In recent years, medical image fusion is extensively used to help doctors improve the accuracy of medical diagnosis by combining multimodality images acquired under different imaging conditions into a single one. Most of the previous methods aim at attaining information as many as possible from source images. However, some of the exacted information is not necessary or useful for medical diagnosis. As Marrpsilas vision stated, human visual system is being adapted for exacting structure features such as lines, edges, contours from the images. In this paper, we try to develop a new image fusion scheme based on structure similarity match measure (SSIM) to exact structure features from the input images to improve the accuracy of diagnosis. The visual experiments and quantitative assessments demonstrate the effectiveness of this method compared to present image fusion schemes.