Adaptive level-set evolution without initial contours for image segmentation

摘要

We propose a novel variational formulation (external en-nergy) for implicit active contours that forces the level-set function tonhave the opposite sign along the edges at convergence. This exter-nnal energy is then incorporated into a variational level-set formulationnwith two extra regularization terms (internal energy). The resultingnevolution of the level-set function is the gradient flow that minimizesnthe overall energy functional. Because of the external energy, thenlevel-set function can be initialized to any bounded function (e.g.,na constant function), which completely eliminates the need of initialncontours. This implies that the new formulation is robust to initializa-ntion or even free of manual initialization. The proposed model is beennapplied to both real and synthetic images with promising results. Be-nsides, our model is especially appropriate for segmentation of realnimages with a complex and /or nonuniform background