Robust active contours based on local threshold preprocessing fitting energies for fast segmentation of inhomogenous images

This letter presents a robust active contour model driven by the local threshold preprocessing fitting energies to actualize fast segmentation of inhomogenous images. First, the local threshold preprocessing is carried out to compute two local intensity means. Second, these two means are sorted and serve as the local area fitting centres, which makes the contour move along the internal or external target boundaries. Finally, the above local area fitting centres are used to build the model's energy functional based on the symmetric cross entropy. Experimental results of the real inhomogenous images confirm that the presented model can segment inhomogenous images much faster and is robust to the setting up of the original contour.