基于GLCM与自适应Gabor滤波器组的纹理图像分割  

Texture Image Segmentation Based on GLCM and Self-adaptive Gabor Filter Bank

作　　者：闵永智[1] 程天栋 殷超[1] 岳彪 肖本郁 马宏锋 

机构地区：[1]兰州交通大学自动化与电气工程学院,兰州730070 [2]兰州工业学院电子信息工程学院,兰州730050

出　　处：《计算机工程》2017年第1期280-286,共7页Computer Engineering

基　　金：国家自然科学基金"基于机器视觉的铁路轨道表面缺陷快速识别与分类方法研究"(61461023);国家自然科学基金"铁路长大隧道路基表面沉降链式图像监测方法及模型"(61663022);甘肃省高原信息工程及控制重点实验室开放课题基金"钢轨表面缺陷机器视觉快速检测"(20161105)

摘　　要：基于Gabor滤波器的纹理图像分割算法存在参数难以选择的问题。为此,提出一种预测图像纹理类型数与Gabor滤波器组参数的分割算法。将图像分割成大小相等的区域块,根据各类纹理特性预测Gabor滤波器组参数,利用各区域块的纹理特征向量预测纹理类型数,并使用预测的滤波器组提取图像纹理特征,通过预测的纹理类型数对图像进行聚类分割。实验结果表明,该算法能以较高的精度与较快的速度分割纹理图像,且受纹理类型数量影响较小。To solve the problem of parameter selection in the algorithm of texture image segmentation based on Gabor filter,a texture image segmentation algorithm is proposed in this paper,which predicts the number of texture types and the parameters of Gabor filter bank.Firstly,the image is divided into regional blocks.Then,the number of texture types is predicted by the texture feature vector of regional blocks,and the parameters of Gabor filter bank are predicted by the characteristics of various texture features.Finally,texture features of the original image is extracted by using the predicted filter bank,and the image is clustered and segmented based on the predicted number of texture types.Experimental results show that the proposed algorithm can process the segmentation in the texture image with higher precision and faster speed,and is less affected by the number of texture types.

关 键 词：GABOR滤波器 纹理图像 纹理类型 灰度共生矩阵 模糊C均值聚类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]