融合改进鬣狗优化和Tsallis熵的图像分割

Image segmentation integrating with improved spotted hyena optimizer and Tsallis entropy

针对传统图像多阈值分割法计算代价随阈值数呈指数增长、分割效率低等问题,提出基于改进鬣狗优化算法结合Tsallis熵的多阈值分割算法。为提高鬣狗觅食的精度和效率,利用混沌映射优化初始种群,提升种群丰富性;设计非线性收敛因子调节机制,均衡全局搜索和局部开采;引入邻域重心对立学习提高全局寻优能力,改善局部最优缺陷。将改进鬣狗优化算法应用于图像分割最优阈值求解问题上,以Tsallis熵评估搜索个体质量优劣。实验结果表明,该算法在图像分割效率和分割精度上都具有明显优势。

Traditional image multi-threshold segmentation method has some shortages,such as the computation cost,which has an exponential growth and low segmentation efficiency.In view of this problem,a multi-threshold image segmentation algorithm based on improved spotted hyena optimizer and Tsallis entropy was put forward.To improve the precision and the efficiency of spotted hyena foraging,the chaotic mapping was used to optimize the initial population,which made initial population have better diversity.The nonlinear convergence factor regulating mechanism was used to balance the global searching and local mining ability for the algorithm.The neighborhood centroid opposite-learning was applied to promote the global searching ability and avoid the local optimum.The improved spotted hyena optimizer was applied on solving the optimal thresholds of image segmentation.Tsallis entropy was used to evaluate the quality of individuals.Experimental results show that,compared with three other similar algorithms,on image segmentation pictorial diagram and quantitative index,the proposed algorithm has better performance on the efficiency and accuracy of image segmentation.