Two preprocessing techniques based on grey level and geometric thickness to improve segmentation results

Two different techniques of performing preprocessing of an image to improve segmentation results are presented. The methods use the grey level thickness of the objects, in order to find the resulting image, by varying the size of a neighbourhood depending on the sum of the included grey levels. The first method, RW, uses the random walk of a particle, defined in the neighbourhood of the position of the particle. The resulting image holds the number of times the particle visits a pixel. Instead of randomization to find the number of visits, the second method, IP, scans the image iteratively and calculates the expected value of the same number. Three different kinds of real world applications are demonstrated to get better segmentation results with the preprocessing techniques included than without.