A Novel Approach for Detection of Edges in Range Images Using Splines

In this paper we propose an approach for detection of edges of range images using a local statistics for detection of crease edges. The jump edges are first located using the ratio of the slopes as a measure. For the detection of crease edges, curvature is estimated a t all the data points using the first and second order derivatives of the smoothing spline. The local maximas of curvature are considered as the candidates for the crease edges. This condition is not a very strong constraint, as a result of which a large number of pixels are marked as crease edges. A statistical test is done over the image, and the distribution of the curvature values within a window is used for the detection of crease edges.