Image fusion with perfect reconstruction DFT/RDFT filter banks and DWT

Abstract

We present a perfect reconstruction DFT/RDFT filter bank algorithm and its application in image fusion in comparison to and in conjunction with the DWT method. In order to achieve perfect reconstruction, it is necessary to modify the input signal slightly with an invertible transformation only when the number of data points is divisible by 4. In image fusion application investigated, a SPOT panchromatic image is fused with a corresponding SPOT multispectral image with three bands. For this purpose, the approximation subband of the multispectral image is merged with the detail subbands of the panchromatic image. The fused image is recovered from the subband images by the perfect reconstruction algorithm. The RDFT and DWT results are comparable. By fusing the two together with another algorithm, better results are obtained