Intrinsic Image Decomposition-based Multi-resolution Image Fusion

Using intrinsic feature fusion, image decomposition offers a sophisticated method for amalgamating multiple images into a unified composite, enhancing visual appeal and informational depth. This process utilizes an intrinsic image decomposition algorithm to break down each input image into its fundamental components: reflectance and illumination. Introducing an innovative super-resolution technique for multi-resolution multispectral image fusion, rooted in Smoothing Filter-based Intensity Modulation (SFIM), the procedure begins by identifying the most suitable guide images to enhance the high-resolution counterparts of each band, aligning with the low-resolution images of the same scene. Subsequently, the super-resolution image is upsampled and sliced, with the average gray value of each slice computed to determine the appropriate guide image slice, guided by a formulated degree value. Following this, the SFIM-enhanced algorithm is applied to the image slices, which are then integrated to generate the final enhanced output. Experimental validation conducted on the Sentinel-2 flood dataset demonstrates the effectiveness of this approach for multispectral image fusion super-resolution. Comparative analysis against alternative image super-resolution methods highlights its superior subjective and objective performance, as well as computational efficiency, establishing its superiority over existing approaches.