SPJFNet: Self-Mining Prior-Guided Joint Frequency Enhancement for Ultra-Efficient Dark Image Restoration

2025-08-05

AI Summary

To address efficiency bottlenecks in low-light image enhancement—stemming from reliance on external priors, redundant multi-stage designs, and indiscriminate frequency-domain processing—this paper proposes the Self-Mined Prior-guided Frequency Enhancement Network (SPFEN). Methodologically, SPFEN introduces: (i) a novel self-mined guidance module that generates lightweight spatial-frequency guidance signals endogenously, eliminating dependence on external priors; (ii) a dual-frequency collaborative framework that decouples high-frequency detail enhancement in the wavelet domain from low-frequency structural restoration in the Fourier domain, balancing computational efficiency and reconstruction fidelity; and (iii) lossless wavelet decomposition coupled with selective enhancement of advantageous frequencies, jointly optimized end-to-end for improved generalization. On multiple benchmarks, SPFEN surpasses state-of-the-art methods with 32% fewer parameters and 41% lower FLOPs, achieving superior speed-accuracy trade-offs.

Technology Category

• Computer Vision: Learning & Optimization for CV
• Search and Optimization: Learning to Search
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Security and Privacy: Privacy-enhancing technologies
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Current dark image restoration methods suffer from severe efficiency bottlenecks, primarily stemming from: (1) computational burden and error correction costs associated with reliance on external priors (manual or cross-modal); (2) redundant operations in complex multi-stage enhancement pipelines; and (3) indiscriminate processing across frequency components in frequency-domain methods, leading to excessive global computational demands. To address these challenges, we propose an Efficient Self-Mining Prior-Guided Joint Frequency Enhancement Network (SPJFNet). Specifically, we first introduce a Self-Mining Guidance Module (SMGM) that generates lightweight endogenous guidance directly from the network, eliminating dependence on external priors and thereby bypassing error correction overhead while improving inference speed. Second, through meticulous analysis of different frequency domain characteristics, we reconstruct and compress multi-level operation chains into a single efficient operation via lossless wavelet decomposition and joint Fourier-based advantageous frequency enhancement, significantly reducing parameters. Building upon this foundation, we propose a Dual-Frequency Guidance Framework (DFGF) that strategically deploys specialized high/low frequency branches (wavelet-domain high-frequency enhancement and Fourier-domain low-frequency restoration), decoupling frequency processing to substantially reduce computational complexity. Rigorous evaluation across multiple benchmarks demonstrates that SPJFNet not only surpasses state-of-the-art performance but also achieves significant efficiency improvements, substantially reducing model complexity and computational overhead. Code is available at https://github.com/bywlzts/SPJFNet.

Problem

• Eliminate external prior reliance in dark image restoration
• Reduce redundant operations in multi-stage enhancement pipelines
• Optimize frequency component processing to cut computational costs

Innovation

• Self-Mining Guidance Module for lightweight endogenous priors
• Lossless wavelet and joint Fourier frequency enhancement
• Dual-Frequency Guidance Framework decouples processing complexity