基于人工蚁群优化的矢量量化码书设计算法

本文提出一种基于人工蚁群优化的矢量量化码书设计新算法.该算法利用人工蚁群系统中蚂蚁通过信息素留存寻找最优路径的机制,结合单只蚂蚁通过拾起、放下物体从而使物体聚堆的行为模式,合理设计放下概率、禁忌列表、信息素更新方式以及相应的参数.与基于进化模拟退火和随机竞争学习的码书设计算法相比,本文提出的算法能获得性能较好的码书,其峰值信噪比比传统的LBG算法提高超过2dB.

Codebook Design for Image Vector Quantization with Ant Colony Optimization

A new codebook design algorithm is proposed for vector quantization image coding based on the ant colony optimization.Inspired by the ability of real ants to find the shortest path through the laying down of pheromone and to cluster corpse in the nest through the picking up and dropping of small objects by individual ant,we proposed the ant clustering based codebook design algorithm with emphasis on the appropriate design of the dropping probability,the tabu list as well as the way pheromone was updated,and the determination of the corresponding parameters.Experimental results are very encouraging and it is found that the new algorithm outperforms the latest evolutionary simulated annealing and stochastic competitive vector quantization codebook design,the improvement of pixel signal to noise ratio exceeds 2dB compared with the conventional LBG algorithm.