Pansharpening of Multispectral Images using Deep Learning Architectures: Trends & Future Directions

The process of combining the geographic and spectrum information of the source pictures into a combined one is called panchromatic. The multispectral image fusing is also referred to as pan-sharpening. The final picture is more reliable for subsequent tasks because it has a higher spatial and spectrum precision than any original image. Many methods have been proposed to enhance the quality of the results obtained by image fusion. The fast growth of deep learning (DL) and artificial intelligence (AI) in recent years has significantly influenced the creation of pan-sharpening techniques. However, a thorough summary of recent advancements in this area aided by the rise of AI and DL is not yet accessible. A wide range of distinct deep-learning methods is examined in this summary piece. The substantial computation needed to use neural network-based methods for practical mapping prevented them from being evaluated. The methods were assessed using four spectrums and three geographic quality metrics. The results showed that most hybrid-based methods had the best spatial quality, methods based on CS had the worst results both spectrally and spatially, and methods based on Sparse representation did better regarding spectral quality.