Full Resolution Image Compression with Recurrent Neural Networks
George Toderici Google Research Damien Vincent Nick Johnston Sung Jin Hwang David Minnen Joel Shor Michele Covell
Abstract
This paper presents a set of full-resolution lossy image compression methods based on neural networks. Each of the architectures we describe can provide variable compression rates during deployment without requiring retraining of the network: each network need only be trained once. All of our architectures consist of a recurrent neural network (RNN)-based encoder and decoder, a binarizer, and a neural network for entropy coding. We compare RNN types (LSTM, associative LSTM) and introduce a new hybrid of GRU and ResNet. We also study “one-shot” versus additive reconstruction architectures and introduce a new scaled-additive framework. We compare to previous work, showing improvements of 4.3%–8.8% AUC (area under the rate-distortion curve), depending on the perceptual metric used. As far as we know, this is the first neural network architecture that is able to outperform JPEG at image compression across most bitrates on the rate-distortion curve on the Kodak dataset images, with and without the aid of entropy coding.
中文速览
用循环神经网络（RNN）做有损图像压缩时，现有方法要么只能处理小尺寸图片，要么无法在不重新训练的前提下灵活调整压缩率，谷歌这篇工作同时解决了这两个问题。研究者设计了一套基于 RNN 的编解码框架，引入了 LSTM、Associative LSTM 以及他们自创的 GRU+ResNet 混合单元，并提出"缩放累加重建"（scaled-additive reconstruction）机制让网络能根据图像内容自适应地处理不同迭代阶段的残差；在此之上，他们还配套设计了基于 PixelRNN 思路的熵编码模块，利用图像块之间的长程依赖进一步压缩比特数。在标准 Kodak 数据集上，该方法在率失真曲线（rate-distortion curve）的绝大多数码率点上首次让神经网络超越了 JPEG，AUC 提升幅度达 4.3%–8.8%（视评估指标而定）。这项工作的意义在于，它证明了单个神经网络无需针对不同码率分别训练，就能在全分辨率自然图像上全面超越沿用数十年的经典编解码标准。
原文 arXiv:1608.05148;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1608.05148v2