Active Contour Model Based on Adaptive Sign Function

Due to the fact that the geometric active contour model is sensitive to the position of initial contours, the distance regularized level set evolution (DRLSE) model must set the initial contour curve inside or outside the target boundary. An adaptive level set evolution (ALSE) for contour extraction is able to reduce the influence of the location of initial contours. However, both of these two models are easy to fall into false boundaries and leak from weak edges, besides, they have poor resistance to noise. This paper provides a novel active contour model, which combines an adaptive sign function with an adaptive edge indication function. This improvement makes the model robust to initial curves, and solves the problems of having slow convergence rate and being easy to leak from weak edges. In addition, a new distance regularization term is presented, which makes the evolution more stable. Experiments on some real images have proved that the proposed model not only improves the accuracy of segmentation and reduces segmentation time, but also enhances the robustness to initial contours. © Copyright 2019, Institute of Software, the Chinese Academy of Sciences. All rights reserved.