A hybrid active contour model driven by global and local intensity information

Abstract

Aiming at the characteristics of intensity inhomogeneity distribution in images, a variational level set image segmentation model combining global and local intensity information is proposed. Local region information is the key to accurately segmenting images. However, the conventional CV model does not utilize the local region information, and the LBF model is susceptible to the initial outline and noise. In this paper, we present a hybrid model driven by new global and local intensity information. A new evolutionary stop function is constructed by using the principle of LBF model, and it is combined with the CV model to obtain an active contour model containing local and global information. By testing various types of real images and synthetic images, the model not only can deal with image with intensity inhomogeneity, but also reduces sensitivity of the model to the initial contour and the iteration number is also decreased.