A study of edge detection algorithms

Edges are image attributes which are useful for image analysis and classification in a wide range of applications. The numerous applications and the subjective approach to edge definition and characterization have promoted the development of a large number of edge detectors (or operators) which may perform well in given applications but poorly in others. In this work we describe a variety of edge detectors and evaluate their performance in terms of some known performance measures. The study is based on computer simulated edges of different shapes, slopes and background noise levels. The performed evaluation together with results in other related works helps to categorize the different edge detection schemes, as well as to better understand the usefulness and limitations of the performance measures used.