Color image segmentation based on fuzzy rule-based reasoning applied to colonoscopic images

A fuzzy color segmentation approach is developed for the analysis of colonoscopic images. The segmentation is made up of two phases: segmentation through histogram space filtering and region merging using fuzzy rule-base reasoning. The first phase involves using a scale-space filter to analyze the hue, saturation, and intensity (HSI) histograms to determine the number of classes and construct a 3-D class grid. The color image is then segmented based on the class grid. In the second phase, region merging based on applying the fuzzy rule-base is employed to guide the combining process of the segmented regions. For fuzzy reasoning, three criteria are evaluated, namely, the edge strength along the boundary, color similarity, and spatial connectivity of adjoining regions. Experimental testing of the proposed method applied on colonoscopic images was conducted, and the results are encouraging.