Abstract/Summary

To effectively improve the fusion efficiency under the premise of ensuring high fusion quality, a novel and efficient hybrid multiscale decomposition fusion algorithm based on support value transform (SVT) and morphology is proposed. Firstly, each source image is decomposed into one low-frequency approximate background image and a series of high-frequency support value images with salient details by SVT. The decomposition can ensure the support value image in each resolution is unique and the decomposed coefficients can better represent the saliency of image details. Secondly, to overcome the drawback of single-scale decomposition cannot extract details adequately, the idea of constructing dual-channel multiscale morphological decomposition based on `a trous algorithm is proposed. Then the low-frequency approximate background image is decomposed by dual-channel multiscale morphological top-bottom hat (TBH) decomposition to extract the bright-dark information. For the high-frequency support value images, the dual-channel multiscale morphological inner-outer gradient (IOG) decomposition is constructed to extract the edge details. Finally, the decomposed coefficients are integrated and the fused image is reconstructed by corresponding inverse transforms. Experiments were conducted on infrared-visible, multi-focus and medical images and the results shown that the proposed method has state-of-the-art fusion performance. Moreover, compared with some advanced algorithms such as NSCT and NSST, the proposed algorithm not only has best fusion stability, but also greatly reduces the fusion processing time under the premise of guaranteeing high fusion quality. In addition, the powers of de-noising in the fusion of noised images and removing edge-halo are also commendable.