Transformations in Learned Image Compression from a Modulation Perspective

TL;DR

This paper introduces a unified transformation approach for learned image compression based on modulation theory, interpreting LIC as a communication system and extending transformations to nonlinear modulation for improved performance.

Contribution

It proposes a novel unified transform method (TSM) for LIC, grounded in communication system principles, and demonstrates its effectiveness across various datasets and architectures.

Findings

Achieves 3.52% BD-rate reduction on Kodak dataset.

Validates the effectiveness and robustness of the proposed method.

Confirms the feasibility of designing LIC from a communication perspective.

Abstract

In this paper, a unified transformation method in learned image compression(LIC) is proposed from the perspective of modulation. Firstly, the quantization in LIC is considered as a generalized channel with additive uniform noise. Moreover, the LIC is interpreted as a particular communication system according to the consistency in structures and optimization objectives. Thus, the technology of communication systems can be applied to guide the design of modules in LIC. Furthermore, a unified transform method based on signal modulation (TSM) is defined. In the view of TSM, the existing transformation methods are mathematically reduced to a linear modulation. A series of transformation methods, e.g. TPM and TJM, are obtained by extending to nonlinear modulation. The experimental results on various datasets and backbone architectures verify that the effectiveness and robustness of the proposed method. More importantly, it further confirms the feasibility of guiding LIC design from a communication perspective. For example, when backbone architecture is hyperprior combining context model, our method achieves 3.52$\%$ BD-rate reduction over GDN on Kodak dataset without increasing complexity.