In the New Space era, commercial microsatellite development has expanded rapidly, driven by the need for
lightweight, low-power, and cost-efficient satellites weighing under 100 kg. Microsatellite constellations, often comprising dozens to hundreds of satellites, are widely employed for Earth observation. This proliferation has created a demand for advanced analytical techniques to utilize microsatellite imagery effectively in applications such as flood monitoring, marine observation, agriculture, and forestry. Flood mapping poses unique challenges, particularly due to inaccuracies in estimating flood-affected areas caused by errors in delineating water body boundaries. High-resolution imagery can mitigate these errors, making super-resolution (SR) techniques critical [1-2]. SR refers to the process of reconstructing high-resolution (HR) images from low-resolution (LR) inputs and has been extensively studied in image processing and deep learning domains. Prominent SR models include Convolutional Neural Networks (CNN), Generative Adversarial Networks (GAN), and Transformer-based architectures.