GSPan: A Continuous Gaussian Primitive Representation for Arbitrary-Scale Pansharpening

2026-06-16

AI Summary

This work addresses the limitation of conventional pansharpening methods, which rely on fixed-grid predictions and struggle to support fusion at arbitrary scales. The authors propose GSPan, a novel framework that models band-wise residual details as continuous, learnable 2D Gaussian primitives. Through a dual-stream hierarchical interaction architecture equipped with a spatial-spectral interactive attention module, GSPan jointly estimates these primitives from panchromatic and multispectral inputs and renders them at any target resolution without retraining. An innovative scale-decoupled asymmetric inference strategy is introduced to significantly enhance computational efficiency for large-scale scenes. Extensive experiments on QuickBird, GaoFen-2, WorldView-3, and WorldView-3-4K datasets demonstrate that GSPan achieves state-of-the-art performance by striking an optimal balance between fusion quality and computational efficiency.

Abstract

Pansharpening aims to generate high-resolution multispectral (HRMS) images by fusing low-resolution multispectral (LRMS) and panchromatic (PAN) observations. Most existing deep learning methods treat pansharpening as fixed-grid prediction, which limits scale adaptation. To address this, we propose GSPan, a framework that introduces 2D Gaussian Splatting (GS) into pansharpening. Instead of directly predicting pixels, GSPan represents band-wise residual details as continuous and learnable 2D Gaussian primitives. We design a Dual-Stream Hierarchical Interaction (DSHI) architecture with a Spatial-Spectral Interactive Attention (SSIA) module to estimate these primitives from complementary PAN and MS observations. The predicted primitives are rendered as a residual detail field and injected into the upsampled MS image. This continuous representation allows GSPan to render fused images on arbitrary target sampling grids without scale-specific retraining. It further enables a Scale-Decoupled Asymmetric Inference (SDAI) strategy, which estimates primitives at a reduced resolution and renders the fused image at the target resolution for efficient large-scene pansharpening. Experiments on QuickBird, GaoFen-2, WorldView-3, and WorldView-3-4K datasets show that GSPan delivers state-of-the-art fusion performance. Moreover, SDAI markedly accelerates inference, achieving a favorable trade-off between computational efficiency and fusion quality. Our results demonstrate the potential of continuous Gaussian residual representations as a flexible and scale-decoupled alternative to fixed-grid prediction.

Problem

• pansharpening
• scale adaptation
• continuous representation
• arbitrary-scale
• fixed-grid prediction

Innovation

• Gaussian Splatting
• continuous representation
• arbitrary-scale pansharpening
• scale-decoupled inference
• residual detail field