| Summary: | The conventional Hough Transform used for detection of objects with known shape and size has proved its robustness. One typical task for this transform can be the detection of stamp(s) on an envelope. Unfortunately, the Hough Transform has an important drawback: the heavy computational effort and, as consequence, a big execution time. This paper introduces a variant of Hough Transform that speeds up the process. One important aid is given by a filtering step based on a fast analysis of a rough deformation model. This method is a combination of Randomized and Local Hough Transform. Experiments were made comparing the modified Hough Transform approach with the UpWrite Transform and they proved that the first approach preserves the quality of the Hough Transform results at a higher speed.
|
|---|