An annular region based clustering method of color image segmentation based on 2-D cloud model and its application in vision system of biomimetic robot fish

In order to solve uncertainties of surface wave, surface reflection and changes of environmental illumination in image processing, this paper proposes a color image segmentation algorithm based on annular region segmentation model and 2-D cloud model. Firstly, according to the morphological feature of objective, an annular region segmentation model is presented by lots of experimental results. Secondly, 2-D cloud models are introduced to describe the color information of every region, and then they are taken as centers of clustering segmentation. Finally, memberships of every pixel are calculated through the expectation curve formula of each 2-D cloud model, by which color image clustering analysis is performed combined with employing maximum likelihood principle. The proposed algorithm has no over-segmentation or lack-segmentation phenomenon, meanwhile, the segmentation results are consistent with the eyes´ perception. Thus, it is ideal. The experimental results have illustrated that such an algorithm can segment the objective accurately and is robust to the uncertainties of environmental illumination.