Learned Image Compression With Separate Hyperprior Decoders

Learned image compression techniques have achieved considerable development in recent years. In this paper, we find that the performance bottleneck lies in the use of a single hyperprior decoder, in which case the ternary Gaussian model collapses to a binary one. To solve this, we propose to use thr...

Full description

Saved in:

Bibliographic Details
Main Authors:	Zhao Zan, Chao Liu, Heming Sun, Xiaoyang Zeng, Yibo Fan
Format:	article
Language:	EN
Published:	IEEE 2021
Subjects:	Learned image compression variational autoencoder convolutional neural networks Gaussian mixture model Electric apparatus and materials. Electric circuits. Electric networks TK452-454.4
Online Access:	https://doaj.org/article/3d5edee77bea48bc9984699c72bba75b
Tags:	Add Tag No Tags, Be the first to tag this record!

id	oai:doaj.org-article:3d5edee77bea48bc9984699c72bba75b
record_format	dspace
spelling	oai:doaj.org-article:3d5edee77bea48bc9984699c72bba75b2021-11-23T00:02:18ZLearned Image Compression With Separate Hyperprior Decoders2644-122510.1109/OJCAS.2021.3125354https://doaj.org/article/3d5edee77bea48bc9984699c72bba75b2021-01-01T00:00:00Zhttps://ieeexplore.ieee.org/document/9623333/https://doaj.org/toc/2644-1225Learned image compression techniques have achieved considerable development in recent years. In this paper, we find that the performance bottleneck lies in the use of a single hyperprior decoder, in which case the ternary Gaussian model collapses to a binary one. To solve this, we propose to use three hyperprior decoders to separate the decoding process of the mixed parameters in discrete Gaussian mixture likelihoods, achieving more accurate parameters estimation. Experimental results demonstrate the proposed method optimized by MS-SSIM achieves on average 3.36% BD-rate reduction compared with state-of-the-art approach. The contribution of the proposed method to the coding time and FLOPs is negligible.Zhao ZanChao LiuHeming SunXiaoyang ZengYibo FanIEEEarticleLearned image compressionvariational autoencoderconvolutional neural networksGaussian mixture modelElectric apparatus and materials. Electric circuits. Electric networksTK452-454.4ENIEEE Open Journal of Circuits and Systems, Vol 2, Pp 627-632 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	Learned image compression variational autoencoder convolutional neural networks Gaussian mixture model Electric apparatus and materials. Electric circuits. Electric networks TK452-454.4
spellingShingle	Learned image compression variational autoencoder convolutional neural networks Gaussian mixture model Electric apparatus and materials. Electric circuits. Electric networks TK452-454.4 Zhao Zan Chao Liu Heming Sun Xiaoyang Zeng Yibo Fan Learned Image Compression With Separate Hyperprior Decoders
description	Learned image compression techniques have achieved considerable development in recent years. In this paper, we find that the performance bottleneck lies in the use of a single hyperprior decoder, in which case the ternary Gaussian model collapses to a binary one. To solve this, we propose to use three hyperprior decoders to separate the decoding process of the mixed parameters in discrete Gaussian mixture likelihoods, achieving more accurate parameters estimation. Experimental results demonstrate the proposed method optimized by MS-SSIM achieves on average 3.36% BD-rate reduction compared with state-of-the-art approach. The contribution of the proposed method to the coding time and FLOPs is negligible.
format	article
author	Zhao Zan Chao Liu Heming Sun Xiaoyang Zeng Yibo Fan
author_facet	Zhao Zan Chao Liu Heming Sun Xiaoyang Zeng Yibo Fan
author_sort	Zhao Zan
title	Learned Image Compression With Separate Hyperprior Decoders
title_short	Learned Image Compression With Separate Hyperprior Decoders
title_full	Learned Image Compression With Separate Hyperprior Decoders
title_fullStr	Learned Image Compression With Separate Hyperprior Decoders
title_full_unstemmed	Learned Image Compression With Separate Hyperprior Decoders
title_sort	learned image compression with separate hyperprior decoders
publisher	IEEE
publishDate	2021
url	https://doaj.org/article/3d5edee77bea48bc9984699c72bba75b
work_keys_str_mv	AT zhaozan learnedimagecompressionwithseparatehyperpriordecoders AT chaoliu learnedimagecompressionwithseparatehyperpriordecoders AT hemingsun learnedimagecompressionwithseparatehyperpriordecoders AT xiaoyangzeng learnedimagecompressionwithseparatehyperpriordecoders AT yibofan learnedimagecompressionwithseparatehyperpriordecoders
_version_	1718417370076151808

Learned Image Compression With Separate Hyperprior Decoders

Similar Items