GAN-Based Quantization Table Optimization for Enhanced JPEG Image Compression
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
The increasing dependence on image-centric applications has created a strong need for compression techniques that reduce storage and bandwidth usage while maintaining visual quality. Although JPEG is widely used, it relies on fixed quantization tables (QTs) that do not adapt to image content, resulting in visible artifacts and loss of fine details at low bitrates. Earlier works attempted to improve JPEG using optimization techniques such as Genetic Algorithms (GA), Particle Swarm Optimization (PSO), and Differential Evolution (DE), but these methods involve high computational cost and are unsuitable for real-time applications. Deep learning-based methods like QNet provide adaptive quantization but require large datasets, complex training, and heavy processing resources, which limits their practicality in resource-constrained environments. To overcome these issues, this work proposes a GAN-inspired JPEG approach that does not involve any GAN training or discriminator–generator learning process. Instead, the method adapts quantization behaviour using local image characteristics such as variance and texture strength, enabling content-aware adjustment in a lightweight manner. The approach is evaluated on standard images using PSNR, MSE, bitrate comparison, and visual analysis, and the results show consistent improvement over standard JPEG, especially at 10% and 25% compression levels, providing better visual quality with similar compression efficiency. Thus, the proposed method offers a practical and efficient solution for high-quality JPEG compression without the need for complex deep learning training.
Fields of Study & Subjects
Advanced Data Compression Techniques
Image and Video Quality Assessment
Video Coding and Compression Technologies
Publication details
- Journal / source
- International Conference on Intelligent Systems for Communication, IoT and Security (ICISCoIS)
- Pages
- 1–6
- Publication date
- 8 January 2026
- Language
- EN