Heuristic edge detection using fuzzy rule-based classifier

Article type: Research Article

Authors: Mansoori, Eghbal G. | Eghbali, Hassan J.

Affiliations: Department of Computer Science and Engineering, Shiraz University, Shiraz, Iran

Note: [] Corresponding author: Hassan J. Eghbali is with the faculty of Computer Science and Engineering of Shiraz University, Shiraz, Iran. E-mail: [email protected]

Abstract: Edge detection is one of the most important preprocessing operations needed for detection and extraction of objects in scene, especially in the field of machine vision. Since the nature of image data is indeterminate and the edges of an object in an image are not very clear and occasionally transition from scene pixels to object ones occurs moderately, so fuzzy reasoning is able to extract useful attributes from approximate and incomplete data and improve the task of edge detection. In this paper a heuristic fuzzy rule-based algorithm for detecting the edge patterns in an image is presented. The Heuristic Fuzzy Edge Detector, HFED, uses three features from a 3 by 3 window size for each central pixel surrounded by its eight neighbors, to classify that pixel as part of an edge or as non-edge patterns. The fuzzy inference system use these features for classification and because of interpolative operation of fuzzy reasoning, the results are comparable with other well-known edge detector, especially in degraded images.

Keywords: Edge detection, block deviation, pixel discrepancy norm, local degree of edge, fuzzy rule-based classification system

Journal: Journal of Intelligent & Fuzzy Systems, vol. 17, no. 5, pp. 457-469, 2006