基于类别共生矩阵的纹理疵点检测方法 邹超 朱德森 肖力

摘要：根据有规则纹理的特点，提出了基于类别的共生矩阵来描述纹理特征，从而很好地将正常纹理与疵点区分开。分析了传统的灰度共生矩阵在计算纹理特征时计算量大，且分辨能力差的缺点．为了克服灰度共生矩

阵在计算量和分辨能力上的缺点，定义了类别共生矩阵．在类别共生矩阵的算法中，首先学习纹理的一些基本特征以确定类别共生矩阵的一些关键参数。如纹理的概率密度分布、纹理的主方向和周期，以及分类准则等重要参数，然后计算类别共生矩阵并提取白疵点增强、黑疵点增强和一致度等三个特征，最后采用异常点检测的方法即可很好地区分正常纹理和疵点．实验证明，该方法比已有的灰度共生矩阵计算量小，并具有更突出的分辨纹理和疵点的能力．-Class-Based Co-occurrence Matrix Texture defect detection method ZOU Chao Zhu Sen Xiao Li

Abstract: According to the rules-texture features is proposed based on categories of co-occurrence matrix to describe the texture features, which will be well distinguished from the normal texture with defects. Analysis of the traditional GLCM texture features in the calculation for calculating the volume, and the disadvantage of poor resolution. In order to overcome the symbiotic Gray Moment

Array in the calculation of capacity and the ability to distinguish the shortcomings of the definition of the categories of co-occurrence matrix. In the categories of co-occurrence matrix algorithm, first learn some basic features of the texture in order to determine the categories of some of the key parameters of co-occurrence matrix. Such as the texture of the probability density distribution of the main texture direction and cycles, as well as the classification criteria and other important parameters,