In this paper, we introduce an image segmentation method based on multiresolution and scale-space theory. Specifically, a nonlinear scale-space stack is constructed by anisotropic wavelet shrinkage. A tree structure is derived from the stack based on the relationships between connected components of adjacent scale layers. Then, Equivalent framework of wavelet shrinkage and anisotropic diffusion is proved. Method used to extract connected component during the procedure of constructing scale-space stack is optimized. Based on these, we propose multiresolution anisotropic wavelet shrinkage segmentation algorithm. Comparative experimental results show that proposed algorithm can keep edge features very well while coping with the interior nonuniform. Furthermore, it has some robustness to noise.