Image fusion based on a hybrid decomposition via bilateral and Gaussian filters

The fusion result of infrared and visible images tends to become blurred, confused and suffers contrast degradation due to different characteristics of the two sources of information. To solve these problems, a fusion algorithm based on a hybrid decomposition via bilateral and Gaussian filters is proposed. Bilateral and Gaussian filters are applied firstly to achieve a hybrid decomposition for infrared and visible images. Small-scale texture details, large-scale edges and coarse image information of the input images are obtained. The sub-information is then merged to inject prominent infrared image features into the visible image, for which the fusion weights are determined according to the infrared image features represented by the decomposed large-scale edges. Finally, all the merged sub-information is assembled to reconstruct the fused image. Experiments demonstrate that the proposed fusion method obviously outperforms the image fusion algorithms based on the conventional multi-scale decomposition.