Image segmentation on adaptive sub-region smoothing
Junruo Gao, Xin Liu, Kun He · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2017
To improve the performance of the active contour segmentation on real images, a new segmentation method is proposed. In this model, we construct a function about Gaussian variance according to sub-regions intensity. Further, to avoid the curve vanishing, we design the convergence condition based on the confidence level of segmentation sub-regions. Experimental results show that the proposed method is less sensitive to noise and can suppress inhomogeneous intensity regions efficiently.