用语义同义性重构图像压缩，提升感知质量
Synonymous Variational Inference for Perceptual Image Compression
- 基于语义同义关系建模潜在表示，优化感知相似性
- 单个渐进式编码器实现媲美现有方法的率失真感知性能
- 为感知图像压缩提供理论支撑，适合图像压缩研究者
近期语义信息论研究表明，语义与句法信息之间存在集合元素关系，表现为同义关系。本文基于这一同义性视角，提出一种同义变分推断（SVI）方法，重新分析感知图像压缩问题。以感知相似性作为典型同义标准，构建理想同义集（Synset），并通过最小化部分语义KL散度，用参数化密度逼近其潜在同义表示的后验分布。该分析理论上证明了感知图像压缩的优化方向遵循三重权衡，涵盖现有率-失真-感知方案。此外，我们提出同义图像压缩（SIC）新范式，对应SVI的分析过程，并实现一个渐进式SIC编解码器以充分发挥模型能力。实验表明，仅用单一渐进式SIC编码器即可达到与现有方法相当的率失真感知性能，验证了所提分析方法的有效性。
原文摘要 · Abstract (English)
Recent contributions of semantic information theory reveal the set-element relationship between semantic and syntactic information, represented as synonymous relationships. In this paper, we propose a synonymous variational inference (SVI) method based on this synonymity viewpoint to re-analyze the perceptual image compression problem. It takes perceptual similarity as a typical synonymous criterion to build an ideal synonymous set (Synset), and approximate the posterior of its latent synonymous representation with a parametric density by minimizing a partial semantic KL divergence. This analysis theoretically proves that the optimization direction of perception image compression follows a triple tradeoff that can cover the existing rate-distortion-perception schemes. Additionally, we introduce synonymous image compression (SIC), a new image compression scheme that corresponds to the analytical process of SVI, and implement a progressive SIC codec to fully leverage the model's capabilities. Experimental results demonstrate comparable rate-distortion-perception performance using a single progressive SIC codec, thus verifying the effectiveness of our proposed analysis method.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。