Target-adaptive CNN-based pansharpening

TL;DR

This paper introduces a lightweight, fast-training CNN for remote sensing pansharpening that adapts effectively to target images, even with sensor mismatches, providing high-quality results with broad usability.

Contribution

It presents a novel target-adaptive CNN approach for pansharpening that maintains high performance across different sensors and mismatched training data.

Findings

Achieves further performance gains over previous CNN models.

Enables fast training and adaptation to target images.

Provides an accessible software tool for practical use.

Abstract

We recently proposed a convolutional neural network (CNN) for remote sensing image pansharpening obtaining a significant performance gain over the state of the art. In this paper, we explore a number of architectural and training variations to this baseline, achieving further performance gains with a lightweight network which trains very fast. Leveraging on this latter property, we propose a target-adaptive usage modality which ensures a very good performance also in the presence of a mismatch w.r.t. the training set, and even across different sensors. The proposed method, published online as an off-the-shelf software tool, allows users to perform fast and high-quality CNN-based pansharpening of their own target images on general-purpose hardware.