Piece-Wise Smooth Multi-Object Area Contours Extraction

Multi-object contours extraction has been widely used in image processing and graphic drawing. Piece-wise constant methods which are based on binary value images are not suitable to piecewise smooth images. An improved level set method is presented for piecewise smooth multi-object area contours extraction. Firstly, the disadvantages and limitations of currently used conventional model are pointed out; Then, bidirectional abstracted forces in gradient vector field which instead of unidirectional normal forces are used in level set function, it can drive the active contours moving towards inside or outside edges of image, while remain the features of topological changes; Thirdly, an adapted signed distance function has been given for deduce the number of re-initialization signed distance function in convergence iteration, it can adjust its values automatically. Experiment results show that the method is feasible and efficient in multi-object area contours extraction.