A ROBUST VISUAL METHOD FOR ASSESSING THE RELATIVE PERFORMANCE OF EDGE-DETECTION ALGORITHMS

A new method for evaluating edge detection algorithms is presented and applied to measure the relative performance of algorithms by Canny, N alwa-Binford, Iverson-Zucker, Bergholm, and Rothwell. The basic measur e of performance is a visual rating score which indicates the perceive d quality of the edges for identifying an object. The process of evalu ating edge detection algorithms with this performance measure requires the collection of a set of gray-scale images, optimizing the input pa rameters for each algorithm, conducting visual evaluation experiments and applying statistical analysis methods. The novel aspect of this wo rk is the use of a visual task and real images of complex scenes in ev aluating edge detectors. The method is appealing because, by definitio n, the results agree with visual evaluations of the edge images.