Improved region-scalable fitting model with robust initialization for image segmentation

摘要

It had been known the region-scalable fitting (RSF) model can handle images with intensity inhomogeneity effectively, but it depends on the position of the initial contour. In this letter, we present a scheme about the improvement on the RSF model in term of the robustness of initialization. In the process of curve evolution, we add a function to exchange the value of fitting inside and outside curve in the regions where the direction of curve evolution is opposite. In this way, the whole curve will evolve along the inner or outer boundary of object. Thus, the improved RSF model will not be trapped in local minima. The experimental results have proved the proposed scheme is robust to initialization. And compared to other improvements, our proposed scheme is simple and efficient.