Polyps in the colorectal part of the bowel appear often, and in many cases these polyps can develop into malign lesions, such as cancer. Colonoscopy is the most efficient way to study the inner surface of the colorectum, and doctors usually are able to detect polyps on a motion picture diagnostic session. However, it is useful to have an automated tool that can help drawing attention to given parts of the image, and later a method for classification the polyps can also be developed. Statistical properties of the colour channels of the images are used as antecedents for a fuzzy decision system, together with edge densities and Renyi entropies-based structural entropy. However promising the processed images are, the variation in the preparation of the diagnosis as well as the practice of the operating personnel can lead to images with significantly different noise and distortion level, thus detecting the polyp can be complicated. In the following considerations image groups are presented that have similarities from the polyp detection point of view, and those type of images are also given, which can spoil a well prepared detecting system.