Variable-Rate Deep Image Compression based on Low-Rank Adaptation by Progressive Learning

2026-06-14

AI Summary

This work addresses the challenges of variable-rate deep image compression, where existing approaches either require multiple models—leading to substantial storage overhead—or rely on a single model that incurs high computational complexity and performance degradation. To overcome these limitations, the authors propose a progressive learning framework based on Low-Rank Adaptation (LoRA), integrating reparameterizable rate-adaptive LoRA modules into the compression network. This design enables efficient variable-rate compression without additional computational cost during inference. Notably, the method achieves competitive compression performance comparable to multi-model strategies while requiring only 10% of the training data, 3% of the training steps, and reducing parameter storage by 99%.

Abstract

In the digital age, image compression is crucial for numerous applications, including web media, streaming services, high-resolution medical imaging, and connected vehicle networks, enabling efficient data storage and transmission. With the increasing demand for high-quality image communication, the need for advanced compression techniques becomes increasingly critical. Numerous Deep Image Compression (DIC) techniques have recently been introduced, showing impressive performance compared to traditional standards. However, variable-rate image compression remains an unresolved issue. Specific DIC methods deploy multiple networks to attain different compression rates, whereas others use a single model, which often results in higher computational complexity and reduced performance. This work proposes a progressive learning approach for variable-rate image compression based on the parameter-efficient fine-tuning method, the Low-Rank Adaptation (LoRA). We introduce an additional LoRA Rate-Adaptive Module (LoRAM) in DIC methods. Due to the re-parameterized merging of LoRA, our proposed method does not introduce additional computational complexity during inference. Compared to methods utilizing multiple models, comprehensive experiments demonstrate that our approach achieves competitive performance, saving 99\% in parameter storage, 90% in datasets, and 97% in training steps.

Problem

• Variable-Rate Image Compression
• Deep Image Compression
• Computational Complexity
• Parameter Efficiency
• Rate Adaptation

Innovation

• Variable-Rate Compression
• Low-Rank Adaptation
• Deep Image Compression
• Parameter-Efficient Fine-Tuning
• Progressive Learning