基于遗传模拟退火算法的Otsu图像分割研究

针对目前基本遗传算法在优化图像分割算法中存在的易于早熟、陷入局部最优的不足,以最大类间方差函数为适应度函数,提出了一种基于改进遗传算法的图像阈值分割算法.对交叉、变异算子进行自适应改进,同时将模拟退火算法融入到遗传算法中,使得对个体的评价更合理,既能克服种群退化现象,又改善算法的全局搜索能力,避免遗传算法陷入局部最优.实验结果显示,与Otsu图像分割法以及基于遗传算法的图像分割方法相比,使用该方法得出的阈值范围更加稳定,执行效率更高,在图像分割中获得的分割效果更佳.

Image Segmentation on Genetic Simulated Annealing Algorithm

To address some defects which basic genetic algorithm is exploded in this day and age, such as easy precocious and local optimum in optimizing image segmentation, the image threshold segmentation algorithm based on improved genetic algorithm is proposed in this article with considering Otsu as fitness function. The cross and mutation operators are optimized adaptively while the simulated annealing algorithm is fused into genetic algorithm. And then, individual evaluation is more rational. Not only can population degradation be overcome, but global search performance of the algorithm can be enriched by utilizing the optimized algorithm. The experimental result indicates that threshold range keeps more stable and operating efficiency is better, compared with Otsu and the basic genetic algorithm. As a result, the obtained image segmentation effect is more apparent and perfect.