The traditional methods based on regional characteristics of image fusion have two limitations. First, the fusion rule of low-frequency sub-image does not make full use of regional energy information and regional edge information at the same time. Second, the fusion rule of high-frequency sub-image even sometimes ignores the directional information. In order to overcome the limitation for getting more the neighborhood information from the traditional region characteristics method, an improved image fusion method is proposed based on the NSST（Non-Subsampled Shearlet Transform）and the modified directional fusion rule. First, the source images are decomposed by the NSST algorithm to obtain a low-frequency sub-image and a series of high-frequency sub-images. Subsequently, the low-frequency sub-image is performed by adaptive contrast Spatial Frequency algorithm based on region energy, and the high-frequency sub-images contained the directional information is performed by a novel Sum-modified-Laplacian fusion rule. Finally, the method is verified by living example.