Hyperspectral imagery provides rich spectral information, but commonly suffers from low spatial resolution. This problem can be addressed with hyperspectral pansharpening which benefits from a high-resolution panchromatic image to enhance the resolution of spectral bands. Another possibility of improving spatial resolution, commonly exploited for remotely-sensed images, benefits from multitemporal fusion of images acquired at different revisits of a satellite. However, such a process, termed multi-image super-resolution, has not been applied to hyperspectral imagery due to the excessive size of hyperspectral cubes. In this paper, we propose a new processing pipeline that combines these two approaches. A multitemporal series of panchromatic images is super-resolved to reconstruct a high-resolution panchromatic image which is subsequently used during pansharpening. We demonstrate the benefits of the proposed pipeline for PRISMA imagery, whose hyperspectral cubes (30 m ground sampling distance) are associated with panchromatic images (5 m ground sampling distance). The latter are upsampled by a factor of 3× relying on multi-image super-resolution, which allows our pansharpening to achieve the upsampling ratio of 18${\times }$ for PRISMA images. We report and discuss the results of our extensive experiments which indicate that the proposed approach offers substantial quantitative and qualitative improvement, confirmed by a mean opinion score survey.
Single-image super-resolution refers to the reconstruction of a high-resolution image from a single low-resolution observation. Although recent deep learning-based methods have demonstrated notable success on simulated datasets—with low-resolution images obtained by degrading and downsampling high-resolution ones—they frequently fail to generalize to real-world settings, such as document scans, which are affected by complex degradations and semantic variability. In this study, we introduce a task-driven, multi-task learning framework for training a super-resolution network specifically optimized for optical character recognition tasks. We propose to incorporate auxiliary loss functions derived from high-level vision tasks, including text detection using the connectionist text proposal network (CTPN), text recognition via a convolutional recurrent neural network (CRNN), keypoints localization using Key.Net, and hue consistency. To balance these diverse objectives, we employ a dynamic weight averaging (DWA) mechanism, which adaptively adjusts the relative importance of each loss term based on its convergence behavior. Experimental evaluation demonstrates that the proposed approach improves text detection, measured with intersection over union, by 1.09% for simulated and 1.94% for real-world datasets containing scanned documents, while preserving overall image fidelity. These improvements are statistically significant as confirmed by the Kruskal–Wallis H test and the post hoc Dunn test with Benjamini–Hochberg p-value correction. Our findings highlight the value of multi-objective optimization in super-resolution models for bridging the gap between simulated training regimes and practical deployment in real-world scenarios.