使用进化策略的图像矢量量化

在分析自组织特征映射（SOFM）算法的基础上，把进化策略引入到矢量量化设计中，在将SOFM算法作为基本聚类算法的前提下，基于等误差原则，利用最大最小原则改变各区域聚类点，使各子区域的误差趋于相等。使用进化策略在子区域误差大的聚类点附近产生新的聚类点，从而调整各个区域的子误差，进一步改善总期望误差。实验证明该方法可克服基本聚类算法落入局部最化的局限。在对Lena图像进行量化的过程中，改善了码书矢量，提高了码书矢量，提高了量化后图像的质量。

Design of Image Vector Quantization Using Evolutionary Strategy

This paper analyzed self organizing feature map (SOFM) algotithm. Based on it, the evolutionary strategy into vector quantization was introduced. After SOFM algorithm is used to decrease the expected distortion, the evolutionary strategy is utilized to modulate the subdistortion of each region in order to improve the expected distortion. The experimental results show that the algorithm can well modulate the subdistortion of each region by overcoming the local optimality.