A Robust Active Contour Segmentation Based on Fractional-Order Differentiation and Fuzzy Energy

Abstract

Vascular diseases cause a wide range of severe health problems. Vessel images are often corrupted by intensity inhomogeneity and blurry boundary, which makes it difficult to segment vessel image to identify vascular lesions. Integrating the fuzzy decision and a special local energy functional, in this paper, a robust active contour model is proposed to segment preprocessed vessel images. First, as for the blurry boundary problem, unlike the traditional method, a fractional-order differential method is used to enhance the original image for accurate segmentation utilizing fully high-frequency marginal features. Then, to deal with intensity inhomogeneity, a novel energy functional is formulated by considering the local fuzzy statistical information of boundaries. At the same time, a double-well potential function is designed to automatically limit the values of the membership function in the range [0, 1] during the curve evolution. Finally, Experiments on synthetic and real images are carried out, showing the accuracy of the proposed model and the robustness to the initial contour when working on vascular images.

Keywords

• Active contour
• vascular segmentation
• fuzzy energy
• double-well potential function
• fractional-order differentiation