In this paper, we propose a new region-based image fusion approach using energy estimation. In order to segment an input image into a set of regions according to the homogeneity of the region of the image, an improved method to solving the piecewise smooth Mumford and Shah model is studied, and the level set based optimal algorithm is integrated with it to speedup convergence. For selecting the best fusion rule on every region of the image an energy model is proposed to measure the fusion quality of the candidates. Furthermore, the optimal fusion rules can be determined by comparing the magnitudes of the energies. By numerical experiment, it has been demonstrated that proposed approach has shown many advantages over previous ones, such as the ability to preserve all relevant information and removal of the side effects like sensitive to error of registration and reduction of contrast.