An Improved Algorithm for Image Segmentation

One-dimensional image segmentation algorithm only considered the distribution of the pixel grayscales, but ignored the correlation between different gray levels, and needed an objective function. To solve these problems, a two-dimensional optimal evolution algorithm (2D-OEA) without an objective function for image segmentation is proposed based on optimal evolution algorithm. The Two-dimensional vectors represented the image's two-dimensional information are regarded as chromosome. Assuming the optimal evolution direction exists, the updating model of evolution direction is established. Then define the chromosomes' coding rules, initialize the group by simple-random-sampling, select chromosomes to crossover and mutate, calculate the fitness values, produce a new population by the selection mechanism and modify the threshold to obtain a stable two-dimensional optimal threshold. The rationalities of the assumption and the updating model have been analyzed in this paper. The experimental results show that the assumption and the updating model are proper. Two-dimensional optimal evolution algorithm (2D-OEA) is a fast, robust and effective algorithm, and it is better than OEA.