arXiv 2410.026402024-10eess.IVcs.CV

RDEIC：利用中继残差扩散加速基于扩散的极限图像压缩

RDEIC: Accelerating Diffusion-Based Extreme Image Compression with Relay Residual Diffusion

Institute of Artificial Intelligence and Robotics, Xi’an Jiaotong University（人工智能与机器人研究所，西安交通大学）；Department of Computer Science and Software Engineering, The University of Western Australia（计算机科学与软件工程系，西澳大学）

Zhiyuan Li, Yanhui Zhou, Hao Wei, Chenyang Ge, Ajmal Mian

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

针对基于扩散的极限图像压缩在极低比特率下保真度和效率受限的问题，提出RDEIC方法，利用压缩特征初始化和残差扩散加速去噪，并采用固定步长微调，在保真度和效率上均达到最优性能。

摘要：

基于扩散的极限图像压缩方法在极低比特率下取得了令人瞩目的性能。然而，受限于从纯噪声开始的迭代去噪过程，这些方法在保真度和效率方面均受到限制。为了解决这两个问题，我们提出了中继残差扩散极限图像压缩（RDEIC），该方法利用压缩特征初始化和残差扩散。具体而言，首先，我们使用图像压缩后的潜在特征并添加噪声，而非纯噪声，作为起始点，以消除去噪过程中不必要的初始阶段。其次，我们直接从稳定扩散（Stable Diffusion）的原始扩散方程推导出一种新颖的残差扩散方程，通过迭代去除添加的噪声以及压缩与目标潜在特征之间的残差来重建原始图像。通过这种方式，我们有效地结合了残差扩散的效率与稳定扩散强大的生成能力。第三，我们提出了一种固定步长微调策略，以消除训练与推理阶段之间的差异，从而进一步提高重建质量。大量实验表明，所提出的RDEIC实现了最先进的视觉质量，并在保真度和效率方面均优于现有的基于扩散的极限图像压缩方法。源代码和预训练模型可在 https://github.com/huai-chang/RDEIC 获取。

英文摘要：

Diffusion-based extreme image compression methods have achieved impressive performance at extremely low bitrates. However, constrained by the iterative denoising process that starts from pure noise, these methods are limited in both fidelity and efficiency. To address these two issues, we present Relay Residual Diffusion Extreme Image Compression (RDEIC), which leverages compressed feature initialization and residual diffusion. Specifically, we first use the compressed latent features of the image with added noise, instead of pure noise, as the starting point to eliminate the unnecessary initial stages of the denoising process. Second, we directly derive a novel residual diffusion equation from Stable Diffusion's original diffusion equation that reconstructs the raw image by iteratively removing the added noise and the residual between the compressed and target latent features. In this way, we effectively combine the efficiency of residual diffusion with the powerful generative capability of Stable Diffusion. Third, we propose a fixed-step fine-tuning strategy to eliminate the discrepancy between the training and inference phases, thereby further improving the reconstruction quality. Extensive experiments demonstrate that the proposed RDEIC achieves state-of-the-art visual quality and outperforms existing diffusion-based extreme image compression methods in both fidelity and efficiency. The source code and pre-trained models are available at https://github.com/huai-chang/RDEIC.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。