Improving edge detection by an objective edge evaluation

Abstract

Edge detection is conventionally performed in an open-loop system. No feedback from the outcome of the detection operation is provided. The quality of the resulting edge image, thus, tends to be uncontrollable. The performance of edge detection can be improved by conducting the operation in a closed-loop system. This system performs an on-line objective edge evaluation. The evaluation process computes a set of edge metrics for the resulting image. An overall edge strength measurement of the image is obtained. No a priori knowledge about the edge distributions in the image is required for the measurement. Feedback of the evaluation adjusts, or guides the adjustment of, the edge operator parameters, thus improving the quality of the resulting edge images.