High-accuracy edge detection with Blurred Edge Model

A high-accuracy edge detection algorithm at sub-pixel level is proposed in this work. A blurred edge model is adopted here, and a least-squared-error based solution is derived. Its applications to both synthetic and real images are presented for evaluation. It is compared with two other sub-pixel edge detectors. One uses a moment-based approach, and the other an interpolation-based approach. The comparison shows higher accuracy of the proposed algorithm, even for image data with significant noise. An application of the proposed algorithm in engineering is also introduced herein.