Colorectal polyps affect a large percentage of the population all over the world, and they can be a basis for more serious conditions such as cancers. As the most reliable method for detecting a polyp in the lower bowel tract is colonoscopy, more and more image processing experiments appear that help to find or characterize such a lesion. The social benefit of such methods is clear, any aid in detecting pre-cancer states saves lives. In the present considerations a fuzzy decision method for finding polyps on a colonoscopy image is presented. As a first step, the image taken during the colonoscopy is cut into tiles of size N by N, thus a rough localization of the lesion within the picture is also possible. The antecedent dimensions consist of statistical characteristics of the colour channels of the tiles, their Renyi entropies, edge density and fitted polynomial coefficients. The method's dependence on the tile-size within the images are also studied, and the success rate increases with the decrease of the tile size between 70 by 70 and 20 by 20 tile sizes.