基于灰度共生矩阵的旋转不变性改进算法

Improved Algorithms of Rotation Invariance Based on GLCM

基于灰度共生矩阵纹理特征提取算法,提出了针对其旋转不变性的改进算法。引入"伪像素点"的概念,使用"伪像素点"进行像素对灰度值统计,提高了纹理特征旋转不变性;在原始算法四个方向的基础上,增加了四个方向,使得纹理特征旋转不变性进一步提高。使用公共图像库对原始算法及两种改进算法进行实验比较,结果表明。

Based on the gray level co-occurrence matrix （GLCM） feature extraction algorithm,two improved algorithms are proposed in this paper.It introduces a concept Faked Pixel,count the gray level of pairs of pixels using Faked Pixel,the rotation invariance is improved.On the basis of the original algorithm,four other directions are added,the rotation invariance of textural features is improved ulteriorly.Two improved algorithms are compared with the original algorithm using public images sets.