| Abstract | Image fusion is the combination of two or more different images to form a new image that contains enhanced information. Consistent with specific application goals, a variety of image products arises
from the many available fusion algorithms. However, there is no universal, quantitative performance measure to estimate image fusion quality. The essential objective of image fusion is that nearly all of the original application-specific information
should be preserved, and artifacts should be minimized in the final product. The wavelet transform, a well-known and solid mathematical tool, has already been applied to multi-sensor image fusion. The wavelet transform allows the decomposition
of an image into its constituent spatial scale layers. Most image fusion techniques, including wavelet analysis, require that the input images of different spatial resolutions and sample sizes first be re-sampled to achieve spatial registration. The
re-sampling could cause a loss of spatial information or might introduce artifacts in the final fused image, especially when the resolutions of the input images are significantly different. In this paper, as a further development of the
application of wavelet analysis to image fusion, we propose a new scheme for multi-resolution image fusion, Preserving Spatial Information and Minimizing Artifacts (PSIMA) with multi-scale wavelet analysis. With the PSIMA scheme, the images are fused
in almost their original pixel size. Therefore, the finest spatial information of the input images can be preserved and artifacts minimized in the final fused product. We demonstrate the PSIMA scheme using RADARSAT-1 ScanSAR and NOAA AVHRR images.
The results show that the PSIMA scheme is superior to conventional wavelet analysis for image fusion in terms of spatial information preservation and artifact rejection. |