Gaussian Regularizing CV Model Using Entropy and Neighborhood Information

Abstract

This paper presents a robust Gaussian regularizing CV model using entropy and local neighborhood information. In the energy functional of this model, the local interior and local exterior energies are weighted by entropy, which improves the evolving curve. In addition, we consider local rather than global image statistics and evolve a contour based on local information, which reduces the great impact of the heterogeneous grays inside of regions and improves the segmentation results in the evolving curve process. Then, the Gaussian kernel is used to regularize the level set function, which not only can keep the level set function smooth and stable, but also remove the traditional Euclidean length term and re-initialization. To reduce the sensitivity to the initialization, we use the Circular Hough Transformation to obtain the initialization automatically in the cardiac experiments. The encouraging results on the medical images indicate that our new algorithm has the advantage of high accuracy and strong robustness.