Active contour model based on local Kullback–Leibler divergence for fast image segmentation

The inhomogeneity of image intensity and noise are the main factors that affect the segmentation results. To overcome these challenges, a new active contour model is designed based on level set method and Kullback–Leibler Divergence. First of all, a new regional measurement of information scale is applied to construct energy functional, instead of Euclidean distance. Test results demonstrate that the Kullback–Leibler Divergence achieves a truly better segmentation. Then, a new Heaviside function has been proposed in this paper, which gives rise to a faster zero-crossing slope than traditional function. In this sense, it can stimulate the evolution of the level set function faster and allocate internal and external energy reasonably. In addition, the activation function has also been improved, which makes itself fluctuates over a smaller range than former activation function. Experiments reveal that the ‘Local Kullback–Leibler Divergency’ (LKLD) model has desired segmentation results both on real-world and medical images. Also, it owns a better noise robustness and is not limited to position of initial contour.