Learning sparsifying transforms for image processing

Saiprasad Ravishankar, Yoram Bresler

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

Abstract

The sparsity of signals and images in a certain analytically defined transform domain or dictionary such as discrete cosine transform or wavelets has been exploited in many applications in signal and image processing. Recently, the idea of learning a dictionary for sparse representation of data has become popular. However, while there has been extensive research on learning synthesis dictionaries, the idea of learning analysis sparsifying transforms has received only little attention. We propose a novel problem formulation and an alternating algorithm for learning well-conditioned square sparsifying transforms from data. We show the superiority of our approach for image representation over analytical sparsifying transforms such as the DCT. We also show promise in image denoising. Denoising using the learnt analysis transforms is not only better than by synthesis dictionaries learnt using the K-SVD algorithm but also faster.

Original languageEnglish (US)
Title of host publication2012 IEEE International Conference on Image Processing, ICIP 2012 - Proceedings
Pages681-684
Number of pages4
DOIs
StatePublished - Dec 1 2012
Event2012 19th IEEE International Conference on Image Processing, ICIP 2012 - Lake Buena Vista, FL, United States
Duration: Sep 30 2012Oct 3 2012

Publication series

NameProceedings - International Conference on Image Processing, ICIP
ISSN (Print)1522-4880

Other

Other2012 19th IEEE International Conference on Image Processing, ICIP 2012
CountryUnited States
CityLake Buena Vista, FL
Period9/30/1210/3/12

Fingerprint

Glossaries
Image processing
Image denoising
Discrete cosine transforms
Singular value decomposition
Signal processing

Keywords

  • Analysis transforms
  • Dictionary learning
  • Image denoising
  • Image representation
  • Sparse representation

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Information Systems

Cite this

Ravishankar, S., & Bresler, Y. (2012). Learning sparsifying transforms for image processing. In 2012 IEEE International Conference on Image Processing, ICIP 2012 - Proceedings (pp. 681-684). [6466951] (Proceedings - International Conference on Image Processing, ICIP). https://doi.org/10.1109/ICIP.2012.6466951

Learning sparsifying transforms for image processing. / Ravishankar, Saiprasad; Bresler, Yoram.

2012 IEEE International Conference on Image Processing, ICIP 2012 - Proceedings. 2012. p. 681-684 6466951 (Proceedings - International Conference on Image Processing, ICIP).

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

Ravishankar, S & Bresler, Y 2012, Learning sparsifying transforms for image processing. in 2012 IEEE International Conference on Image Processing, ICIP 2012 - Proceedings., 6466951, Proceedings - International Conference on Image Processing, ICIP, pp. 681-684, 2012 19th IEEE International Conference on Image Processing, ICIP 2012, Lake Buena Vista, FL, United States, 9/30/12. https://doi.org/10.1109/ICIP.2012.6466951
Ravishankar S, Bresler Y. Learning sparsifying transforms for image processing. In 2012 IEEE International Conference on Image Processing, ICIP 2012 - Proceedings. 2012. p. 681-684. 6466951. (Proceedings - International Conference on Image Processing, ICIP). https://doi.org/10.1109/ICIP.2012.6466951
Ravishankar, Saiprasad ; Bresler, Yoram. / Learning sparsifying transforms for image processing. 2012 IEEE International Conference on Image Processing, ICIP 2012 - Proceedings. 2012. pp. 681-684 (Proceedings - International Conference on Image Processing, ICIP).
@inproceedings{c70ceb8c943441d2ba835a31e4b645d4,
title = "Learning sparsifying transforms for image processing",
abstract = "The sparsity of signals and images in a certain analytically defined transform domain or dictionary such as discrete cosine transform or wavelets has been exploited in many applications in signal and image processing. Recently, the idea of learning a dictionary for sparse representation of data has become popular. However, while there has been extensive research on learning synthesis dictionaries, the idea of learning analysis sparsifying transforms has received only little attention. We propose a novel problem formulation and an alternating algorithm for learning well-conditioned square sparsifying transforms from data. We show the superiority of our approach for image representation over analytical sparsifying transforms such as the DCT. We also show promise in image denoising. Denoising using the learnt analysis transforms is not only better than by synthesis dictionaries learnt using the K-SVD algorithm but also faster.",
keywords = "Analysis transforms, Dictionary learning, Image denoising, Image representation, Sparse representation",
author = "Saiprasad Ravishankar and Yoram Bresler",
year = "2012",
month = "12",
day = "1",
doi = "10.1109/ICIP.2012.6466951",
language = "English (US)",
isbn = "9781467325332",
series = "Proceedings - International Conference on Image Processing, ICIP",
pages = "681--684",
booktitle = "2012 IEEE International Conference on Image Processing, ICIP 2012 - Proceedings",

}

TY - GEN

T1 - Learning sparsifying transforms for image processing

AU - Ravishankar, Saiprasad

AU - Bresler, Yoram

PY - 2012/12/1

Y1 - 2012/12/1

N2 - The sparsity of signals and images in a certain analytically defined transform domain or dictionary such as discrete cosine transform or wavelets has been exploited in many applications in signal and image processing. Recently, the idea of learning a dictionary for sparse representation of data has become popular. However, while there has been extensive research on learning synthesis dictionaries, the idea of learning analysis sparsifying transforms has received only little attention. We propose a novel problem formulation and an alternating algorithm for learning well-conditioned square sparsifying transforms from data. We show the superiority of our approach for image representation over analytical sparsifying transforms such as the DCT. We also show promise in image denoising. Denoising using the learnt analysis transforms is not only better than by synthesis dictionaries learnt using the K-SVD algorithm but also faster.

AB - The sparsity of signals and images in a certain analytically defined transform domain or dictionary such as discrete cosine transform or wavelets has been exploited in many applications in signal and image processing. Recently, the idea of learning a dictionary for sparse representation of data has become popular. However, while there has been extensive research on learning synthesis dictionaries, the idea of learning analysis sparsifying transforms has received only little attention. We propose a novel problem formulation and an alternating algorithm for learning well-conditioned square sparsifying transforms from data. We show the superiority of our approach for image representation over analytical sparsifying transforms such as the DCT. We also show promise in image denoising. Denoising using the learnt analysis transforms is not only better than by synthesis dictionaries learnt using the K-SVD algorithm but also faster.

KW - Analysis transforms

KW - Dictionary learning

KW - Image denoising

KW - Image representation

KW - Sparse representation

UR - http://www.scopus.com/inward/record.url?scp=84875864283&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84875864283&partnerID=8YFLogxK

U2 - 10.1109/ICIP.2012.6466951

DO - 10.1109/ICIP.2012.6466951

M3 - Conference contribution

AN - SCOPUS:84875864283

SN - 9781467325332

T3 - Proceedings - International Conference on Image Processing, ICIP

SP - 681

EP - 684

BT - 2012 IEEE International Conference on Image Processing, ICIP 2012 - Proceedings

ER -