Uncover Treasures in DCT: Advancing JPEG Quality Enhancement by Exploiting Latent Correlations

2025-06-26

AI Summary

To address the low efficiency and weak performance of JPEG image quality enhancement in the DCT domain, this paper proposes AJQE—a novel adaptive JPEG quality enhancement framework. We first identify and model two inherent correlations among DCT coefficients: intra-block local frequency-domain correlation and inter-block global correlation. Based on this, we design a lightweight correlation-aware feature extraction module and construct an adaptive enhancement network compatible with diverse pixel-domain architectures, enabling end-to-end DCT-domain image restoration. Unlike conventional pixel-domain pipelines involving decode-enhance-reencode—costly in computation and memory—AJQE operates directly in the DCT domain, preserving structural consistency while significantly improving efficiency. Experiments demonstrate that AJQE achieves an average PSNR gain of 0.35 dB over state-of-the-art pixel-domain methods, with 60.5% higher throughput and substantially reduced computational cost, establishing a new paradigm for efficient compressed-domain image enhancement.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Distributed Search

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Joint Photographic Experts Group (JPEG) achieves data compression by quantizing Discrete Cosine Transform (DCT) coefficients, which inevitably introduces compression artifacts. Most existing JPEG quality enhancement methods operate in the pixel domain, suffering from the high computational costs of decoding. Consequently, direct enhancement of JPEG images in the DCT domain has gained increasing attention. However, current DCT-domain methods often exhibit limited performance. To address this challenge, we identify two critical types of correlations within the DCT coefficients of JPEG images. Building on this insight, we propose an Advanced DCT-domain JPEG Quality Enhancement (AJQE) method that fully exploits these correlations. The AJQE method enables the adaptation of numerous well-established pixel-domain models to the DCT domain, achieving superior performance with reduced computational complexity. Compared to the pixel-domain counterparts, the DCT-domain models derived by our method demonstrate a 0.35 dB improvement in PSNR and a 60.5% increase in enhancement throughput on average.

Problem

• Enhancing JPEG quality by exploiting DCT coefficient correlations
• Reducing computational costs of pixel-domain JPEG enhancement
• Improving performance of DCT-domain JPEG enhancement methods

Innovation

• Exploits latent DCT coefficient correlations
• Adapts pixel-domain models to DCT
• Enhances JPEG quality with lower complexity