Abstract: In this paper, in response to the question of the segmentation of images with intensity inhomogeneity and noise, a new adaptive weights image segmentation model was proposed based on the combination of the local and global region intensity information. Firstly, the local and global energy functionals were constructed based on the local and global image intensity information respectively. Secondly, a new adaptive weights function based on the denoising methods of wavelet transform and wavelet threshold was defined by constructing the edge characterization of matrix insensitive to noise. Finally, the defined weights function was utilized to combine local and global terms adaptively to obtain the energy functional of the proposed model. The model’s level-set function was deduced from the variational method, and the finite difference method was employed to realize numerical solution. Experimental results show that the proposed model, which combines the advantages of the Chan-Vese model and the Local Binary Fitting model, can effectively segment images with intensity inhomogeneity and noise and is robust to the positions and shapes of initial contour of evolution curve.

Key words: image segmentation, level-set method, wavelet transform, adaptive weight, intensity inhomogeneous image