Improved Lossy Image Compression with Priming and Spatially Adaptive Bit Rates for Recurrent Networks

Abstract

We propose a method for lossy image compression based on recurrent, convolutional neural networks that outperforms BPG (4:2:0), WebP, JPEG2000, and JPEG as measured by MS-SSIM. We introduce three improvements over previous research that lead to this state-of-the-art result. First, we show that training with a pixel-wise loss weighted by SSIM increases reconstruction quality according to several metrics. Second, we modify the recurrent architecture to improve spatial diffusion, which allows the network to more effectively capture and propagate image information through the network's hidden state. Finally, in addition to lossless entropy coding, we use a spatially adaptive bit allocation algorithm to more efficiently use the limited number of bits to encode visually complex image regions. We evaluate our method on the Kodak and Tecnick image sets and compare against standard codecs as well recently published methods based on deep neural networks.

Learning Image and Video Compression Through Spatial-Temporal Energy Compaction

“…Learning Compression Recently, end-to-end image compression has attracted great attention. Some approaches proposed to use recurrent neural networks (RNNs) to encode the residual information between the raw image and the reconstructed images in several iterations, such as the work [8,9] optimized by mean-squared error (MSE) or the work [10] optimized by MS-SSIM [30]. Some generative adversarial networks (GANs) based techniques are proposed in [18,19,20] for high subjective reconstruction quality at extremely low bit rates.…”

“…Although autoencoders are basically applied to dimensionality reduction tasks [5], they are able to achieve better compression performance. Most recent learning based compression approaches, including recurrent neural networks [8,9,10], convolutional neural networks [11,12,13,14,15,16,17] and generative adversarial networks [18,19,20], have all adopted the autoencoder architecture. Next, the temporal redundancy of video compression can be intuitively reduced by using learning based video prediction, generation and interpolation approaches.…”

Learned Lossless Image Compression with A Hyperprior and Discretized Gaussian Mixture Likelihoods

“…For example, some image compression approaches use generative models to learn the distribution of images using adversarial training [6,7,8] to achieved impressive subjective quality at extremely low bit rate. Some works use recurrent neural networks to compress the residual information recursively, such as [9,10,11] to realize scalable coding. Some approaches propose a hyperpriorbased and context-adaptive context model to compress codes effectively in [12,13,14].…”

Improved hybrid layered image compression using deep learning and traditional codecs

“…Experimental results using the Kodak and Tecnick datasets show that the proposed scheme outperforms the state-of-the-art deep learning-based layered coding scheme and traditional codecs including BPG in both PSNR and MS-SSIM metrics across a wide range of bit rates, when the images are coded in the RGB444 domain.Keywords deep learning-based image coding · layered image coding · residual coding · convolutional neural network · autoencoderRecently with the rapid development of deep learning theory, especially after the successful applications of convolutional neural networks (CNN) in computer vision, deep learning has been applied to many areas, including image compression. Some deep learning-based methods have outperformed traditional image codecs such as JPEG, JPEG2000, and the H.265/HEVC-based BPG image codec [13,1,2,6,8,11,12,21], demonstrating its great potentials. In [17], Toderici et al proposed a variable-rate image compression framework for thumbnail images by using the recurrent neural arXiv:1907.06566v1 [eess.IV]…”

“…Recently with the rapid development of deep learning theory, especially after the successful applications of convolutional neural networks (CNN) in computer vision, deep learning has been applied to many areas, including image compression. Some deep learning-based methods have outperformed traditional image codecs such as JPEG, JPEG2000, and the H.265/HEVC-based BPG image codec [13,1,2,6,8,11,12,21], demonstrating its great potentials. In [17], Toderici et al proposed a variable-rate image compression framework for thumbnail images by using the recurrent neural arXiv:1907.06566v1 [eess.IV]…”