JND-Based Perceptual Optimization For Learned Image Compression

TL;DR

This paper introduces a JND-based perceptual loss and a distortion-aware adjustor to improve the perceptual quality of learned image compression, aligning compression results more closely with human visual perception.

Contribution

It proposes a novel JND-based perceptual loss and a distortion-aware adjustor that enhance perceptual quality in learned image compression, with high scalability and plug-and-play capability.

Findings

Improved perceptual quality over baseline models at the same bit rate.

Effective integration of JND-based loss with existing compression schemes.

Demonstrated scalability across various learned image compression methods.

Abstract

Recently, learned image compression schemes have achieved remarkable improvements in image fidelity (e.g., PSNR and MS-SSIM) compared to conventional hybrid image coding ones due to their high-efficiency non-linear transform, end-to-end optimization frameworks, etc. However, few of them take the Just Noticeable Difference (JND) characteristic of the Human Visual System (HVS) into account and optimize learned image compression towards perceptual quality. To address this issue, a JND-based perceptual quality loss is proposed. Considering that the amounts of distortion in the compressed image at different training epochs under different Quantization Parameters (QPs) are different, we develop a distortion-aware adjustor. After combining them together, we can better assign the distortion in the compressed image with the guidance of JND to preserve the high perceptual quality. All these designs enable the proposed method to be flexibly applied to various learned image compression schemes with high scalability and plug-and-play advantages. Experimental results on the Kodak dataset demonstrate that the proposed method has led to better perceptual quality than the baseline model under the same bit rate.