CAJ | 학술논문

为了克服传统基于聚类的图像分割算法需要指定聚类数目以及依赖初始值等缺点,提出了一种基于动态小生境的人工鱼群算法的图象分割方法. 该算法将图像分割问题转化为根据图像像素特征对像素的自动聚类问题.采用更为简单的个体描述方式,每条人工鱼表示一个分割区域的一个可行解,并对进化过程中的人工鱼进行动态的划分小生境,每个小生境对应了图像分割问题中一个分割区域. 通过对鱼群行为的模拟及种群的动态划分实现了对图像分割问题的分割区域中心和区域数的同时进化,实现了一种新的聚类算法,并实现了对图像的自动分割. 实验结果表明:该算法可以自动地估计分割的区域数,并获得较好的分割性能.
위료극복전통기우취류적도상분할산법수요지정취류수목이급의뢰초시치등결점,제출료일충기우동태소생경적인공어군산법적도상분할방법. 해산법장도상분할문제전화위근거도상상소특정대상소적자동취류문제.채용경위간단적개체묘술방식,매조인공어표시일개분할구역적일개가행해,병대진화과정중적인공어진행동태적화분소생경,매개소생경대응료도상분할문제중일개분할구역. 통과대어군행위적모의급충군적동태화분실현료대도상분할문제적분할구역중심화구역수적동시진화,실현료일충신적취류산법,병실현료대도상적자동분할. 실험결과표명:해산법가이자동지고계분할적구역수,병획득교호적분할성능.
In order to overcome the defects in the traditional clustering-based image segmentation algorithm, e.g., it needs to specify the number of clusters, it is sensitive to initial value, and so on, an image segmentation method based on dynamic niche artificial fish-swarm algorithm ( DNAF) is presented in this paper. In the new algorithm, the image segmentation problem is transformed into an automatic pixel clustering process based on the pixel features of the image. A simpler representation is adopted, each artificial fish represents a single feasible solution of one seg-mented area. Moreover, the dynamic identification of the fish niches is performed at each generation to automatical-ly evolve the optimal number of regions. Each fish niche corresponds to one segmentation region in the image seg-mentation problem. Therefore, the proposed DNAF algorithm implements simultaneous evolution in the center of the segmentation region and the optimal number of regions through simulation on the behaviors of fish swarm and the dynamic division of population. It thereby achieves a new clustering algorithm and automatic segmentation of an im-age. Experiment results demonstrate that the DNAF algorithm is able to automatically estimate the number of the segmented regions, and an excellent segmentation performance can be attained.