基于人类视觉特性的纹理分割方法  

Adaptive texture segmentation method based on property of HVS

作　　者：毕胜[1] 梁德群[1] 

机构地区：[1]大连海事大学信息工程学院,辽宁大连116026

出　　处：《计算机应用》2006年第5期1015-1017,共3页journal of Computer Applications

基　　金：交通部博士基金资助项目(200332522501)

摘　　要：纹理分割是将一幅图像依据纹理不同分成若干个不同的区域,目前广泛采用的是利用滤波器族(如Gabor)对图像进行分解。但由于图像纹理表现的各异性,通常在滤波器参数的选择上不能做到自适应,导致提取的特征不明显,分类效果不好,使用范围受限。文中提出了一种基于人类视觉系统(HumanVisualSystem,HVS)机理的纹理分割方法,不但可以模拟人类观察纹理直觉处理阶段确定纹理区的个数并且粗略地划分区域,而且可以模拟专注处理阶段自动选择Gabor滤波器的个数及参数。该算法符合HVS区分纹理机理,计算过程简单、方便。针对各纹理选取的特征明显,分类效果好。Texture segmentation involves subdividing an image into differently textured regions. Many texturesegmentation schemes are based on a filter-bank model, e.g. Gabor filter-bank. But these algorithms usually are limited because of the texture＇s variety and the parameters selection difficulty. These deficiencies lead to the ineflieient extracted feature, the worse classification results and the limited application domain. In this paper, a texture segmentation algorithm based on the HVS mechanism was presented. It earl give the number of texture type and approximate areas in the image mimicking the preattentive stage, and can also give the Gabor filters parameters mimicking the focused attention stage. This algorithm corresponds to the HVS property, and the approach is easy to implement and eomputationally efficient. The features are characteristic, and the classification results have higher accuracy.

关 键 词：人类视觉系统 纹理分割 GABOR滤波器 纹理特征度量 滤波 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]