Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MarsSQE: Stereo Quality Enhancement for Martian Images Using Bi-level Cross-view Attention

Mai Xu, Yinglin Zhu, Qunliang Xing, Jing Yang, Xin Zou

TL;DR

This work addresses quality degradation in compressed stereo Martian images transmitted from Mars. It introduces MarsSQE, a bi-level cross-view attention network that exploits intra-view and cross-view information through patch-level and pixel-level attention to enhance both left and right views. A stereo Martian image dataset with rich cross-view correlations is established, and MarsSQE achieves superior PSNR/SSIM and rate-distortion performance compared with monocular and binocular baselines. The results demonstrate a practical path to higher-quality Martian imagery, improving scientific analysis and navigation under bandwidth-constrained communication.

Abstract

Stereo images captured by Mars rovers are transmitted after lossy compression due to the limited bandwidth between Mars and Earth. Unfortunately, this process results in undesirable compression artifacts. In this paper, we present a novel stereo quality enhancement approach for Martian images, named MarsSQE. First, we establish the first dataset of stereo Martian images. Through extensive analysis of this dataset, we observe that cross-view correlations in Martian images are notably high. Leveraging this insight, we design a bi-level cross-view attention-based quality enhancement network that fully exploits these inherent cross-view correlations. Specifically, our network integrates pixel-level attention for precise matching and patch-level attention for broader contextual information. Experimental results demonstrate the effectiveness of our MarsSQE approach.

MarsSQE: Stereo Quality Enhancement for Martian Images Using Bi-level Cross-view Attention

TL;DR

This work addresses quality degradation in compressed stereo Martian images transmitted from Mars. It introduces MarsSQE, a bi-level cross-view attention network that exploits intra-view and cross-view information through patch-level and pixel-level attention to enhance both left and right views. A stereo Martian image dataset with rich cross-view correlations is established, and MarsSQE achieves superior PSNR/SSIM and rate-distortion performance compared with monocular and binocular baselines. The results demonstrate a practical path to higher-quality Martian imagery, improving scientific analysis and navigation under bandwidth-constrained communication.

Abstract

Stereo images captured by Mars rovers are transmitted after lossy compression due to the limited bandwidth between Mars and Earth. Unfortunately, this process results in undesirable compression artifacts. In this paper, we present a novel stereo quality enhancement approach for Martian images, named MarsSQE. First, we establish the first dataset of stereo Martian images. Through extensive analysis of this dataset, we observe that cross-view correlations in Martian images are notably high. Leveraging this insight, we design a bi-level cross-view attention-based quality enhancement network that fully exploits these inherent cross-view correlations. Specifically, our network integrates pixel-level attention for precise matching and patch-level attention for broader contextual information. Experimental results demonstrate the effectiveness of our MarsSQE approach.
Paper Structure (8 sections, 4 equations, 5 figures, 3 tables)

This paper contains 8 sections, 4 equations, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Stereo Martian Image Dataset
  3. MarsSQE Approach
  4. Experiments
  5. Experimental Settings
  6. Evaluation
  7. Ablation Study
  8. Conclusion

Figures (5)

  • Figure 1: Overall framework of the proposed MarsSQE approach.
  • Figure 2: Sample images in our dataset of four different landforms.
  • Figure 3: Illustration of the patch-level attention module (left) and the pixel-level attention module (right).
  • Figure 4: Rate-Distortion (RD) curves of MarsSQE and compared approaches. The rate is measured in bits per pixel (bpp).
  • Figure 5: Visual comparison of our MarsSQE and other approaches under QF=30.