Active contour model for image segmentation with dilated convolution filter

ACMs have been demonstrated to be highly suitable as image segmentation models for computer vision tasks. Among other ACM, the local region-based show better performance because they extract information regarding intensity in neighborhood and embed it into energy minimization function guide active contour boundary of desired object. However, online noisy inhomogeneous is still a challenging task ACM models. To overcome this challenge, This paper proposes novel model, named model with dilated convolution filter (ACLD). The ACLD integrates form signed pressure force function. Then, Gaussian kernel applied using instead discrete regularizing level set formulation. Finally, constant stopping condition, automatically stops at object boundaries. proposed shows improved results visually combined less computational time case synthetic natural images compared state-of-the-art Further, on ISIC2017 dataset, yields highest accuracy.