CNN-based rate-distortion modeling for H.265/HEVC

Bin XU; Xiang PAN; Yan ZHOU; Yiming LI; Daiqin YANG; Zhenzhong CHEN

doi:10.1109/VCIP.2017.8305151

CNN-based rate-distortion modeling for H.265/HEVC

Bin XU, Xiang PAN, Yan ZHOU, Yiming LI, Daiqin YANG, Zhenzhong CHEN^*

^*Corresponding author for this work

Research output: Book Chapters | Papers in Conference Proceedings › Conference paper (refereed) › Research › peer-review

28 Citations (Scopus)

Abstract

In this paper, we propose a convolutional neural network (CNN)-based rate-distortion (R-D) modeling method for H.265/HEVC. A fully convolutional neural network (CNN) is designed to learn end-to-end, pixels-to-pixels mappings from the original images to the structural similarity (SSIM) maps indicating distortion. The rate information is predicted through a CNN with fully connected layers as well. When compared to traditional CNN methods, the proposed mappings to the distortion or rate information. The experiments demonstrate the feasibility of our CNN-based framework for rate-distortion modeling.

Original language	English
Title of host publication	IEEE Visual Communications and Image Processing, VCIP 2017
Publisher	IEEE
Pages	1-4
Number of pages	4
ISBN (Electronic)	9781538604625
ISBN (Print)	9781538604632
DOIs	https://doi.org/10.1109/VCIP.2017.8305151
Publication status	Published - Dec 2017
Externally published	Yes
Event	2017 IEEE Visual Communications and Image Processing (VCIP 2017) - St. Petersburg, United States Duration: 10 Dec 2017 → 13 Dec 2017

Conference

Conference	2017 IEEE Visual Communications and Image Processing (VCIP 2017)
Country/Territory	United States
City	St. Petersburg
Period	10/12/17 → 13/12/17

Bibliographical note

Publisher Copyright:
© 2017 IEEE.

Funding

This work was supported in part by National Natural Science Foundation of China (No. 61471273), National Hightech R&D Program of China (863 Program, 2015AA015903), Natural Science Foundation of Hubei Province of China (No. 2015CFA053) and LIESMARS Special Research Funding.

Keywords

convolutional neural network (CNN)
deep learning
H.265/HEVC
Rate-distortion model
structural similarity

Access to Document

10.1109/VCIP.2017.8305151

Cite this

@inproceedings{da226bf1e90e4d96a6510fe191d1ec93,

title = "CNN-based rate-distortion modeling for H.265/HEVC",

abstract = "In this paper, we propose a convolutional neural network (CNN)-based rate-distortion (R-D) modeling method for H.265/HEVC. A fully convolutional neural network (CNN) is designed to learn end-to-end, pixels-to-pixels mappings from the original images to the structural similarity (SSIM) maps indicating distortion. The rate information is predicted through a CNN with fully connected layers as well. When compared to traditional CNN methods, the proposed mappings to the distortion or rate information. The experiments demonstrate the feasibility of our CNN-based framework for rate-distortion modeling.",

keywords = "convolutional neural network (CNN), deep learning, H.265/HEVC, Rate-distortion model, structural similarity",

author = "Bin XU and Xiang PAN and Yan ZHOU and Yiming LI and Daiqin YANG and Zhenzhong CHEN",

note = "Publisher Copyright: {\textcopyright} 2017 IEEE.; 2017 IEEE Visual Communications and Image Processing (VCIP 2017) ; Conference date: 10-12-2017 Through 13-12-2017",

year = "2017",

month = dec,

doi = "10.1109/VCIP.2017.8305151",

language = "English",

isbn = "9781538604632",

pages = "1--4",

booktitle = "IEEE Visual Communications and Image Processing, VCIP 2017",

publisher = "IEEE",

}

TY - GEN

T1 - CNN-based rate-distortion modeling for H.265/HEVC

AU - XU, Bin

AU - PAN, Xiang

AU - ZHOU, Yan

AU - LI, Yiming

AU - YANG, Daiqin

AU - CHEN, Zhenzhong

PY - 2017/12

Y1 - 2017/12

N2 - In this paper, we propose a convolutional neural network (CNN)-based rate-distortion (R-D) modeling method for H.265/HEVC. A fully convolutional neural network (CNN) is designed to learn end-to-end, pixels-to-pixels mappings from the original images to the structural similarity (SSIM) maps indicating distortion. The rate information is predicted through a CNN with fully connected layers as well. When compared to traditional CNN methods, the proposed mappings to the distortion or rate information. The experiments demonstrate the feasibility of our CNN-based framework for rate-distortion modeling.

AB - In this paper, we propose a convolutional neural network (CNN)-based rate-distortion (R-D) modeling method for H.265/HEVC. A fully convolutional neural network (CNN) is designed to learn end-to-end, pixels-to-pixels mappings from the original images to the structural similarity (SSIM) maps indicating distortion. The rate information is predicted through a CNN with fully connected layers as well. When compared to traditional CNN methods, the proposed mappings to the distortion or rate information. The experiments demonstrate the feasibility of our CNN-based framework for rate-distortion modeling.

KW - convolutional neural network (CNN)

KW - deep learning

KW - H.265/HEVC

KW - Rate-distortion model

KW - structural similarity

UR - https://www.scopus.com/pages/publications/85050623983

U2 - 10.1109/VCIP.2017.8305151

DO - 10.1109/VCIP.2017.8305151

M3 - Conference paper (refereed)

AN - SCOPUS:85050623983

SN - 9781538604632

SP - 1

EP - 4

BT - IEEE Visual Communications and Image Processing, VCIP 2017

PB - IEEE

T2 - 2017 IEEE Visual Communications and Image Processing (VCIP 2017)

Y2 - 10 December 2017 through 13 December 2017

ER -

CNN-based rate-distortion modeling for H.265/HEVC

Abstract

Conference

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this