A robust active contour model driven by pre-fitting bias correction and optimized fuzzy c-means algorithm for fast image segmentation

Abstract

Active contour model (ACM) is an effective method for image segmentation that has been widely used in various research fields. For images with severe intensity inhomogeneity, most existing ACMs show a poor segmentation performance. Moreover, robustness of these models to initial contour and noise is unsatisfactory. To seek better approaches to these issues, this paper proposes an ACM driven by pre-fitting bias correction and optimized fuzzy c-means (FCM) algorithm, which is robust and achieves a fast segmentation. Firstly, an optimized FCM algorithm is presented, by which bias field is pre-estimated. Secondly, a criterion function for local intensity is defined, then integrated with respect to the center for a global criterion. Thirdly, the above theory is introduced into ACM according to the property of level set function. Fourthly, a novel regularization method is proposed for variational level set. Experiments on real and synthetic images prove that the proposed model can effectively segment images with severe intensity inhomogeneity. Compared with the bias correction model, there is no more time-consuming convolutions in iterations so that computational amount of our model is enormously reduced. Furthermore, the model has better robustness to both noise and initialization, segmentation efficiency and accuracy than most region-based models.