一个基于信息熵和水平集方法的噪声图像分割模型

Variational level set methods(VLSMs)have been widely usedin the field of image segmentation, but many models are sensitive to noise and the initialization of active contour. To address these issues, a new VLSMcombiningthe global and local imageinformation is proposed in this paper,based on information entropy theory. The proposed model is based on the Gaussian distribution fitting of the gray level of the image, which utilizes the information entropy of the global and local distribution to measure the internal difference of the image in the corresponding regions, and uses the relative entropy between them to measure the difference between the global and local regions. The resulting energy functional is minimized by gradientdescent, which controls the evolution curve to move towards the target boundary and eventually achievesimage segmentation. Compared to some representativeVLSMs, the proposed method yields better robustness to noise and initial contours.