computational complexity. The boundary strength (bS) decision algorithm of H.264/AVC deblocking filter is adopted in the future-facing MVC (multi-view video coding). Unfortunately, with the illumination compensation technology introduced into MVC, the traditional decision algorithm shows its disadvantages.This dissertation does thorough research on deblocking filter, and proposes an optimized ASIC design of H.264/AVC deblocking filter aiming at further improving its filtering performance. Furthermore, in order to make up for the weaknesses of the conventional algorithm, this dissertation proposes a more effective decision algorithm for MVC deblocking filter.On one hand, in ASIC design of H.264/AVC deblocking filter, this dissertation proposes an optimized filtering order for vertical and horizontal edges grouping processing by studying the filtering edge features. In order to save inter-stage parameter buffer, it proposes the method of calculating filtering control parameters in entropy coding stage. A five-stage pipelined edge filter is presented to implement the parallel filtering and increase processing speed. And the hardware cost is efficiently decreased by optimizing shift logic of the transposer, reusing pipeline register set and optimizing the implementation of rounding operation. The experimental results show that, compared with state-of-the-art designs, the proposed architecture has better comprehensive performance.On the other hand, this dissertation studies on the illumination compensation of MVC. Through intensive analysis and formula deduction on the coding process of transformed residual coefficients, it reveals that the essence of illumination compensation is to add predictive coding on the residual DC coefficients before its quantization. According to the cause of block artifact resulted by illumination compensation, it presents an optimized bS decision algorithm. The proposed algorithm grasps the essential feature of illumination compensation, simplifies the decision procedure, and keeps consistent with bS decision algorithm in H.264/AVC. The experimental results demonstrate that, the new decision algorithm can improve both subjective and objective visual quality. With comparison to existing MVC decision algorithm, the proposed algorithm not only has lower computational complexity, but also achieves further bit-savings.更多还原