An adaptive active contour model with Sobolev gradient
Song Qi, Mingxing Wang, Leo Selavo · International journal of imaging and robotics · 2014
Image segmentation is a key step in image and vision technology. In this paper, an adaptive active contour model is proposed which is based on both L^2 gradient and Sobolev gradient. In this new model, the evolving PDE consists of two forces, one is the adaptive force which uses L2 gradient and the other is length regularization force which uses Sobolev gradient. Due to the adaptive force, the level set function can be easily initialized to a piecewise constant function rather than the widely-used signed distance function. In addition, the re-initialization procedure is not necessary, and the evolving PDE is robust with the initialization. The experimental results show that the proposed model can segment robustly two-phase piecewise constant images in a few iterations with accuracy boundary.