A new local segmentation model based on binary level set for images with intensity inhomogeneity is proposed. The local segmentation property is guaranteed with the gradient function of the binary level set function (LSF), and the curve evolution precision can be accurate to one-pixel width. Due to utilizing more statistical information which contains local interior, local exterior, and global interior information, the new segmentation energy is more adaptable for local segmentation in the case of intensity inhomogeneity. Since the contour is initialized as a small shape inside the desired object and inflated afterwards in local segmentation, a morphological closing operation is used to regularize the binary LSF, which not only can promote curve inflating, but also can maintain the binary property of the LSF. Experiments on medical images show that the proposed model is more effective and robust than the Chan-Vese (CV), local binary fitting (LBF), localizing region-based active contours (LR-AC), and selective binary and Gaussian filtering regularized level set method (SB-GFRLS) models in local segmentation for images with intensity inhomogeneity.