Universal Efficient Variable-rate Neural Image Compression

TL;DR

This paper introduces universal modules Energy-based Channel Gating and Bit-rate Modulator that enable existing neural image compression models to achieve arbitrary bit-rates and reduce computational complexity, enhancing practicality.

Contribution

The paper presents two novel modules that can be integrated into existing models to control bit-rate and reduce computation without retraining.

Findings

Over 50% FLOPs reduction in convolution layers.

Models can output arbitrary bit-rates with a single trained model.

Enhanced computational efficiency and rate flexibility.

Abstract

Recently, Learning-based image compression has reached comparable performance with traditional image codecs(such as JPEG, BPG, WebP). However, computational complexity and rate flexibility are still two major challenges for its practical deployment. To tackle these problems, this paper proposes two universal modules named Energy-based Channel Gating(ECG) and Bit-rate Modulator(BM), which can be directly embedded into existing end-to-end image compression models. ECG uses dynamic pruning to reduce FLOPs for more than 50\% in convolution layers, and a BM pair can modulate the latent representation to control the bit-rate in a channel-wise manner. By implementing these two modules, existing learning-based image codecs can obtain ability to output arbitrary bit-rate with a single model and reduced computation.