CVGIP: GRAPHICAL MODELS AND IMAGE PROCESSING Vol. 53, No. 3, May, pp. 258-268, 1991 Rule-Based Multitemplate Edge Detector’ D. BRZAKOVIC Department of Electrical and Computer Engineering, University of Tennessee, Knoxville, Tennessee 37996 R. PATTON I-MATH Associates Inc., . Box 560788, Orlando, Florida 32856 AND R. L. WANG Department of Electrical and Computer Engineering, University of Tennessee, Knoxville, Tennessee 37996 Received November 20. 1989; accepted July 20, 1990 Number of edge pixels detected that co- This paper describes an edge detection algorithm that employs incide with the true edge pixels Number of detected edge pixels directional templates of various sizes and rule-based decision Number of true edge pixels making to assign confidence values to candidate edge pixels. High Quantitative measurements of edge de- confidence values indicate a high likelihood of pixels being edge tector performance pixels, while low confidence values are associated with pixels Areas associated with different edge ge- likely to be noise. Consequently, an edge image is generated by ometries retaining pixels associated with high confidence values. The per- Threshold values formance of the al