A qualitative profile-based approach to edge detection

Edge detection is a fundamental problem of computer vision and has been widely investigated. We propose a new framework for edge detection based on edge profiles. Our approach is partly inspired by the well-known model called “active contour” which is based on minimization of the energy composed of the terms of continuity and smoothness of the contour, along with edge attraction forces of the image. We will introduce similar constructs called “worms” in which profile-forced are used. Unlike a snake which is initialized by an arbitrary curve, a worm can be initialized at a seed point. Our edge model, based on one-dimensional qualitative edge profile fitting and edge consistency, will produce one continuous edge from an initial seed point. A “profile” is defined as a finite cross-section of a two-dimensional image along a line segment. By “qualitative edge profile fitting”, we mean that, in the vicinity of an edge point, the profile from that edge point will give a local minimum value, relative to some evaluation function. The value itself is not important. “Edge consistency” means that all the profiles on the same edge should be consistent. Appropriate evaluation functions are needed for different types of edge profiles, such as step edges, ramp edges, etc. an evaluation function must meet the requirement that it will produce local minima at the positions where edges of a given type occurs in the profile. Instead of subjective thresholding, image noise is measured statistically and used as a systematic way of filtering false edges. Once an edge point is localized, parameters of the edge, such as colors on each side and edge width, can be determined by the profile of that edge point. The rest of the edge points on the same edge can be found by extended from that point, using the specific evaluation function and those edge parameters. The software architecture for achieving consistent and high quality results in this setting is a non-trivial problem we address. We develop the necessary algorithms and implement them. The results are compared to widely used edge detectors such as the Canny detector.