Active Contours Driven by Saliency Detection for Image Segmentation

Abstract

Aming at the over-segmentation problem of the active contour models, a new model based on the LBF (Local Binary Fitting) model driven by saliency detection is proposed. The proposed method consists of two main innovations: (1) The target object is located quickly and the initial contour is generated automatically by saliency detection method, which solves the problem that the LBF model is sensitive to the initial position, and the different targets can be segmented by selecting different initial contours. (2) The saliency detection results are transformed into priori energy functions, which are added to the energy model to prevent over-segmentation during the iterative process. We applied the proposed method to some gray images and real images, the simulation results show better segmentation accuracy.