竞争聚类耦合码字定向移动的图像重构矢量量化算法设计

为了解决模糊聚类-矢量量化算法都是强迫模糊分区转变为Crisp集合,降低码本质量,使重构图像丢失了丰富的边缘纹理信息;且其计算代价较高,以及严重依赖初始化等不足.提出了竞争聚类耦合码字定向移动的图像重构模糊矢量量化算法.引入竞争聚类,基于C均值和模糊C均值,设计最优目标函数;结合拉格朗日乘子,推导出最优目标函数的聚类中心和隶属度的求解模型;定义迁移规则,构造了码字定向移动机制,使得小簇群向大簇群移动,形成更大的簇群;以失真度和相似度为评估准则,构造量化反馈机制,优化重构图像.仿真结果显示:与其他矢量量化机制相比,本文算法的编码图像相似度更高,PSNR(peak signal to noise ratio)最高,得到的重构图像质量最佳;且本文算法的时间成本最低.

Optimization Design of Image Reconstruction Vector Quantization Algorithm Based on Competitive Clustering and Codeword Directional Movement

In order to solve these defects such as low codebook quality induced by forced fuzzy partition into the Crisp collection to result in losing rich texture information of reconstructed image,and the high computational cost,as well as heavily dependence on initialization.The fuzzy vector quantization algorithm based on competitive clusterring and codeword directional movement is designed.The optimal objective function is designed by introducing the competitive clustering and on the basis of C-means and fuzzy C-means to balance the efficiency and the performance of this algorithm.The migration rule is defined to propose the codeword directional movement strategy to make the small clusters move to large clusters for forming the bigger clusters.Taking the distortion and similarity as the evaluation criteria to construct the quantization feedback mechanism to optimize the reconstructed image.Finally,the performance of this algorithm is tested by simulation platform.And the results show that comparing with other vector quantization mechanism,the distortion value of this algorithm is less,the PSNR is higher to obtain the best reconstructed image quality,and the time cost is lowest.