基于共生矩阵纹理特征提取的改进算法

摘要

After studly for GLCM and its improved algorithm, especialy for its excessive computational burden, a novel method based on grey level co-occurrence hybrid structure (GLCHS) and discrete Fourier transform is presented to achieve texture feature extraction. First,the result image is devided into several blocks which can reduce the number of gray levels. Then, gray normalization is done to reduce the range of the gray value. Finally, the GLCHS is used to compute the five dimensional vectors to describe the image' s texture. The experiment results indicate that the improved method reduces the computational complexity and greatly reduces the time of feature extraction.%深入研究灰度共生矩阵及其改进算法,针对其计算量大、耗时等问题,提出一种基于灰度共生混合结构和离散傅立叶变换的方法来实现纹理特征的提取.对傅立叶变换后的频谱图进行分块计算,以此降低计算时的灰度级,再采用正规化的方式减少特征的分布范围,并利用灰度共生混合结构算法计算一个5维的特征向量来描述图像的纹理.实验结果表明,改进后的算法降低了计算复杂度,极大地减少了图像纹理特征提取的时间.