Learning to Balance: Decoupled Siamese Diffusion Transformer for Reference-Based Remote Sensing Image Super-Resolution

2026-05-18

AI Summary

This work addresses the challenges of texture artifacts and detail loss in reference-based image super-resolution, which arise from either over-reliance on or insufficient utilization of reference information. To this end, the authors propose DS-DiT, a novel approach that, for the first time, enables decoupled interaction between low-resolution structural priors and reference textures within a diffusion model. Specifically, a decoupled Siamese diffusion Transformer separates these two information streams in the attention layers to independently guide the reconstruction of noisy latent variables. The method further incorporates a block-wise weighting module for adaptive fusion of multi-source conditions and introduces a self-guided inference strategy that requires no additional training. Extensive experiments demonstrate that DS-DiT consistently achieves superior quantitative performance and visual quality across multiple remote sensing datasets and large-scale super-resolution factors compared to existing methods.

Abstract

Diffusion-based methods demonstrate significant potential for remote sensing image super-resolution at large scaling factors, particularly in reference-based super-resolution (RefSR) where high-resolution reference images provide critical fine-grained texture priors. However, existing methods often suffer from a trade-off between over-reliance on reference information, which leads to texture artifacts, and underutilization, which results in insufficient detail recovery. To address these issues, we propose DS-DiT, a Decoupled Siamese Diffusion Transformer method that decouples low-resolution and reference interactions at the attention level. By enabling low-resolution structural priors and reference texture information to interact independently with the noisy latent, the framework effectively mitigates inter-source competition. Furthermore, to compensate for the limited local modeling ability of global attention, we introduce a Patch-Level Weights (PLW) module that adaptively modulates the fusion of conditional sources. In addition, this siamese architecture facilitates an autoguidance strategy during inference, which enhances reconstruction by exploiting the prediction discrepancy between strong and weak reference conditions. This approach boosts generation quality without additional training. Experimental results across multiple datasets and scaling factors demonstrate that DS-DiT outperforms existing methods in both quantitative metrics and visual fidelity.

Problem

• reference-based super-resolution
• remote sensing image
• texture artifacts
• detail recovery
• diffusion models

Innovation

• Decoupled Siamese Diffusion Transformer
• Reference-Based Super-Resolution
• Attention Decoupling
• Patch-Level Weights
• Autoguidance