1 新算法描述及原理 - 中国图象图形学报.doc

中图法分类号：TP301.6 文献标识码：A 文章编号：1006-8961(200 ) - - 论文引用格式： 阈值标记的分水岭彩色图像分割 张海涛1， 李雅男1 1.辽宁工程技术大学软件学院，葫芦岛 125105 摘 要：目的：针对传统分水岭算法中产生的过分割问题，提出一种新的基于阈值标记的分水岭彩色图像分割算法。方法：该方法将分水岭算法直接应用到原始梯度图像上而不是简化之后的图像，这样做的目的是可以保护边缘信息不受损失；利用不同尺寸结构元求取彩色图像形态学梯度，解决了关于保护边缘和图像简化之间的矛盾。同时新算法设计一种阈值自动选取与标记提取方法，从梯度的低频成分中提取与物体相关的局部极小值，用这些极小值构成的二值图像强制标定原始梯度图像，在修改后的梯度上进行分水岭分割。结果：在仿真实验中，利用本文算法针对不同RGB彩色图像进行分割，获得准确、连续封闭的分割边界，与其他同类方法相比，得到符合人类视觉的最小分割区域数，同时在运行效率上也有很大提高。结论：该方法可以自适应提取标记而不需要先验知识，有效解决了分水岭算法的过分割问题，相对于传统的算法，提高了分割性能，有较好的适用性和鲁棒性，可将其应用于机器视觉、生物医学以及高光谱遥感图像分割领域。 关键词 ：彩色图像分割；多尺度梯度；最大熵阈值；巴特沃斯低通滤波；标记提取；分水岭 Watershed algorithm of color image segmentation based on threshold mark Zhang Haitao, Li Yanan 1.School of Software，Liaoning Technical UniversiHuludao，125105 Abstract: Objective: In view of the segmentation problem caused by the traditional watershed algorithm, a new improved watershed algorithm for color image segmentation based on threshold mark is proposed.Method: The method applied watershed algorithm directly to the original gradient image but not the simplified image, the purpose of which is to insure the loss edge information can be avoided. With different size structure element for color morphological gradient image, solved the contradiction between the protection of the edge and the image is simplified. At the same time, the new algorithm designed a method of threshold automaticly selection and marker extraction--extracting local minimum value related to the objects from the low-frequency components of the gradient, with these minimal values constituting the binary image force calibration of the original gradient image,then used watershed segmentation on the modified gradient.e Result In the simulation experiment,the algorithm in this paper is compared with similar segmentation method.Algorithm obtained accurate and continuous closed boundary for the different RGB color images segmented,which got the minimum segmentation number that conforming to human vision,and improved th