粒子群优化算法在图像矢量量化码书设计中的应用  

Codebook design of image vector quantization based on particle swarm optimization

作　　者：张绪冰[1] 关泽群[1] 徐景中[1] 

机构地区：[1]武汉大学遥感信息工程学院,武汉430079

出　　处：《计算机应用》2007年第12期3051-3054,共4页journal of Computer Applications

基　　金：教育部长江学者和创新团队发展计划资助项目(IRT0438)

摘　　要：鉴于经典的LBG码书设计算法易陷入局部最优解,首次采用粒子群优化算法来设计图像矢量量化的最优码书,并提出了粒子群矢量量化(PSO-VQ)算法和粒子一致性操作(PCO)。在PSO-VQ算法中,每个粒子表示一个码书,以粒子群进化的方式对初始码书进行迭代而获得最优码书,PCO操作对各初始码书中的码矢量按其灰度均值排序,使不同码书的内部结构基于码矢量灰度均值达到基本一致,确保了结果向全局最优解收敛。实验证明,PSO-VQ算法在解码图像的PSNR值和主观效果上都优于LBG算法,同时拓展了粒子群优化算法的应用领域。The LBG algorithm depends upon the initial codebook and is prone to converge to a local optimal solution. To solve this problem, Particle Swarm Optimization （PSO） was adopted to design the optimal codebook of image Vector Quantization （VQ） and PSO Vector Quantization （PSO-VQ） algorithm was presented. According to PSO-VQ, a particle indicated a codebook, and the optimal codebook was obtained from iterations of the initial codebooks by method of the particle evolvement. To ensure the solution converge to the global optimal codebook, the Particle Coherent Operation （PCO） was also proposed, by which the code vectors of each initial codebook were sorted in ascending order based on the gray mean, and so that the inner structures of all the particles were essentially identical. The experimental results show that the PSO-VQ algorithm is better than LBG in terms of the PSNR and subjective effect of the decoded image. Meanwhile, it extends the application of the PSO.

关 键 词：粒子群优化 矢量量化 码书 图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TN919.81[自动化与计算机技术—计算机科学与技术]