Studies Of Motion Estimation Algorithms Based On Swarm Intelligence Optimization

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.