Abstract: To solve the problem that single threshold cannot take care of the situation of each pixel when illumination varies, noise exists or background image gray changes dramatically, an improved image binary segmentation based on local ant colony algorithm was proposed. Within different image areas, different iteration numbers and steps were set for ants to achieve superior image segmentation results. The experimental results indicate the proposed method can enhance segmentation accuracy and reduce running time.

Key words: image segmentation, Ant Colony Optimization (ACO), K-means

摘要： 为了解决光照不均匀、有噪声，或者背景灰度变化较大时，采用单一阈值不能兼顾图像各个像素的实际情况，提出一种利用局部蚁群算法对图像进行阈值分割的改进算法。对图像进行分块，在每块内分别设定阈值进行分割，可以有效减少像素错误归类的现象。该算法对图像的不同区域设置不同的迭代次数和蚂蚁走的步数，获得了更好的分割效果。实验表明该算法可以提高分割的精度，缩短程序运行的时间。

关键词: 图像分割, 蚁群算法, K-均值