A Model Based Edge Location with Subpixel Precision

Estimation of edge features in an image (sub-pixel position, direction, curvature and change in intensity at both sides of the edge) starting from the calculus of the gradient vector in each pixel is not exact, even in ideal images. In this paper we present a new edge detector based on an edge and adquisition model that does not assume continuity in the image values. Edge features are obtained with high precision using this detector, even in noisy images. When noise is too high, we also propose an iterative method to enhance the image quality, and keep precision in the features obtained. Finally some synthetic and real examples are shown.