Real-Time LiDAR Super-Resolution via Frequency-Aware Multi-Scale Fusion

2025-11-10

AI Summary

To address the challenge of achieving high-fidelity 3D perception with low-cost, low-resolution LiDAR sensors, this paper proposes FLASH—a novel framework that pioneers frequency-domain modeling for LiDAR point cloud super-resolution. Methodologically, FLASH introduces a frequency-aware windowed attention mechanism leveraging Fast Fourier Transform (FFT) to explicitly capture the periodic structural patterns inherent in LiDAR scans; replaces fixed skip connections with adaptive multi-scale feature fusion for position-sensitive dynamic feature aggregation; and employs a lightweight Transformer backbone synergistically integrating FFT, CBAM, and hierarchical feature interaction to enable joint optimization across both frequency and spatial domains. Evaluated on KITTI, FLASH consistently outperforms state-of-the-art methods—including TULIP—with particularly pronounced improvements in long-range accuracy. Moreover, it achieves real-time, high-fidelity reconstruction in a single forward pass.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Computer Vision: Multi-modal Vision
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

LiDAR super-resolution addresses the challenge of achieving high-quality 3D perception from cost-effective, low-resolution sensors. While recent transformer-based approaches like TULIP show promise, they remain limited to spatial-domain processing with restricted receptive fields. We introduce FLASH (Frequency-aware LiDAR Adaptive Super-resolution with Hierarchical fusion), a novel framework that overcomes these limitations through dual-domain processing. FLASH integrates two key innovations: (i) Frequency-Aware Window Attention that combines local spatial attention with global frequency-domain analysis via FFT, capturing both fine-grained geometry and periodic scanning patterns at log-linear complexity. (ii) Adaptive Multi-Scale Fusion that replaces conventional skip connections with learned position-specific feature aggregation, enhanced by CBAM attention for dynamic feature selection. Extensive experiments on KITTI demonstrate that FLASH achieves state-of-the-art performance across all evaluation metrics, surpassing even uncertainty-enhanced baselines that require multiple forward passes. Notably, FLASH outperforms TULIP with Monte Carlo Dropout while maintaining single-pass efficiency, which enables real-time deployment. The consistent superiority across all distance ranges validates that our dual-domain approach effectively handles uncertainty through architectural design rather than computationally expensive stochastic inference, making it practical for autonomous systems.

Problem

• Achieving high-quality 3D perception from low-resolution LiDAR sensors
• Overcoming spatial-domain limitations with restricted receptive fields
• Enabling real-time super-resolution without computationally expensive inference

Innovation

• Dual-domain processing combining spatial and frequency analysis
• Frequency-Aware Window Attention with FFT for global patterns
• Adaptive Multi-Scale Fusion with learned feature aggregation