Neighbor embedding based single image super-resolution using hybrid feature and adaptive weight decay regularization

Yonggun Lee, Yoonsik Choe

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In this paper, a novel single image super-resolution technique based on neighbor embedding is proposed. Conventional neighbor embedding technique uses gradient feature vector in neighbor selection. However, since gradient feature is high-frequency feature, it is not appropriate for feature representation in weak-edge and texture regions. Therefore, instead of using only gradient feature as in neighbor embedding, hybrid feature which combines gradient feature for edge region and luminance norm feature for non-edge region is used. Also, adaptive weight decay regularization is applied to shrink unstable weights for neighbors. Consequently, experimental result shows that the proposed method can reconstruct more details in weak-edges, textures, and strong-edges as well as conventional method.

Original languageEnglish
Title of host publicationProceedings 2014 IEEE 4th International Conference on Consumer Electronics - Berlin, ICCE-Berlin
EditorsFrancisco J. Bellido, Dietmar Hepper, Hans L. Cycon, Alexander Huhn
PublisherIEEE Computer Society
Pages185-187
Number of pages3
EditionFebruary
ISBN (Electronic)9781479961658
DOIs
Publication statusPublished - 2015 Feb 5
Event2014 4th IEEE International Conference on Consumer Electronics - Berlin, ICCE-Berlin - Berlin, Germany
Duration: 2014 Sep 72014 Sep 10

Publication series

NameIEEE International Conference on Consumer Electronics - Berlin, ICCE-Berlin
NumberFebruary
Volume2015-February
ISSN (Print)2166-6814
ISSN (Electronic)2166-6822

Conference

Conference2014 4th IEEE International Conference on Consumer Electronics - Berlin, ICCE-Berlin
CountryGermany
CityBerlin
Period14/9/714/9/10

Fingerprint

Textures
Luminance

All Science Journal Classification (ASJC) codes

  • Electrical and Electronic Engineering
  • Industrial and Manufacturing Engineering
  • Media Technology

Cite this

Lee, Y., & Choe, Y. (2015). Neighbor embedding based single image super-resolution using hybrid feature and adaptive weight decay regularization. In F. J. Bellido, D. Hepper, H. L. Cycon, & A. Huhn (Eds.), Proceedings 2014 IEEE 4th International Conference on Consumer Electronics - Berlin, ICCE-Berlin (February ed., pp. 185-187). [7034223] (IEEE International Conference on Consumer Electronics - Berlin, ICCE-Berlin; Vol. 2015-February, No. February). IEEE Computer Society. https://doi.org/10.1109/ICCE-Berlin.2014.7034223
Lee, Yonggun ; Choe, Yoonsik. / Neighbor embedding based single image super-resolution using hybrid feature and adaptive weight decay regularization. Proceedings 2014 IEEE 4th International Conference on Consumer Electronics - Berlin, ICCE-Berlin. editor / Francisco J. Bellido ; Dietmar Hepper ; Hans L. Cycon ; Alexander Huhn. February. ed. IEEE Computer Society, 2015. pp. 185-187 (IEEE International Conference on Consumer Electronics - Berlin, ICCE-Berlin; February).
@inproceedings{302f5aac3e1e4fb5934b4ab967dc5234,
title = "Neighbor embedding based single image super-resolution using hybrid feature and adaptive weight decay regularization",
abstract = "In this paper, a novel single image super-resolution technique based on neighbor embedding is proposed. Conventional neighbor embedding technique uses gradient feature vector in neighbor selection. However, since gradient feature is high-frequency feature, it is not appropriate for feature representation in weak-edge and texture regions. Therefore, instead of using only gradient feature as in neighbor embedding, hybrid feature which combines gradient feature for edge region and luminance norm feature for non-edge region is used. Also, adaptive weight decay regularization is applied to shrink unstable weights for neighbors. Consequently, experimental result shows that the proposed method can reconstruct more details in weak-edges, textures, and strong-edges as well as conventional method.",
author = "Yonggun Lee and Yoonsik Choe",
year = "2015",
month = "2",
day = "5",
doi = "10.1109/ICCE-Berlin.2014.7034223",
language = "English",
series = "IEEE International Conference on Consumer Electronics - Berlin, ICCE-Berlin",
publisher = "IEEE Computer Society",
number = "February",
pages = "185--187",
editor = "Bellido, {Francisco J.} and Dietmar Hepper and Cycon, {Hans L.} and Alexander Huhn",
booktitle = "Proceedings 2014 IEEE 4th International Conference on Consumer Electronics - Berlin, ICCE-Berlin",
address = "United States",
edition = "February",

}

Lee, Y & Choe, Y 2015, Neighbor embedding based single image super-resolution using hybrid feature and adaptive weight decay regularization. in FJ Bellido, D Hepper, HL Cycon & A Huhn (eds), Proceedings 2014 IEEE 4th International Conference on Consumer Electronics - Berlin, ICCE-Berlin. February edn, 7034223, IEEE International Conference on Consumer Electronics - Berlin, ICCE-Berlin, no. February, vol. 2015-February, IEEE Computer Society, pp. 185-187, 2014 4th IEEE International Conference on Consumer Electronics - Berlin, ICCE-Berlin, Berlin, Germany, 14/9/7. https://doi.org/10.1109/ICCE-Berlin.2014.7034223

Neighbor embedding based single image super-resolution using hybrid feature and adaptive weight decay regularization. / Lee, Yonggun; Choe, Yoonsik.

Proceedings 2014 IEEE 4th International Conference on Consumer Electronics - Berlin, ICCE-Berlin. ed. / Francisco J. Bellido; Dietmar Hepper; Hans L. Cycon; Alexander Huhn. February. ed. IEEE Computer Society, 2015. p. 185-187 7034223 (IEEE International Conference on Consumer Electronics - Berlin, ICCE-Berlin; Vol. 2015-February, No. February).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Neighbor embedding based single image super-resolution using hybrid feature and adaptive weight decay regularization

AU - Lee, Yonggun

AU - Choe, Yoonsik

PY - 2015/2/5

Y1 - 2015/2/5

N2 - In this paper, a novel single image super-resolution technique based on neighbor embedding is proposed. Conventional neighbor embedding technique uses gradient feature vector in neighbor selection. However, since gradient feature is high-frequency feature, it is not appropriate for feature representation in weak-edge and texture regions. Therefore, instead of using only gradient feature as in neighbor embedding, hybrid feature which combines gradient feature for edge region and luminance norm feature for non-edge region is used. Also, adaptive weight decay regularization is applied to shrink unstable weights for neighbors. Consequently, experimental result shows that the proposed method can reconstruct more details in weak-edges, textures, and strong-edges as well as conventional method.

AB - In this paper, a novel single image super-resolution technique based on neighbor embedding is proposed. Conventional neighbor embedding technique uses gradient feature vector in neighbor selection. However, since gradient feature is high-frequency feature, it is not appropriate for feature representation in weak-edge and texture regions. Therefore, instead of using only gradient feature as in neighbor embedding, hybrid feature which combines gradient feature for edge region and luminance norm feature for non-edge region is used. Also, adaptive weight decay regularization is applied to shrink unstable weights for neighbors. Consequently, experimental result shows that the proposed method can reconstruct more details in weak-edges, textures, and strong-edges as well as conventional method.

UR - http://www.scopus.com/inward/record.url?scp=84937896734&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84937896734&partnerID=8YFLogxK

U2 - 10.1109/ICCE-Berlin.2014.7034223

DO - 10.1109/ICCE-Berlin.2014.7034223

M3 - Conference contribution

AN - SCOPUS:84937896734

T3 - IEEE International Conference on Consumer Electronics - Berlin, ICCE-Berlin

SP - 185

EP - 187

BT - Proceedings 2014 IEEE 4th International Conference on Consumer Electronics - Berlin, ICCE-Berlin

A2 - Bellido, Francisco J.

A2 - Hepper, Dietmar

A2 - Cycon, Hans L.

A2 - Huhn, Alexander

PB - IEEE Computer Society

ER -

Lee Y, Choe Y. Neighbor embedding based single image super-resolution using hybrid feature and adaptive weight decay regularization. In Bellido FJ, Hepper D, Cycon HL, Huhn A, editors, Proceedings 2014 IEEE 4th International Conference on Consumer Electronics - Berlin, ICCE-Berlin. February ed. IEEE Computer Society. 2015. p. 185-187. 7034223. (IEEE International Conference on Consumer Electronics - Berlin, ICCE-Berlin; February). https://doi.org/10.1109/ICCE-Berlin.2014.7034223