Lightweight and Fast Real-time Image Enhancement via Decomposition of the Spatial-aware Lookup Tables

2025-08-22

AI Summary

Traditional 3D LUT-based image enhancement methods lack spatial awareness due to pointwise color mapping; while existing spatially aware approaches mitigate this limitation, they incur high parameter counts and resolution-dependent computational overhead. This paper proposes a Decomposed Spatially-aware 3D LUT (DS-LUT), which linearly decomposes the original 3D LUT into multiple low-dimensional LUTs, applies Singular Value Decomposition (SVD) to compress redundancy, and introduces a lightweight feature fusion module to efficiently model spatial context. DS-LUT preserves both color fidelity and spatial adaptability while significantly reducing model parameters (up to 72% reduction) and inference latency (3.1× speedup on 4K images), with computational complexity independent of input resolution. Extensive experiments demonstrate that DS-LUT achieves state-of-the-art performance across multiple benchmarks, striking an optimal balance between real-time efficiency and enhancement quality.

Technology Category

Computer Vision: 3D Computer VisionMachine Learning: Hardware-aware MLKnowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMsSearch and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingUser Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

The image enhancement methods based on 3D lookup tables (3D LUTs) efficiently reduce both model size and runtime by interpolating pre-calculated values at the vertices. However, the 3D LUT methods have a limitation due to their lack of spatial information, as they convert color values on a point-by-point basis. Although spatial-aware 3D LUT methods address this limitation, they introduce additional modules that require a substantial number of parameters, leading to increased runtime as image resolution increases. To address this issue, we propose a method for generating image-adaptive LUTs by focusing on the redundant parts of the tables. Our efficient framework decomposes a 3D LUT into a linear sum of low-dimensional LUTs and employs singular value decomposition (SVD). Furthermore, we enhance the modules for spatial feature fusion to be more cache-efficient. Extensive experimental results demonstrate that our model effectively decreases both the number of parameters and runtime while maintaining spatial awareness and performance.

Problem

• Reduces parameters and runtime in spatial-aware 3D LUT methods
• Addresses redundancy in image-adaptive lookup table generation
• Maintains spatial awareness while improving computational efficiency

Innovation

• Decomposes 3D LUTs via SVD into low-dimensional linear sums
• Enhances cache-efficient spatial feature fusion modules
• Generates image-adaptive LUTs by targeting redundant table parts