复合形态滤波和改进型分水岭算法的图像分割  

Image Segmentation Based on Compound Morphology Filter and Modified Watershed Algorithm

作　　者：夏平[1,2] 刘馨琼[1,2] 向学军[2] 万钧力[2] 

机构地区：[1]三峡大学智能视觉与图像信息研究所,湖北宜昌443002 [2]三峡大学电气信息学院,湖北宜昌443002

出　　处：《三峡大学学报（自然科学版）》2008年第5期60-63,共4页Journal of China Three Gorges University：Natural Sciences

基　　金：湖北省教育厅自然科研基金项目(D200513001);宜昌市科技发展计划项目(A2007107-08)

摘　　要：提出了一种基于复合形态学滤波和改进型分水岭的图像分割算法.为克服传统分水岭算法进行图像分割时产生的过度分割和耗时的不足,定义了由开-闭以及闭-开加权组合形成复合形态学滤波器,应用该滤波器对图像进行滤波;其次,设计一种应用内、外部标记的形态学分水岭算法,确定区域中梯度低频成分的局部极小值作为内部标记,区域之间确定为外部标记,内、外标记随处理信息的变化而变化,从而实现了对图像信息的自适应分割.仿真结果表明,采用本文算法实现的图像分割,其抗噪的MSE、PSNR性能明显地优于开-闭、闭-开单一运算的抗噪性能;同时,减小了梯度图中由于噪声造成虚假局部极小值而产生的过分割现象,准确地提取图像信息的轮廓、边缘等信息.Image segmentation algorithm based on the compound morphology filter and modified watershed has been put forward. First of all, we define the compound morphology filter with opening-closing operators and closing-opening operators, which overcomes the over-segmentation and time-consuming shortage of the traditional watershed algorithm. Secondly, we design a new morphology watershed with inner and external marks, and also define the regional minima to inner marks from the low frequency components of the gradients and external marks between the region; the inner and external marks change along with the image information, thus which has realized to the image information adaptive segmentation. Experimental results show that the new algorithm can overcome over-segmentation in the traditional morphology watershed segmentation algorithm, which could accurately obtain the image edge and contour.

关 键 词：图像分割 形态滤波 分水岭算法 梯度 

分 类 号：TN911.73[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]