The possibilities of improving the effectiveness of fuzzy Hough transform calculations in the detection of colonoscopy image processing and polyp detection by gradient based preprocessing is studied. For the Hough transforms a black and white image consisting of line segments is necessary, thus most of the times an edge detected image is used as the basis of the transform. Here, the gradient magnitude values corresponding to the Canny edge pixels of the image are used for determining, which are the typical magnitude values for the polyp contours, in order to remove part of the non-contour edges from the image. Three publicly available databases with images and ground truth masks are used for testing, whether a general threshold for the gradients is applicable, but based on the histograms it seems to be not possible to generate a database independent normalized gradient based domain that can be used for sorting out the unnecessary edges.