Temporally-Similar Structure-Aware Spatiotemporal Fusion of Satellite Images

TL;DR

This work tackles the challenge of producing high-spatial, high-temporal-resolution satellite image sequences under realistic noise. It introduces Temporally-Guided Total Variation (TGTV) and Temporally-Guided Edge Constraint (TGEC) to preserve spatial structure while denoising, formulating a constrained optimization solved by a preconditioned primal-dual splitting method with operator-based diagonal preconditioning. Adaptive alpha updating enables robust performance across diverse noise types, and empirical results on simulated and real data show TSSTF matches or surpasses state-of-the-art methods, especially under noise. The framework provides practical parameter guidelines and demonstrates strong potential for real-world ST fusion tasks, with future work addressing significant land-cover changes and cloud occlusions.

Abstract

This paper proposes a spatiotemporal (ST) fusion framework robust against diverse noise for satellite images, named Temporally-Similar Structure-Aware ST fusion (TSSTF). ST fusion is a promising approach to address the trade-off between the spatial and temporal resolution of satellite images. In real-world scenarios, observed satellite images are severely degraded by noise due to measurement equipment and environmental conditions. Consequently, some recent studies have focused on enhancing the robustness of ST fusion methods against noise. However, existing noise-robust ST fusion approaches often fail to capture fine spatial structure, leading to oversmoothing and artifacts. To address this issue, TSSTF introduces two key mechanisms: Temporally-Guided Total Variation (TGTV) and Temporally-Guided Edge Constraint (TGEC). TGTV is a weighted total variation-based regularization that promotes spatial piecewise smoothness while preserving structural details, guided by a reference high spatial resolution image acquired on a nearby date. TGEC enforces consistency in edge locations between two temporally adjacent images, while allowing for spectral variations. We formulate the ST fusion task as a constrained optimization problem incorporating TGTV and TGEC, and develop an efficient algorithm based on a preconditioned primal-dual splitting method. Experimental results demonstrate that TSSTF performs comparably to state-of-the-art methods under noise-free conditions and outperforms them under noisy conditions.

Figures (16)

• Figure 1: Illustration of our TSSTF.
• Figure 2: Our problem setting
• Figure 3: Illustration of the weights in (\ref{['eq: weight']}). A noise-attenuated guide image $\mathbf{h}_{\mathrm{r}}^{\prime}$ in (\ref{['eq: guide image']}) is first constructed from the reference HR image. The weights are then computed from this guide image according to (\ref{['eq: weight']}). In the right panel, each arrow corresponds to a weight; the darker the arrow, the greater the weight.
• Figure 4: Illustration of TGTV. TGTV evaluates the four neighborhood differences with adaptive weights derived from the guided HR image $\mathbf{h}_{\mathrm{r}}'$.
• Figure 5: Illustration of edge similarity. When the reference and target dates are temporally close, the locations of edges are likely to coincide. On the other hand, edge intensities may differ due to changes in the spectral brightness of surrounding regions over time.