Fuzzy triangle contour characterization by subspace based methods of array processing

Fuzzy paradigm was considered from several aspects in image segmentation. For the first time, we derive a signal processing model out of an image which contains a fuzzy contour. We propose to adapt subspace-based methods of array processing which are originally dedicated to multiple incoherently distributed sources, to retrieve the orientation and spread parameters of fuzzy contours. A set of experiments performed on hand-made and real-world images shows that the proposed methods estimate accurately the expected orientation and spread parameters of fuzzy contours, and exhibit a small computational load.