一种快速检测纹理斑块的方法

提出一种检测一致纹理图像中不一致纹理斑块的快速算法。该算法利用图像中每个位置的代表块与纹理原型的距离作为纹理相似性测度，并在此基础上进行纹理分类。这里，采用了统计中广泛使用的Kolmogorov-Smirnov距离，并研究了纹理原型的估计方法，以及为加快搜索速度而采用的层次搜索算法。实验表明了该方法的有效性。

An efficient algorithm for texture fault detection

A method for fast detection of texture faults in a texture image is presented. Based on the distance of the representative block of each site in the image and the texture prototype, which measures the similarity of two textures, we can do texture discrimination. Here we use the Kolmogorov Smirnov distance well known in statistic. We also study the estimation method of the texture prototype, and develop a hierarchical method for quick search. The experiment results show the efficiency of this algorithm.