In view of the fact that the complementarily of functional and structural information is not strong in the multimodal medical image fusion method, which easily appears the phenomenon of edge distortion and contour blur, a medical image fusion scheme based on the salient features coupling local extreme value decomposition is proposed. The local extremum is introduced to decompose the source image into a series of smoothing sub-images in different scales. Canny operator is used to obtain the edge weighted mapping to keep the structure information of source image. The color perception mapping is extracted by the context aware operator. The fusion criterion based on the edge and color saliency feature functions as the weighted mapping coefficients are defined respectively to get smooth and detail fusion image. A new image is reconstructed from the smooth and detail image. Experimental results show that in CT and MRI, CT and PET image fusion, the edges and contours obtained by the proposed method are clearer and richer in details compared with the current fusion algorithms. The proposed algorithm has good fusion quality which has certain application value in the fields of medicine, remote sensing and infrared detection.