WaveDiffUR: A diffusion SDE-based solver for ultra magnification super-resolution in remote sensing images

TL;DR

This paper targets ultra-resolution remote sensing super-resolution (UR-SR), framing the problem as a conditional diffusion stochastic differential equation and introducing WaveDiffUR, a wavelet-domain solver that splits the UR process into manageable sub-steps. It couples a plug-and-play pre-trained SR module for low-frequency reconstruction with a high-frequency upscaling pathway, and enforces dynamic cross-scale pyramid (CSP) constraints to maintain spectral-spatial fidelity at extreme magnifications. The method achieves superior performance in UR across multiple public datasets, with large gains in PSNR, SRE, and perceptual metrics, while maintaining efficiency through wavelet-based decomposition and self-cascade inference. These advancements enable scalable, high-fidelity remote sensing analysis at unprecedented magnifications, with potential impacts on environmental monitoring, urban planning, and agriculture.

Abstract

Deep neural networks have recently achieved significant advancements in remote sensing superresolu-tion (SR). However, most existing methods are limited to low magnification rates (e.g., 2 or 4) due to the escalating ill-posedness at higher magnification scales. To tackle this challenge, we redefine high-magnification SR as the ultra-resolution (UR) problem, reframing it as solving a conditional diffusion stochastic differential equation (SDE). In this context, we propose WaveDiffUR, a novel wavelet-domain diffusion UR solver that decomposes the UR process into sequential sub-processes addressing conditional wavelet components. WaveDiffUR iteratively reconstructs low-frequency wavelet details (ensuring global consistency) and high-frequency components (enhancing local fidelity) by incorporating pre-trained SR models as plug-and-play modules. This modularity mitigates the ill-posedness of the SDE and ensures scalability across diverse applications. To address limitations in fixed boundary conditions at extreme magnifications, we introduce the cross-scale pyramid (CSP) constraint, a dynamic and adaptive framework that guides WaveDiffUR in generating fine-grained wavelet details, ensuring consistent and high-fidelity outputs even at extreme magnification rates.

Figures (14)

• Figure 1: A comparison of SR performance across several existing PDF-based SR models using deep learning approaches, including our previous model shi2022latent.
• Figure 2: An Illustration of the proposed self-cascade UR pyramid framework, consisting of 1) DWT/IDWT: cyclically decompose a low-resolution image into wavelet domain, and restore the high-resolution image from the up-scaled wavelet-domain components. 2) SR pipeline: integrates a plug-and-play tunable SR module into the framework to reconstruct the low-frequency wavelet components of a high-resolution image from it's low-resolution counterpart. 3) Upscaler: progressively adapts the high-frequency wavelet components of the low-resolution images to match those of the higher-resolution images.
• Figure 3: An overall of the proposed CSP-WaveDiffUR.
• Figure 4: An illustration of the cross-scale pyramid encoder.
• Figure 5: An illustration of the cross-scale high-frequency restoration model.