Dictionary learning for scalable sparse image representation with applications

Research output: Contribution to journalArticle

Abstract

This paper introduces a novel design for the dictionary learning algorithm, intended for scalable sparse representation of high motion video sequences and natural images. The proposed algorithm is built upon the foundation of the K-SVD framework originally designed to learn non-scalable dictionaries for natural images. Proposed design is mainly motivated by the main perception characteristic of the Human Visual System (HVS) mechanism. Specifically, its core structure relies on the exploitation of the high-frequency image components and contrast variations in order to achieve visual scene objects identification at all scalable levels. Proposed design is implemented by introducing a semi-random Morphological Component Analysis (MCA) based initialization of the K-SVD dictionary and the regularization of its atom’s update mechanism. In general, dictionary learning for sparse representations leads to state-of-the-art image restoration results for several different problems in the field of image processing. In experimental section we show that these are equally achievable by accommodating all dictionary elements to tailor the scalable data representation and reconstruction, hence modeling data that admit sparse representation in a novel manner. Performed simulations include scalable sparse recovery for representation of static and dynamic data changing over time (e.g., video) together with application to denoising and compressive sensing.
LanguageEnglish
Pages55-74
Number of pages20
JournalAdvances in Signal Processing
Volume2
Issue number2
DOIs
Publication statusPublished - 5 Jan 2014

Fingerprint

Glossaries
Singular value decomposition
Image reconstruction
Learning algorithms
Data structures
Image processing
Recovery
Atoms

Keywords

  • scalable video representation
  • sparse coding
  • contrast variation
  • denoising
  • compressive sensing

Cite this

@article{55f6ddbe167a4c13a2788a3eb3ee29dd,
title = "Dictionary learning for scalable sparse image representation with applications",
abstract = "This paper introduces a novel design for the dictionary learning algorithm, intended for scalable sparse representation of high motion video sequences and natural images. The proposed algorithm is built upon the foundation of the K-SVD framework originally designed to learn non-scalable dictionaries for natural images. Proposed design is mainly motivated by the main perception characteristic of the Human Visual System (HVS) mechanism. Specifically, its core structure relies on the exploitation of the high-frequency image components and contrast variations in order to achieve visual scene objects identification at all scalable levels. Proposed design is implemented by introducing a semi-random Morphological Component Analysis (MCA) based initialization of the K-SVD dictionary and the regularization of its atom’s update mechanism. In general, dictionary learning for sparse representations leads to state-of-the-art image restoration results for several different problems in the field of image processing. In experimental section we show that these are equally achievable by accommodating all dictionary elements to tailor the scalable data representation and reconstruction, hence modeling data that admit sparse representation in a novel manner. Performed simulations include scalable sparse recovery for representation of static and dynamic data changing over time (e.g., video) together with application to denoising and compressive sensing.",
keywords = "scalable video representation, sparse coding, contrast variation, denoising, compressive sensing",
author = "Bojana Begovic and Vladimir Stankovic and Lina Stankovic",
note = "DOI is as given but does not work",
year = "2014",
month = "1",
day = "5",
doi = "10.13189/asp.2014.020204",
language = "English",
volume = "2",
pages = "55--74",
journal = "Advances in Signal Processing",
issn = "2332-6883",
number = "2",

}

Dictionary learning for scalable sparse image representation with applications. / Begovic, Bojana; Stankovic, Vladimir; Stankovic, Lina.

In: Advances in Signal Processing, Vol. 2, No. 2, 05.01.2014, p. 55-74.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Dictionary learning for scalable sparse image representation with applications

AU - Begovic, Bojana

AU - Stankovic, Vladimir

AU - Stankovic, Lina

N1 - DOI is as given but does not work

PY - 2014/1/5

Y1 - 2014/1/5

N2 - This paper introduces a novel design for the dictionary learning algorithm, intended for scalable sparse representation of high motion video sequences and natural images. The proposed algorithm is built upon the foundation of the K-SVD framework originally designed to learn non-scalable dictionaries for natural images. Proposed design is mainly motivated by the main perception characteristic of the Human Visual System (HVS) mechanism. Specifically, its core structure relies on the exploitation of the high-frequency image components and contrast variations in order to achieve visual scene objects identification at all scalable levels. Proposed design is implemented by introducing a semi-random Morphological Component Analysis (MCA) based initialization of the K-SVD dictionary and the regularization of its atom’s update mechanism. In general, dictionary learning for sparse representations leads to state-of-the-art image restoration results for several different problems in the field of image processing. In experimental section we show that these are equally achievable by accommodating all dictionary elements to tailor the scalable data representation and reconstruction, hence modeling data that admit sparse representation in a novel manner. Performed simulations include scalable sparse recovery for representation of static and dynamic data changing over time (e.g., video) together with application to denoising and compressive sensing.

AB - This paper introduces a novel design for the dictionary learning algorithm, intended for scalable sparse representation of high motion video sequences and natural images. The proposed algorithm is built upon the foundation of the K-SVD framework originally designed to learn non-scalable dictionaries for natural images. Proposed design is mainly motivated by the main perception characteristic of the Human Visual System (HVS) mechanism. Specifically, its core structure relies on the exploitation of the high-frequency image components and contrast variations in order to achieve visual scene objects identification at all scalable levels. Proposed design is implemented by introducing a semi-random Morphological Component Analysis (MCA) based initialization of the K-SVD dictionary and the regularization of its atom’s update mechanism. In general, dictionary learning for sparse representations leads to state-of-the-art image restoration results for several different problems in the field of image processing. In experimental section we show that these are equally achievable by accommodating all dictionary elements to tailor the scalable data representation and reconstruction, hence modeling data that admit sparse representation in a novel manner. Performed simulations include scalable sparse recovery for representation of static and dynamic data changing over time (e.g., video) together with application to denoising and compressive sensing.

KW - scalable video representation

KW - sparse coding

KW - contrast variation

KW - denoising

KW - compressive sensing

UR - http://www.hrpub.org/download/20140105/ASP4-19101837.pdf

U2 - 10.13189/asp.2014.020204

DO - 10.13189/asp.2014.020204

M3 - Article

VL - 2

SP - 55

EP - 74

JO - Advances in Signal Processing

T2 - Advances in Signal Processing

JF - Advances in Signal Processing

SN - 2332-6883

IS - 2

ER -