Boosting Resolution of VHR Remote Sensing Images using CNN

Remote sensing is increasingly pivotal for security and defense, supporting critical decision-making processes. Achieving high-resolution imagery is essential to enhance the accuracy and reliability of these applications. Pansharpening, a technique that fuses high-resolution panchromatic images with lower-resolution multispectral images to produce high-resolution multispectral data, facilitates more effective ground element detection. Traditional pansharpening algorithms, however, often struggle with local misalignments between panchromatic and multispectral images, leading to artifacts and compromised spectral and spatial fidelity. This paper introduces a novel pansharpening algorithm that mitigates these challenges through an advanced local coregistration technique. By automatically improving the alignment of both inter-and intra-band components during the pansharpening process, our method ensures a more precise and accurate fusion. Additionally, our approach incorporates a super-resolution technique that enhances the resolution of multispectral data up to 15 cm. Experimental evaluations using WorldView-3 satellite images demonstrate the effectiveness of our approach, showcasing significant improvements in visual quality and opening new possibilities for advanced remote sensing applications.