Multi-focus image fusion based on joint sparse representation and optimum theory

Image fusion can be regarded as how to integrate complementary information in images with different focal region together. However, except for complementary information, there is also redundant information in source images. To obtain a better fused image, an efficient image fusion method based on sparse representation and optimal solution was proposed in this paper. Firstly, we obtained adaptive dictionaries based on source images themselves by K-means singular value decomposition. Then, we combined a fixed dictionary with adaptive dictionaries to obtain the joint dictionary. By sparse coding source images with the final joint dictionary, complementary and redundant components could be apart. Next, the optimum theory was employed to fuse complementary components and an optimal solution could be obtained by orthogonal matching pursuit. A fused image was constructed by sparse representation at last. Experimental results showed the proposed method had better visual effects and objective valuable index values.