针对图像分割中模糊划分熵算法在多阈值选取时存在的效率低、计算重复的问题,提出了一种递推人工蜂群的模糊划分熵多阈值分割算法(RAFPEA).首先选择附加边界条件及灰度权重的隶属函数来构建图像的模糊熵模型,并将该模型中不同变量的组合计算转化为递推过程,进而保存此过程中不重复的瞬间递推值,然后引入人工蜂群算法,利用预存的递推结果来计算蜂群寻优时的个体适应度值,从而减少重复计算,达到快速寻优的目的.实验结果表明:RAFPEA的均一度与精确的穷举模糊划分熵法相同,但运行时间仅为穷举、遗传的模糊划分熵算法的5％；随着阈值数量的增加,运行时间稳定不变,在确保精度的前提下,可高效地对图像进行多阈值分割.%A new recursive artificial bee colony fuzzy partition entropy algorithm (RAFPEA) for multi-thresholding image segmentation is proposed to solve the inefficiency and repeated computation in fuzzy partition entropy approach for selecting the thresholds in the process of image segmentation. The membership functions with attached boundary conditions and gray weights are selected to build the image fuzzy entropy model. The combined computation of different variables in this model is converted to the recursive process and the no-repetitive results of the processing moments are stored. Then the artificial bee colony algorithm (ABCA) uses the stored results to calculate the fitness value of individual species in the ABCA so that the repeated calculations can be reduced and the optimal thresholds can be searched effectively. Experimental results and comparisons with common algorithms indicate that the run time accounts for 5% of ones of the fuzzy partition entropy approaches based on exhaustive algorithm and genetic algorithm. And the uniformity obtained by the proposed scheme is equivalent to the one via exhaustive search. Moreover, as the number of required thresholds increases, the run time keeps stable. The RAFPEA can effectively segment images by multiple thresholds with ensured precision.
展开▼
