Towards Arbitrary-Scale Spacecraft Image Super-Resolution via Salient Region-Guidance

TL;DR

This work tackles arbitrary-scale super-resolution for spacecraft imagery by introducing SGSASR, a saliency-guided framework that emphasizes the spacecraft core region. It combines a spacecraft image feature extraction encoder (SFEEM) with a two-stage latent-modulation decoder (LMASSRM); the encoder uses SCRRB to locate salient regions and AFFEM to adaptively fuse core-region features, while the decoder employs latent modulation via LMGB and ASRB to render HR outputs at arbitrary scales. Key contributions include the SCRRB for robust saliency-guided focus, the AFFEM for dynamic feature fusion, and the LMASSRM's FiLM-based modulation and render-MLP for continuous, scale-free reconstruction. Empirical results on radar and optical spacecraft datasets show SGSASR achieving state-of-the-art performance in PSNR/SSIM across multiple scales with competitive efficiency, highlighting its practical potential for onboard processing and mission-critical tasks.

Abstract

Spacecraft image super-resolution seeks to enhance low-resolution spacecraft images into high-resolution ones. Although existing arbitrary-scale super-resolution methods perform well on general images, they tend to overlook the difference in features between the spacecraft core region and the large black space background, introducing irrelevant noise. In this paper, we propose a salient region-guided spacecraft image arbitrary-scale super-resolution network (SGSASR), which uses features from the spacecraft core salient regions to guide latent modulation and achieve arbitrary-scale super-resolution. Specifically, we design a spacecraft core region recognition block (SCRRB) that identifies the core salient regions in spacecraft images using a pre-trained saliency detection model. Furthermore, we present an adaptive-weighted feature fusion enhancement mechanism (AFFEM) to selectively aggregate the spacecraft core region features with general image features by dynamic weight parameter to enhance the response of the core salient regions. Experimental results demonstrate that the proposed SGSASR outperforms state-of-the-art approaches.

Figures (12)

• Figure 1: Example of spacecraft image, with a large black space background.
• Figure 2: The overall architecture of salient region-guided spacecraft image arbitrary-scale super-resolution network (SGASR).
• Figure 3: The architecture of spacecraft image feature extraction enhancement module (SFEEM).
• Figure 4: (a) Spacecraft core region recognition block (SCRRB). (b) Nonlinear activation free block (NAFBlock). (c) Simplified channel attention (SCA). (d) Simple gate. (e) Adaptive-weighted feature fusion enhancement mechanism (AFFEM).
• Figure 5: The architecture of latent modulation-based arbitrary-scale super-resolution module (LMASSRM).