In this project, we propose a novel region-based active contour model for image segmentation with a variational level set formulation. By introducing a local binary fitting energy, the proposed method is able to utilize accurate local image information for accurate recovery of desired object boundary.
Our method is able to segment images with nonhomogeneous regions, weak object boundaries, and vessellike structures. Comparison with typical region based active contour models shows the advantages of our method in terms of accuracy and efficiency.
In addition, the proposed method has promising application for image denoising. In our future work, we will extend our method to handle regions with triple junctions. This could be achieved by using a multiphase level set framework. In addition, our current algorithm will be implemented with narrow band techniques to further speed up the computation.
Zhang, Kaihua, Huihui Song, and Lei Zhang. “Active contours driven by local image fitting energy.” Pattern recognition 43, no. 4 (2010): 1199-1206.