Robust Local Region-Based Active Contour for Inhomogeneous Image Segmentation

Abstract

This paper proposes a robust local region-based active contour model for inhomogeneous image segmentation. In the model’s energy function, the gray intensity of each pixel within a region is fitted to a local weighted mean of the original image. To solve the local minimization problem of above energy, the consistency of local contrast between the inner and outer of curve at each pixel in the curve is imposed as a weighting for the local fitting energy. A normalization constraint for the curve is added to the energy, and the minimization of the energy is implemented with level set method. Experimental results reveal that compared to related methods the proposed method can improve the segmentation result, being robust to the initial contour.