Image Segmentation by Active Contour Model with a New Data Fidelity

Amir Razi,Weiwei Wang,Xiang-Chu Feng
DOI: https://doi.org/10.1109/CMVIT.2017.18
2017-01-01
Abstract:Image intensity inhomogeneity and low contrast are two substantial challenges for image segmentation. In this work, we propose a new active contour method fo rsegmenting images with inhomogeneous intensity and low contrast. To deal with the intensity inhomogeneity, we use a local image fitting term. To deal with low contrast, we use a strictly convex and differential function as the data fidelity measure. It has several advantages over the L2 norm and the L1 norm. First, it can well approximate the L1 norm, thus is robust to low contrast and outliers. Second, it possesses the convexity and differentiability of the L2 norm, thus the minimization problem has a global optimal solution and the problem can be easily solved by using the gradient descent method. The experimental results show that the proposed method is robust to intensity inhomogeneity, low contrast and outliers.
What problem does this paper attempt to address?