基于双自适应遗传算法的Otsu图像分割研究

Image Segmentation Based on Double Adaptive Genetic Algorithm

为了克服标准遗传算法在图像分割过程中存在易早熟、易陷入局部最优的缺点,同时针对类间最大方差法仅对灰度直方图分布呈双峰的图像效果显著的不足,提出了一种基于双自适应遗传算法的改进Otsu图像分割算法。该方法将传统的自适应遗传算法与新算法进行融合形成双自适应遗传算法,同时考虑目标背景可变的Otsu图像分割算法,使得对个体的评价更为合理,改善种群的全局搜索能力。实验结果表明:与传统Otsu图像分割法及基于遗传算法的图像分割方法相比,该算法求得出的阈值范围更加稳定,且可以获得更优的图像分割效果,有利于计算机视觉的后续处理。

tandard genetic algorithm in image segmentation has the shortcomings,such as easy prematurting and falling intolocal optimization. At the same time,the disadvantage of Otsu is that the gray histogram distribution is bimodal. In order to overcomethese an improved Otsu image segmentation algorithm based on double adaptive genetic algorithm is proposed in this paper. Themethod combines the traditional AGA with the new algorithm,considering Otsu image segmentation algorithm with target back-ground variation. And then,individual evaluation is more rational. Not only it can improve the global search ability of the popula-tion. Results show that,threshold range keeps more stable and the better image segmentation effect when it is compared with Otsuand the standard GA. It is beneficial to the subsequent processing of computer vision.