Kernel Space Diffusion Model for Efficient Remote Sensing Pansharpening

2025-05-25

AI Summary

To address the dual bottlenecks of insufficient global prior modeling and high inference latency in remote sensing pan-sharpening, this paper proposes a kernel-space implicit diffusion framework. Instead of performing diffusion in pixel space, our method operates directly in the convolutional kernel space, enabling efficient global context integration via low-rank tensor decomposition and unified factor learning for dynamic kernel generation. We further design a structure-aware multi-head attention mechanism to enhance geometric consistency and introduce a two-stage adaptive training strategy to improve convergence and generalization. Evaluated on WorldView-3, Gaofen-2, and QuickBird datasets, our approach achieves state-of-the-art performance, with significant gains in PSNR and SSIM. Moreover, it accelerates inference by 5.2× over pixel-space diffusion models, effectively balancing accuracy and efficiency.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Kernel Methods
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Pansharpening is a fundamental task in remote sensing that integrates high-resolution panchromatic imagery (PAN) with low-resolution multispectral imagery (LRMS) to produce an enhanced image with both high spatial and spectral resolution. Despite significant progress in deep learning-based approaches, existing methods often fail to capture the global priors inherent in remote sensing data distributions. Diffusion-based models have recently emerged as promising solutions due to their powerful distribution mapping capabilities; however, they suffer from significant inference latency, which limits their practical applicability. In this work, we propose the Kernel Space Diffusion Model (KSDiff), a novel approach that leverages diffusion processes in a latent space to generate convolutional kernels enriched with global contextual information, thereby improving pansharpening quality while enabling faster inference. Specifically, KSDiff constructs these kernels through the integration of a low-rank core tensor generator and a unified factor generator, orchestrated by a structure-aware multi-head attention mechanism. We further introduce a two-stage training strategy tailored for pansharpening, enabling KSDiff to serve as a framework for enhancing existing pansharpening architectures. Experiments on three widely used datasets, including WorldView-3, GaoFen-2, and QuickBird, demonstrate the superior performance of KSDiff both qualitatively and quantitatively. Code will be released upon possible acceptance.

Problem

• Enhancing remote sensing images with high spatial and spectral resolution
• Reducing inference latency in diffusion-based pansharpening models
• Capturing global priors in remote sensing data distributions

Innovation

• Leverages diffusion processes in latent space
• Generates convolutional kernels with global context
• Uses two-stage training for pansharpening enhancement