End-To-End Learned Image Compression With Fixed Point Weight Quantization

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 nonlinear 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.

NNCD-IQA: A new neural networks based compressed database for image quality assessment

“…The above architecture [4] has been then considered as a reference model in many other deep learning based image compression algorithms [5,8,25,47]. Among them, we retain here the method proposed in [5].…”

Q-LIC: Quantizing Learned Image Compression with Channel Splitting

“…[37] developed a very heuristic method to train an integer LIC from scratch. [38] proposed a weight clipping method to reduce the weight quantization error, and its advanced version with a layer-by-layer weight fine tuning was presented in [39]. [40] proposed a range preprocessing to bound the dynamic range, and then performed a range-adaptive quantization.…”

“…[40] proposed a range preprocessing to bound the dynamic range, and then performed a range-adaptive quantization. For [38], [39], though weight can be quantized with small coding loss, the activation in the main path is not taken into consideration for the quantization. For [37], [40], both weights and activations in the main and hyper path are quantized.…”