纹理模糊的图像分割  

Image Segmentation on Texture Blurring

作　　者：何坤[1] 郑秀清[2] 张永来[2] 

机构地区：[1]四川大学计算机学院,四川成都610065 [2]四川师范大学信息技术学院,四川成都610066

出　　处：《四川大学学报（工程科学版）》2015年第4期111-117,共7页Journal of Sichuan University (Engineering Science Edition)

基　　金：四川省科技支撑计划资助项目(2013SF0157)

摘　　要：传统活动轮廓分割算法对纹理丰富的图像存在过分割和欠分割现象,为了抑制纹理对图像分割的影响,分析了各向同性和异性扩散的机理,设计了保边纹理模糊函数,以图像梯度为变量建立了全变分图像纹理模糊模型。根据对象轮廓与边缘之间的关系利用固定点迭代算法提取相应的卡通分量,对卡通分量运用水平集算法对图像进行分割。实验结果表明,所设计的模糊平滑了图像纹理,保护了对象轮廓。相对于传统CV分割算法,提出的分割算法抑制了纹理对图像分割的影响,弥补了传统CV的不足。Traditional active contour segmentation algorithm existed over-segmentation and under-segmentation for texture-rich images. In order to suppress the texture on the influence of the image segmentation,the diffuse mechanism of isotropic and anisotropic was analyzed,the edge preserving and texture blurring function was designed,and the total variation texture blurring model was established by using the image gradient as variable. According to the relationship between object contour and edge,the cartoon component was extracted by fixed point iteration algorithm,component segmentation using the level set algorithm. Experimental results showed that this fuzzy function blurred the image texture,protected object contour. Compared with the traditional CV segmentation algorithm,this algorithm suppressed the texture on the influence of the image segmentation,made up for the shortcomings of the traditional CV algorithm.

关 键 词：图像分割 纹理模糊 各向异性 水平集 

分 类 号：TP394.1[自动化与计算机技术—计算机应用技术]