End-to-End Learned Image Compression with Quantized Weights and Activations

TL;DR

This paper presents a method to significantly reduce the complexity of learned image compression networks by quantizing weights and activations, achieving comparable or better performance than traditional codecs like BPG while enabling hardware implementation.

Contribution

It introduces a comprehensive quantization scheme for LIC, including weight and activation quantization with fine-tuning, and demonstrates substantial memory reduction with minimal performance loss.

Findings

Memory cost reduced by 75% with negligible performance loss

Quantized LIC outperforms BPG in MS-SSIM

First complete analysis of coding gain and memory cost for quantized LIC

Abstract

End-to-end Learned image compression (LIC) has reached the traditional hand-crafted methods such as BPG (HEVC intra) in terms of the coding gain. However, the large network size prohibits the usage of LIC on resource-limited embedded systems. This paper reduces the network complexity by quantizing both weights and activations. 1) For the weight quantization, we study different kinds of grouping and quantization scheme at first. A channel-wise non-linear quantization scheme is determined based on the coding gain analysis. After that, we propose a fine tuning scheme to clip the weights within a certain range so that the quantization error can be reduced. 2) For the activation quantization, we first propose multiple non-linear quantization codebooks with different maximum dynamic ranges. By selecting an optimal one through a multiplexer, the quantization range can be saturated to the greatest extent. In addition, we also exploit the mean-removed quantization for the analysis transform outputs in order to reduce the bit-width cost for the specific channel with the large non-zero mean. By quantizing each weight and activation element from 32-bit floating point to 8-bit fixed point, the memory cost for both weight and activation can be reduced by 75% with negligible coding performance loss. As a result, our quantized LIC can still outperform BPG in terms of MS-SSIM. To our best knowledge, this is the first work to give a complete analysis on the coding gain and the memory cost for a quantized LIC network, which validates the feasibility of the hardware implementation.