【作者】 张萍；

【导师】 魏平；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2011， 博士

【摘要】 随着当今视频技术的飞速发展,新的视频编码标准陆续提出,并得到广泛的应用。运动估计是视频编码系统的一个重要组成部分,可以有效去除视频序列相邻图像间存在的时间冗余,极大提高编码效率。然而运动估计庞大的计算量大大增加了视频编码系统的运算复杂度,因此寻找简单高效的快速运动估计算法一直是视频编码领域的研究热点。同时,新的视频编码标准中采用的新技术使得运动估计算法又面临新的挑战。如何有效的将快速运动估计算法和这些新技术结合,提高编码性能,是目前针对运动估计算法研究的重点。本文针对现有的快速运动估计算法的局限性,深入研究了群智能优化技术中的粒子群优化和生物地理分布优化,同时对H.264/AVC视频编码标准和可分级视频编码标准中的新技术进行了深入研究,主要的研究内容及创新点如下:1)提出了基于变异粒子群的快速运动估计算法。在粒子群迭代过程中加入变异操作,防止粒子群进化停滞,增加搜索能力。在利用粒子群算法进行全局搜索的同时,有效结合运动矢量的特性,选取合适的粒子种群,采用适当的终止策略,降低运算复杂度。2)提出了基于单纯形粒子群优化的快速运动估计算法。在基于变异粒子群的快速运动估计算法基础上,利用单纯形优化的局部搜索性能扩展搜索能力,克服粒子群进化的早熟收敛缺陷,进一步提高搜索精度。3)提出了基于混沌的生物地理分布优化算法。针对生物地理分布优化算法早熟收敛的缺陷,利用混沌初始化种群,提高种群的遍历性,同时在迭代过程中加入混沌搜索,有效避免陷入局部最优,提高全局搜索能力。4)提出了基于生物地理分布优化的快速运动估计算法。利用生物地理分布优化算法优越的全局优化能力和混沌算法精细的局部搜索能力,有效结合运动矢量的特性,选取合适的初始种群,采用相同点检查和适当的终止策略,提高搜索精度降低运算复杂度。5)提出了H.264/AVC编码标准中基于生物地理分布优化的自适应快速运动估计算法。根据H.264/AVC中运动估计的新技术,结合H.264/AVC编码特点和生物地理分布优化特性,采用自适应搜索策略、动态搜索范围技术和自适应提前终止策略,在保证编码器原有失真度的前提下,节省运动估计的搜索时间,提高整体编码效率。6)提出了可分级视频编码标准中针对增强层的自适应快速运动估计算法。根据可分级视频编码中的层间预测技术,结合可分级编码标准的编码特点,充分利用增强层和基本层的相关性,选择合适的增强层的预测运动矢量,适当调整增强层的搜索策略和搜索范围,根据层间残差预测特点进一步调整运动估计步骤,在保持增强层的率失真性能的同时,降低增强层的编码复杂度,提高整体编码效率。更多还原

【Abstract】 With the fast growing of today’s video technology, new video coding standards have been continuously proposed and widely applied. Motion estimation, as an important part of video coding system, can greatly improve the efficiency of coding by reducing the temporal redundancy of adjacent picture in video sequence. However, it will introduce huge amount of calculation into video coding system. Looking for simple and efficient fast motion estimation algorithm is a research hotspot in the field of video coding. Meanwhile, the application of new techniques in new video coding standards is another challenge. How to combine fast motion estimation algorithm with new technologies to improve coding performance becomes the focus of present motion estimation algorithm research.This dissertation focuses on the limitations of previous fast motion estimation algorithms, studies the particle swarm optimization and biogeography-based optimization of swarm intelligence optimization in a deep-going way, and researches into new technologies which are applied to both H.264/AVC video coding and scalable video coding. The main research contents and innovation points are as follows:1) A fast motion estimation algorithm based on mutation particle swarm is proposed. Mutation operation is added into iteration of particle swarm to prevent the stagnation of particle swarm evolutionary and increase searching ability. Appropriate particle swarms and termination strategy are adopted based on motion vectors properties during global search of particle swarm algorithm to reduce computational complexity.2) A fast motion estimation algorithm based on simplex particle swarm optimization is proposed. In order to further improve the searching accuracy, premature convergence of particle swarm evolutionary is overcome by extensional search capability of simple method based on the mutation particle swarm fast motion estimation algorithm.3) The biogeography-based optimization algorithm based on chaotic is proposed. In view of the premature convergence of biogeography-based optimization algorithm, ergodicity of the populations is improved by utilizing chaos initialization and global search capability is improved by adding chaotic search during iteration to avoid local optimal.4) A fast motion estimation algorithm based on biogeography-based optimization is proposed. Searching accuracy is improved and computational complexity is reduced by adopting similarities check and proper termination strategies and combining the global optimization capability of biogeography-based optimization and fine local search ability of chaotic algorithm with properties of motion vectors to select appropriate initial populations.5) An adaptive fast motion estimation algorithm based on biogeography-based optimization in H.264/AVC coding standards is proposed. Based on the new motion estimation technologies of H.264/AVC, combined with the characteristics of H.264/AVC encoding and biogeography-based optimization, and adopting the adaptive searching strategy, dynamic searching range and adaptive-advanced terminal strategy, motion estimation search time is saved and overall coding efficiency is improved on the guarantee of the original distortion degree.6) An adaptive fast motion estimation algorithm for enhancement layer in scalable video coding (SVC) is proposed. According to the inter-layer prediction technology of SVC, combined with the encoding characteristics of SVC and making full use of the correlation between enhancement layer and base layer, proper prediction motion vectors for enhancement layer are selected. The search strategy and search range are adjusted appropriately and the process of motion estimation is also changed base on inter-layer residual prediction.Thus, code complexity of enhancement layer is reduced on the guarantee of the rate-distortion performance and overall coding efficiency is improved.更多还原