The effect of image feature qualifiers on fuzzy colorectal polyp detection schemes using KH interpolation - towards hierarchical fuzzy classification of coloscopic still images

Previous studies showed that intensities, intensity variation, edge densities, structural entropies of colonoscopy images and their wavelet transforms are good candidates for being selected as antecedents in fuzzy decision methods using KH interpolation for determining whether an image segment contains colorectal polyp segment. In the present consideration, we check which possible antecedent dimensions need interpolation, whether the average and variation of the gradients makes the classification more effective and whether some of the features can be omitted for some classes of images. The method is tested on three available databases consisting of images of three different resolutions, and according to the results, different resolutions, different types of polyps require different classification schemes, thus a hierarchical decision system needs to be built.