Boundary detection by contextual non-linear smoothing

摘要

In this paper we present a two-step boundary detection algorithm. The first step is a nonlinear smoothing algorithm which is based on an orientation-sensitive probability measure. By incorporating geometrical constraints through the coupling structure, we obtain a robust nonlinear smoothing algorithm, where many nonlinear algorithms can be derived as special cases. Even when noise is substantial, the proposed smoothing algorithm can still preserve salient boundaries. Compared with anisotropic diffusion approaches, the proposed nonlinear algorithm not only performs better in preserving boundaries but also has a non-uniform stable state, whereby reliable results are available within a fixed number of iterations independent of images. The second step is simply a Sobel edge detection algorithm without non-maximum suppression and hysteresis tracking. Due to the proposed nonlinear smoothing, salient boundaries are extracted effectively. Experimental results using synthetic and real images are provided.