ProGIC: Progressive and Lightweight Generative Image Compression with Residual Vector Quantization

2026-03-03

AI Summary

This work proposes a lightweight progressive image compression framework based on residual vector quantization (RVQ) to address the limitations of existing generative approaches, which suffer from large model sizes and rigid architectures that hinder efficient deployment at low bitrates. The method employs multi-stage residual encoding to enable coarse-to-fine reconstruction and progressive bitstream generation. Its backbone integrates depthwise-separable convolutions with compact attention modules, significantly reducing computational overhead while preserving generative fidelity. Evaluated on the Kodak dataset, the proposed approach achieves 57.57% and 58.83% bitrate savings over MS-ILLM under DISTS and LPIPS metrics, respectively, and offers more than a tenfold acceleration in encoding and decoding speed. Furthermore, it supports cross-platform deployment across GPU/CPU and facilitates progressive transmission.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Recent advances in generative image compression (GIC) have delivered remarkable improvements in perceptual quality. However, many GICs rely on large-scale and rigid models, which severely constrain their utility for flexible transmission and practical deployment in low-bitrate scenarios. To address these issues, we propose Progressive Generative Image Compression (ProGIC), a compact codec built on residual vector quantization (RVQ). In RVQ, a sequence of vector quantizers encodes the residuals stage by stage, each with its own codebook. The resulting codewords sum to a coarse-to-fine reconstruction and a progressive bitstream, enabling previews from partial data. We pair this with a lightweight backbone based on depthwise-separable convolutions and small attention blocks, enabling practical deployment on both GPUs and CPU-only devices. Experimental results show that ProGIC attains comparable compression performance compared with previous methods. It achieves bitrate savings of up to 57.57% on DISTS and 58.83% on LPIPS compared to MS-ILLM on the Kodak dataset. Beyond perceptual quality, ProGIC enables progressive transmission for flexibility, and also delivers over 10 times faster encoding and decoding compared with MS-ILLM on GPUs for efficiency.

Problem

• Generative Image Compression
• Low-bitrate
• Model Flexibility
• Practical Deployment
• Progressive Transmission

Innovation

• Progressive Compression
• Residual Vector Quantization
• Lightweight Architecture
• Generative Image Compression
• Efficient Deployment