Gaussian-based approach to subpixel detection of blurred and unsharp edges

In this paper the problem of edge detection with subpixel accuracy is considered. In particular, the precise detection of significantly blurred edges is regarded. A new method for subpixel edge detection is introduced. The method attempts to reconstruct image gradient function at the edge using the Gaussian function. The results of subpixel edge detection in the artificially created and the real images obtained by the introduced approach are presented and compared with the results of previously proposed methods. In particular, the moment based methods, the gravity center method and the parabola fitting method are considered in the comparison. The presented results prove the robustness of the introduced approach against the averaging and the Gaussian blur. Additionally, the comparison shows, that the introduced approach outperforms the existing state-of-art methods for subpixel edge detection.