- Rating
- 3
- Confidence
- 4(+2)
- Informativeness
- 2
- Presentation
- 3(+1)
- Soundness
- 3
Read review
Paper summary
This paper proposes OSSNet, a pansharpening framework that formulates observation-guided feature refinement as a learned three-level multigrid-inspired V-cycle. It introduces an SVD-free Chebyshev-Guided Approximation Mechanism to construct data-adaptive, row-orthogonal projections, together with nullspace compensation to preserve complementary residuals. Experiments on multiple satellite datasets demonstrate competitive reconstruction quality, strong cross-sensor generalization, and substantially fewer parameters and FLOPs.
Critical strengths
1.OSSNet achieves competitive reconstruction quality while using significantly fewer parameters and FLOPs, demonstrating a favorable accuracy–efficiency trade-off.
2.The proposed SVD-free Chebyshev-Guided Approximation Mechanism constructs data-adaptive, approximately row-orthogonal projections without repeated SVD, avoiding the numerical instability of SVD when embedded in neural network training.
Critical weaknesses
The proposed method shows no clear advantage in full-resolution evaluation.
Questions for authors
In Section 3.1, the authors introduce a fixed random subspace to avoid constructing the full covariance matrix. However, Eq. (16) directly computes the channel covariance matrix. Could the authors further clarify the motivation for introducing the random projection and its role in the actual OSSNet implementation?
5 comments
Thank you for the detailed rebuttal. To better validate the performance trade-offs and practical efficiency of the proposed method, I have two remaining requests:
- Ablation on Chebyshev Order : Please evaluate different values (with fixed) in terms of reconstruction performance, parameters, FLOPs, training time, and inference latency.
- Practical Runtime: Beyond parameters and FLOPs, please report the training time and inference latency across all ablation variants (especially Chebyshev CSS vs. Average Pooling).
Thank you for the follow-up and for specifying these two additional evaluations. We address them separately below.
For computational benchmarking in both evaluations, we use an NVIDIA GeForce RTX 5060 Ti GPU with a batch size of 1. The network inputs are an upsampled four-band multispectral (LMS) tensor of size and a panchromatic (PAN) tensor of size (batch channels height width). Both inputs are therefore represented at the PAN spatial resolution; at a scale factor of 4, the corresponding original low-resolution multispectral image has a spatial size of . These settings apply to the computational benchmarks; the WV3 and WV2 reconstruction results are reported separately below.
1. Ablation on Chebyshev order with fixed
Our previous comparison varied and jointly and therefore did not isolate the effect of polynomial order. We now evaluate with fixed at 1e-4. Table 4 provides the complementary sensitivity analysis for at fixed .
| WV3 PSNR ↑ | ERGAS ↓ | SAM ↓ | WV2 PSNR ↑ | HQNR ↑ | ↓ | ↓ | |
|---|---|---|---|---|---|---|---|
| 1 | 37.09 | 2.27 | 3.32 | 29.46 | 0.9450 | 0.023 | 0.032 |
| 2 | 37.11 | 2.26 | 3.31 | 29.87 | 0.9476 | 0.023 | 0.030 |
| 3 | 37.22 | 2.23 | 3.29 | 30.24 | 0.9523 | 0.023 | 0.027 |
| 4 | 37.22 | 2.24 | 3.28 | 30.10 | 0.9499 | 0.024 | 0.027 |
| 5 | 37.23 | 2.24 | 3.28 | 29.98 | 0.9511 | 0.025 | 0.024 |
| 6 | 37.24 | 2.27 | 3.31 | 30.15 | 0.9499 | 0.025 | 0.026 |
| 7 | 37.22 | 2.24 | 3.29 | 29.84 | 0.9505 | 0.025 | 0.026 |
| 8 | 37.15 | 2.25 | 3.31 | 30.08 | 0.9483 | 0.025 | 0.028 |
Reconstruction quality generally improves from low to moderate orders, while further increases yield smaller and non-monotonic changes. achieves the highest WV3 PSNR in this sweep, whereas achieves the highest HQNR and WV2 PSNR. Increasing from 1 to 5 reduces from 0.032 to 0.024, and increases from 0.023 to 0.025. The sweep characterizes the spatial–spectral trade-off under a fixed orthogonality weight.
| Parameters (M) | FLOPs (G) | Latency (ms) | Peak memory (MB) | |
|---|---|---|---|---|
| 1 | 0.17 | 3.89 | 6.66 | 73.66 |
| 2 | 0.17 | 3.95 | 6.81 | 73.67 |
| 3 | 0.17 | 4.10 | 6.87 | 73.68 |
| 4 | 0.17 | 4.15 | 7.00 | 73.69 |
| 5 | 0.17 | 4.20 | 7.21 | 73.70 |
| 6 | 0.17 | 4.26 | 7.33 | 73.70 |
| 7 | 0.18 | 4.30 | 7.54 | 73.71 |
| 8 | 0.18 | 4.35 | 7.71 | 73.72 |
Across the sweep, FLOPs increase from 3.89 to 4.35 G, while parameter counts remain within 0.17–0.18 M at the reported precision. The measured latency ranges from 6.66 to 7.71 ms, and peak memory changes only slightly.
We sincerely thank the reviewer for the positive assessment of our accuracy–efficiency trade-off and SVD-free Chebyshev mechanism. We address both points below.
W1. Full-Resolution Evaluation and Underlying Trade-offs (Table 2 & Appendix D, Tables 7–8)
We clarify OSSNet's baseline efficiency–accuracy standing, root-cause mechanism, and our targeted parameterization that achieves full-resolution SOTA across all benchmarks:
Q1. Random Projection Ω in Section 3.1 vs. Direct Covariance in Eq. (16)
CSS and CCS address different compression scales, and the code and equations are fully consistent. Section 3.1 (Eqs. 1–4) gives the general projected-SVD reference rather than requiring every sketch to use random projection: