ShiftLUT: Spatial Shift Enhanced Look-Up Tables for Efficient Image Restoration

2026-02-28

AI Summary

This work addresses the high computational and memory costs of existing lookup table (LUT)-based image restoration methods when enlarging the receptive field, which hinders their deployment on edge devices. To overcome this limitation, the authors propose ShiftLUT, a novel framework that effectively expands the receptive field through learnable spatial shifting (LSS). ShiftLUT employs an asymmetric dual-branch architecture to optimize computational resource allocation and introduces error-bounded adaptive sampling (EAS) for feature-level LUT compression. Compared to TinyLUT, ShiftLUT achieves a 3.8× larger receptive field and improves average PSNR by over 0.21 dB across multiple benchmark datasets, while maintaining low memory footprint and inference latency suitable for edge deployment.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Computer Vision: Adversarial Attacks & Robustness
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Look-Up Table based methods have emerged as a promising direction for efficient image restoration tasks. Recent LUT-based methods focus on improving their performance by expanding the receptive field. However, they inevitably introduce extra computational and storage overhead, which hinders their deployment in edge devices. To address this issue, we propose ShiftLUT, a novel framework that attains the largest receptive field among all LUT-based methods while maintaining high efficiency. Our key insight lies in three complementary components. First, Learnable Spatial Shift module (LSS) is introduced to expand the receptive field by applying learnable, channel-wise spatial offsets on feature maps. Second, we propose an asymmetric dual-branch architecture that allocates more computation to the information-dense branch, substantially reducing inference latency without compromising restoration quality. Finally, we incorporate a feature-level LUT compression strategy called Error-bounded Adaptive Sampling (EAS) to minimize the storage overhead. Compared to the previous state-of-the-art method TinyLUT, ShiftLUT achieves a 3.8$\times$ larger receptive field and improves an average PSNR by over 0.21 dB across multiple standard benchmarks, while maintaining a small storage size and inference time.

Problem

• Look-Up Table
• Image Restoration
• Receptive Field
• Computational Overhead
• Storage Overhead

Innovation

• Learnable Spatial Shift
• Asymmetric Dual-Branch Architecture
• Error-bounded Adaptive Sampling
• Look-Up Table
• Efficient Image Restoration