Saliency-based variational active contour model for image with intensity inhomogeneity

Abstract

Variational active contour model (ACM) is used to segment or subdivide an image into the desired object. This segmentation technique in region-based ACM can be divided into two classes: global segmentation and selective segmentation. Selective segmentation, in which only a particular desired object is segmented from an input image, is preferable to the global model because the selective segmentation model proves to be very useful, especially in medical image analysis. However, when it comes to segmenting an image with inhomogeneous intensity, these models seem to give unsatisfactory results. In this paper, we propose a new variational selective ACM mainly used for segmentation of images with inhomogeneous intensity, by incorporating saliency image map and local image fitting ideas. In addition, the euler-lagrange equation (EL) was provided to solve the proposed model. A total of thirty sets of medical images were used to test the model. Numerical results show that the suggested model outperforms other existing models, with the hausdorff distance of the proposed model being 47.78% less than the competing model, and the dice and jaccard values being around 17.54% and 33.65% higher, respectively, than the competing model.