Abstract: In view of more optimization difficulty with higher dimensions in the high-dimensional multilevel thresholding and the low optimization efficiency of biogeography-based optimization (BBO) algorithm, this paper proposes an efficient BBO (EBBO) algorithm applied to high-dimensional OTSU multilevel image thresholding. Firstly, a novel little disturbance mutation operator is taken, and the operator is blended into the heuristic operator to get a new migration operator, which simplifies the processes of the proposed algorithm, such as not computing mutation probability or setting its parameters, in order to improve the efficiency. Secondly, the roulette wheel selection is     replaced with the example learning approach and the immigration rate is calculated once before the iteration loop   instead of calculating many times in the original algorithm to reduce the computational complexity further. Thirdly, a greedy selection operator is used instead of the original elitist selection operator to accelerate the convergence process. Finally, the EBBO algorithm is applied to high-dimensional OTSU multilevel image thresholding. The experimental results show that the performance of proposed EBBO is better, faster and more stable than standard BBO, EBO, BDE and MKTO, and there are fewer parameters to be tuned in the EBBO.

Key words: intelligent optimization algorithm, biogeography-based optimization algorithm, image segmentation, multilevel thresholding, maximal between-cluster variance

摘要： 针对高维多阈值分割由于维数增加带来的优化难度加大的问题以及标准生物地理学优化（biogeography- based optimization，BBO）算法效率不高的问题，提出了一种用于高维OTSU多阈值分割的高效生物地理学优化（efficient BBO，EBBO）算法。首先构建新型随机扰动变异算子，然后将此变异算子融合到启发式迁移算子中形成一种高性能的混合迁移算子，去掉了计算变异概率和设置变异参数等环节，以便提高算法的效率；其次将基于迁出率的赌轮选择方式改成无需迁出率的榜样学习选择方案，并将迁入率的多次计算变成一次计算，进一步提高算法的效率；然后将BBO算法中的精英保留方案换成贪婪选择方法，以加快算法的收敛速度；最后将这种EBBO算法应用到高维OTSU多阈值分割中。实验结果表明，与当前的EBO算法、BDE算法、MKTO算法以及BBO算法相比，EBBO算法在高维多阈值分割中不仅有更好的优化性能和更快的运行速度，而且减少了参数设置。

关键词: 智能优化算法, 生物地理学优化算法, 图像分割, 多阈值分割, 最大类间方差法