Image segmentation with GA optimized fuzzy reasoning

In this paper, we present an image segmentation approach, in which fuzzy reasoning is employed in two phases, namely, a denoising phase and a region merging phase. Denoising is realized adaptively, depending on local image characteristics via the use of a set of fuzzy rules. It is an iterative process that terminates based on the variation of information content in the denoised (filtered) image. The denoised image is processed using the Canny operator to generate a gradient magnitude image. This gradient magnitude image is then segmented using the watershed transform. A fuzzy reasoning driven similarity measure is then used to merge the regions resulted from the initial watershed transform segmentation. A genetic algorithm is employed to optimize the segmentation results. Experiments have shown that using the proposed approach, high quality denoising and segmentation performance can be achieved on natural as well as medical images.