A Novel Image Fusion Method based on Translation Invariant Wavelet Transforms

In this paper, a novel method based on translation-invariant wavelet decomposition has been introduced to the pixel level multisensor image fusion. The proposed fusion architecture is related to the “shiftdecompose-fuse-shift” technique and consists of many steps. First, the source images to be shifted in both horizontal and vertical directions. The shifted images will be translated into the wavelet domain and by repeating “shift-translate” to get the decomposition of source images. Second, the different subband coefficients of the fused image are combined with the proposed fusion rule. Finally, the fused image will be obtained with the inverse translate and shift. Experimental results demonstrate that the proposed method fuses the useful information from source images and outperforms the discrete wavelet transform (DWT) and stationary wavelet transform (SWT).