Robust Active Contour Model Using Patch-Based Signed Pressure Force and Optimized Fractional-Order Edge

Active contour models (ACMs) are the most widely used method for image segmentation. However, global fitting ACMs cannot effectively segment inhomogeneous images and local fitting ACMs suffer from noise and the initial position of the contour. To overcome these shortcomings, we propose a novel ACM that consists mainly of a local fitting term, an edge-based term and an external force term. The Jensen-Shannon divergence (JSD) based local fitting term is implemented to address intensity inhomogeneity. The edge-based term is formulated on Caputo-Fabrizio (CF) based fractional-order Gaussian derivatives and applied to compute the weighted area of the region inside the contour. The patch-based external force is designed to improve the robustness of the developed ACM to noise and the initial position of the contour. To further improve the robustness of the proposed model to noise, the input image is first replaced with its local robust statistics. Experimental results demonstrate that the developed model is not only robust to noise and the initial contour but also effective in dealing with intensity inhomogeneity.