In the traditional K order singular value decomposition（ KSVD）-based methods

the dictionary learning process is time-consuming and the learned dictionary can’t represent the source images well. Therefore

a novel dictionary learning method is proposed for the medical images fusion problem. In which

these image blocks with rich energy feature and detail information are firstly filtrated to form the training set and the dictionary is learned from the training set. Next

the sparse model is constructed according to the learned dictionary and the sparse coefficients are solved by the orthogonal matching pursuit（ OMP） algorithm. Finally

the ‘max absolute’rule is employed to obtain the fused coefficient and the final fused image is obtained. The experiment has verified that the proposed method is effective for different medical images.