不规则图像灰度共生矩阵生成方式的比较  

Different Algorithms Comparison for Gray-level Co-occurrence Matrix of Irregular Graph

作　　者：史正星[1] 付强[2] 张国鹏[1] 卢虹冰[1] 

机构地区：[1]第四军医大学生物医学工程系,西安710032 [2]唐都医院泌尿外科,西安710038

出　　处：《医疗卫生装备》2010年第4期10-12,共3页Chinese Medical Equipment Journal

基　　金：国家自然科学基金项目(60772020)

摘　　要：目的:对不规则图形的灰度共生矩阵生成方法进行对比,通过实验总结出不同方法的适用范围。方法:勾勒图像中感兴趣区域(ROI),采用取原始图像中相应点法和邻域近似法分别生成灰度共生矩阵,并计算相应的纹理特征,将得到的纹理特征进行t检验。结果:对乳腺癌图像,2种不同方法生成的灰度共生矩阵没有表现出统计学差异;然而在膀胱癌组和正常膀胱壁组,统计显示不同方法生成的灰度共生矩阵存在差异。结论:在图像中ROI与周围区域不存在强对比且ROI边界点数量有限的情况下,2种方法差别不大,采用取原始图像中相应点灰度值的方法更简单;当ROI与周围区域存在强烈对比并且ROI较为狭长、边界点较多的情况下,邻域近似求解的方法更为适用。Objective Two algorithms were compared for calculation of Gray-level Co-occurrence Matrix（GLCM） in irregular images. Methods First, ROI （region of interest） area was segmented from whole images. Second, GLCM and texture features of each group were calculated. Third, a sample t-test between two groups which used different algorithms was performed, Results As to breast cancer images, there was no significant difference between the two algorithms. However, for bladder cancer images and normal bladder wall images, t-test showed significant difference between them. Conclusion When background is similar with the ROI area and ROI area isn＇t long and narrow, the first algorithm is better. On the contrary, when the ROI area show high contrast with the background and ROI area is long and narrow, the second algorithm is better.[Chinese Medical Equipment Journal, 2010, 31 （4） ： 10-12]

关 键 词：不规则图像 灰度共生矩阵 图像处理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]