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

2024-12-30

AI Summary

To address severe compression artifacts and significant quality degradation in rover-downlinked stereo images from Mars—caused by stringent deep-space bandwidth constraints—this paper introduces the first Mars-specific stereo image dataset and proposes the Dual-level Cross-view Attention Quality Enhancement Network (DCVA-Net). DCVA-Net is the first to explicitly model and exploit the inherent high cross-view correlation in Martian stereo imagery, introducing a novel dual-level (pixel-wise and patch-wise) cross-view attention mechanism that jointly fuses local correspondence and global contextual information. It further incorporates a cross-view feature alignment module to enhance inter-view consistency. Extensive experiments demonstrate that DCVA-Net achieves substantial gains over state-of-the-art methods in PSNR (+1.27 dB) and SSIM (+0.032), effectively suppressing compression artifacts. The results validate the critical role of cross-view priors under extreme compression conditions.

Technology Category

• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multi-instance/Multi-view Learning

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Web Mining and Content Analysis: Web data integration and cleaning

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.

Problem

• Mars Rover
• Stereo Images
• Data Compression

Innovation

• MarsSQE
• Stereo Image Enhancement
• Multi-view Similarity Features