AFLLC: A Novel Active Contour Model Based on Adaptive Fractional Order Differentiation and Local-Linearly Constrained Bias Field

Abstract

In this work, we propose a novel active contour model based on adaptive fractional order differentiation and the local-linearly constrained bias field for coping with images caused by complex intensity inhomogeneity and noise. First, according to the differentiation properties of Fourier transform, we employ the Fourier transform and the Inverse Fourier transform to obtain a global nonlinear boundary enhancement image. In order to overcome the difficulty of setting the optimal order manually, an adaptive selection strategy of the order of fractional differentiation is presented by normalizing the average gradient amplitude. Then, according to the image model, the energy functional is constructed in terms of the level set by taking the fractional differentiation image as the guiding image. In energy functional, local linear functions are used to describe the bias field and construct the local region descriptor since they can flexibly deal with local intensity variety and ensure the overall data fitting. Finally, experimental results demonstrate our proposed model achieves encouraging performance compared with state-of-the-art methods on two datasets.