结合蚁群和自动区域生长的彩色图像分割算法

摘要

In order to overcome some shortcomings of traditional region growing algorithm , such as bad initial seed pixels and poor robustness of growing order, this paper puts forward a method of color image segmentation based on ant colony optimization region growing. First of all, according to the given threshold, we can select automatically the seed pixels by using ant colony algorithm, then select growth and adjacent termination criterion on the basis of the distance di and similarity value d (Hi,Hj), finally optimize segmentation result by using mathematical morphologymethod. We can find that this algorithm has obvious advantages in the segmentation accuracy compared with JSEG and SRG algorithm.%为克服传统区域生长算法对初始种子像素选择以及生长顺序鲁棒性较差等缺点，提出了一种基于蚁群算法优化区域生长的彩色图像分割方法。首先，根据给定阈值，利用蚁群算法自动选取种子像素，然后，根据相邻距离 di 和相似度值 d（Hi，Hj）的值选取生长及终止准则，最后利用数学形态学方法对分割结果进行优化。通过与 JSEG 和 SRG 算法比较发现，所提出的改进算法在分割准确性上具有明显优势。