Parallel contour recognition using matched filter

An approach to parallel contour recognition is proposed. This approach is based on a matched filter in the spectral domain in parallel to selected geometrical features of contours such as length of the contour, area, radius and center of gravity. In the experimental part of this paper, the selected contours were extracted from the 2D image using the OCE (object-oriented contour extraction) method for contour extraction. Then, these contours were corrupted by Gaussian noise and the number of errors (misrecognized contours) were recorded. Several transforms like Walsh, PWL and Fourier were applied for the matched filter. The recognition results for different signal to noise ratios are shown.