Non-strictly Sparse Source Modelling Using Heavy-tailed Distributions for DCT Coefficients

Masoud Aramideh; Ehsan Namjoo; Mahdi Nooshyar

doi:10.1109/IranianCEE.2019.8786650

Non-strictly Sparse Source Modelling Using Heavy-tailed Distributions for DCT Coefficients

Masoud Aramideh
, Ehsan Namjoo
, Mahdi Nooshyar

Faculty of Science & Engineering

Shahid Chamran University of Ahvaz

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

In this study heavy-tailed distributions including Pareto, Weibull, Levy, and Log-normal distributions are proposed to model the DCT coefficients of real-world sources. It will be shown that the proposed heavy-tailed distributions are more accurate to model non-strictly sparse sources rather than Laplace and Cauchy distributions. Maximum Likelihood Estimation (MLE) method is utilized to parameter estimation of the proposed distributions. Finally, the Shannon Lower Bound (SLB) on the rate distortion function of each model is calculated, accordingly a framework is provided that could be used to assess the rate distortion performance of any coding scheme that is supposed to compress DCT coefficients as outcome of a sparse source.

Original language	English
Title of host publication	ICEE 2019 - 27th Iranian Conference on Electrical Engineering
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	1570-1575
Number of pages	6
ISBN (Electronic)	9781728115085
DOIs	https://doi.org/10.1109/IranianCEE.2019.8786650
Publication status	Published - Apr 2019
Event	27th Iranian Conference on Electrical Engineering, ICEE 2019 - Yazd, Iran, Islamic Republic of Duration: 30 Apr 2019 → 2 May 2019

Publication series

Name	ICEE 2019 - 27th Iranian Conference on Electrical Engineering

Conference

Conference	27th Iranian Conference on Electrical Engineering, ICEE 2019
Country/Territory	Iran, Islamic Republic of
City	Yazd
Period	30/04/19 → 2/05/19

Keywords

heavy-tailed distribution
Levy distribution
Log-normal distribution
lossy source coding
maximum likelihood estimation
Pareto distribution
sparse sources
Weibull distribution

Access to Document

10.1109/IranianCEE.2019.8786650

Cite this

Aramideh, M., Namjoo, E., & Nooshyar, M. (2019). Non-strictly Sparse Source Modelling Using Heavy-tailed Distributions for DCT Coefficients. In ICEE 2019 - 27th Iranian Conference on Electrical Engineering (pp. 1570-1575). Article 8786650 (ICEE 2019 - 27th Iranian Conference on Electrical Engineering). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/IranianCEE.2019.8786650

@inproceedings{bc3867ad1c5844f581ca222c71e06675,

title = "Non-strictly Sparse Source Modelling Using Heavy-tailed Distributions for DCT Coefficients",

abstract = "In this study heavy-tailed distributions including Pareto, Weibull, Levy, and Log-normal distributions are proposed to model the DCT coefficients of real-world sources. It will be shown that the proposed heavy-tailed distributions are more accurate to model non-strictly sparse sources rather than Laplace and Cauchy distributions. Maximum Likelihood Estimation (MLE) method is utilized to parameter estimation of the proposed distributions. Finally, the Shannon Lower Bound (SLB) on the rate distortion function of each model is calculated, accordingly a framework is provided that could be used to assess the rate distortion performance of any coding scheme that is supposed to compress DCT coefficients as outcome of a sparse source.",

keywords = "heavy-tailed distribution, Levy distribution, Log-normal distribution, lossy source coding, maximum likelihood estimation, Pareto distribution, sparse sources, Weibull distribution",

author = "Masoud Aramideh and Ehsan Namjoo and Mahdi Nooshyar",

note = "Publisher Copyright: {\textcopyright} 2019 IEEE.; 27th Iranian Conference on Electrical Engineering, ICEE 2019 ; Conference date: 30-04-2019 Through 02-05-2019",

year = "2019",

month = apr,

doi = "10.1109/IranianCEE.2019.8786650",

language = "English",

series = "ICEE 2019 - 27th Iranian Conference on Electrical Engineering",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "1570--1575",

booktitle = "ICEE 2019 - 27th Iranian Conference on Electrical Engineering",

}

Aramideh, M, Namjoo, E & Nooshyar, M 2019, Non-strictly Sparse Source Modelling Using Heavy-tailed Distributions for DCT Coefficients. in ICEE 2019 - 27th Iranian Conference on Electrical Engineering., 8786650, ICEE 2019 - 27th Iranian Conference on Electrical Engineering, Institute of Electrical and Electronics Engineers Inc., pp. 1570-1575, 27th Iranian Conference on Electrical Engineering, ICEE 2019, Yazd, Iran, Islamic Republic of, 30/04/19. https://doi.org/10.1109/IranianCEE.2019.8786650

Non-strictly Sparse Source Modelling Using Heavy-tailed Distributions for DCT Coefficients. / Aramideh, Masoud; Namjoo, Ehsan; Nooshyar, Mahdi.
ICEE 2019 - 27th Iranian Conference on Electrical Engineering. Institute of Electrical and Electronics Engineers Inc., 2019. p. 1570-1575 8786650 (ICEE 2019 - 27th Iranian Conference on Electrical Engineering).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Non-strictly Sparse Source Modelling Using Heavy-tailed Distributions for DCT Coefficients

AU - Aramideh, Masoud

AU - Namjoo, Ehsan

AU - Nooshyar, Mahdi

PY - 2019/4

Y1 - 2019/4

N2 - In this study heavy-tailed distributions including Pareto, Weibull, Levy, and Log-normal distributions are proposed to model the DCT coefficients of real-world sources. It will be shown that the proposed heavy-tailed distributions are more accurate to model non-strictly sparse sources rather than Laplace and Cauchy distributions. Maximum Likelihood Estimation (MLE) method is utilized to parameter estimation of the proposed distributions. Finally, the Shannon Lower Bound (SLB) on the rate distortion function of each model is calculated, accordingly a framework is provided that could be used to assess the rate distortion performance of any coding scheme that is supposed to compress DCT coefficients as outcome of a sparse source.

AB - In this study heavy-tailed distributions including Pareto, Weibull, Levy, and Log-normal distributions are proposed to model the DCT coefficients of real-world sources. It will be shown that the proposed heavy-tailed distributions are more accurate to model non-strictly sparse sources rather than Laplace and Cauchy distributions. Maximum Likelihood Estimation (MLE) method is utilized to parameter estimation of the proposed distributions. Finally, the Shannon Lower Bound (SLB) on the rate distortion function of each model is calculated, accordingly a framework is provided that could be used to assess the rate distortion performance of any coding scheme that is supposed to compress DCT coefficients as outcome of a sparse source.

KW - heavy-tailed distribution

KW - Levy distribution

KW - Log-normal distribution

KW - lossy source coding

KW - maximum likelihood estimation

KW - Pareto distribution

KW - sparse sources

KW - Weibull distribution

UR - https://www.scopus.com/pages/publications/85070995164

U2 - 10.1109/IranianCEE.2019.8786650

DO - 10.1109/IranianCEE.2019.8786650

M3 - Conference contribution

AN - SCOPUS:85070995164

T3 - ICEE 2019 - 27th Iranian Conference on Electrical Engineering

SP - 1570

EP - 1575

BT - ICEE 2019 - 27th Iranian Conference on Electrical Engineering

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 27th Iranian Conference on Electrical Engineering, ICEE 2019

Y2 - 30 April 2019 through 2 May 2019

ER -

Aramideh M, Namjoo E, Nooshyar M. Non-strictly Sparse Source Modelling Using Heavy-tailed Distributions for DCT Coefficients. In ICEE 2019 - 27th Iranian Conference on Electrical Engineering. Institute of Electrical and Electronics Engineers Inc. 2019. p. 1570-1575. 8786650. (ICEE 2019 - 27th Iranian Conference on Electrical Engineering). doi: 10.1109/IranianCEE.2019.8786650

Non-strictly Sparse Source Modelling Using Heavy-tailed Distributions for DCT Coefficients

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this