Detection of smooth texture in facial images for the evaluation of unnatural contrast enhancement

Nur Halilah Binti Ismail, Soong Der Chen

Research output: Contribution to journalArticle

Abstract

This paper presents an algorithm for detecting smooth texture in facial images which is prone to unnatural contrast enhancement. The algorithm consists of texture analysis and machine learning algorithm. Wavelet decomposition is used for texture analysis. Smooth texture tends to have small variance among the wavelet coefficients within the same scale. This paper proposes to divide image into 32×32 sub-image with overlapping of 16 pixels, then perform wavelet decomposition with 5 scales. The final feature is a 5 dimensional vector consists of the variance of the wavelet coefficients from each of the 5 scales. Support Vector Machine (SVM) is used for feature classification. The SVM classifier was trained using 468 samples consist of samples from skin areas (smooth texture) and non-smooth area (eye and nose) of 78 test images. The performance of the classifier was evaluated using k-fold cross validation with k range from 2 to 10. The performance was excellent with the average accuracy for each value of k above 95%. The performance was also very consistent across different set of test images with standard deviation range from 1% ~ 4%.

Original languageEnglish
Pages (from-to)215-220
Number of pages6
JournalJournal of Theoretical and Applied Information Technology
Volume85
Issue number2
Publication statusPublished - 20 Mar 2016

Fingerprint

Contrast Enhancement
Texture
Textures
Evaluation
Texture Analysis
Wavelet decomposition
Wavelet Decomposition
Wavelet Coefficients
Support vector machines
Support Vector Machine
Classifiers
Classifier
Cross-validation
Range of data
Standard deviation
Learning algorithms
Skin
Divides
Overlapping
Learning systems

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

@article{8092a09113fa4f42a4da18a339e0d07a,
title = "Detection of smooth texture in facial images for the evaluation of unnatural contrast enhancement",
abstract = "This paper presents an algorithm for detecting smooth texture in facial images which is prone to unnatural contrast enhancement. The algorithm consists of texture analysis and machine learning algorithm. Wavelet decomposition is used for texture analysis. Smooth texture tends to have small variance among the wavelet coefficients within the same scale. This paper proposes to divide image into 32×32 sub-image with overlapping of 16 pixels, then perform wavelet decomposition with 5 scales. The final feature is a 5 dimensional vector consists of the variance of the wavelet coefficients from each of the 5 scales. Support Vector Machine (SVM) is used for feature classification. The SVM classifier was trained using 468 samples consist of samples from skin areas (smooth texture) and non-smooth area (eye and nose) of 78 test images. The performance of the classifier was evaluated using k-fold cross validation with k range from 2 to 10. The performance was excellent with the average accuracy for each value of k above 95{\%}. The performance was also very consistent across different set of test images with standard deviation range from 1{\%} ~ 4{\%}.",
author = "Ismail, {Nur Halilah Binti} and Chen, {Soong Der}",
year = "2016",
month = "3",
day = "20",
language = "English",
volume = "85",
pages = "215--220",
journal = "Journal of Theoretical and Applied Information Technology",
issn = "1992-8645",
publisher = "Asian Research Publishing Network (ARPN)",
number = "2",

}

Detection of smooth texture in facial images for the evaluation of unnatural contrast enhancement. / Ismail, Nur Halilah Binti; Chen, Soong Der.

In: Journal of Theoretical and Applied Information Technology, Vol. 85, No. 2, 20.03.2016, p. 215-220.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Detection of smooth texture in facial images for the evaluation of unnatural contrast enhancement

AU - Ismail, Nur Halilah Binti

AU - Chen, Soong Der

PY - 2016/3/20

Y1 - 2016/3/20

N2 - This paper presents an algorithm for detecting smooth texture in facial images which is prone to unnatural contrast enhancement. The algorithm consists of texture analysis and machine learning algorithm. Wavelet decomposition is used for texture analysis. Smooth texture tends to have small variance among the wavelet coefficients within the same scale. This paper proposes to divide image into 32×32 sub-image with overlapping of 16 pixels, then perform wavelet decomposition with 5 scales. The final feature is a 5 dimensional vector consists of the variance of the wavelet coefficients from each of the 5 scales. Support Vector Machine (SVM) is used for feature classification. The SVM classifier was trained using 468 samples consist of samples from skin areas (smooth texture) and non-smooth area (eye and nose) of 78 test images. The performance of the classifier was evaluated using k-fold cross validation with k range from 2 to 10. The performance was excellent with the average accuracy for each value of k above 95%. The performance was also very consistent across different set of test images with standard deviation range from 1% ~ 4%.

AB - This paper presents an algorithm for detecting smooth texture in facial images which is prone to unnatural contrast enhancement. The algorithm consists of texture analysis and machine learning algorithm. Wavelet decomposition is used for texture analysis. Smooth texture tends to have small variance among the wavelet coefficients within the same scale. This paper proposes to divide image into 32×32 sub-image with overlapping of 16 pixels, then perform wavelet decomposition with 5 scales. The final feature is a 5 dimensional vector consists of the variance of the wavelet coefficients from each of the 5 scales. Support Vector Machine (SVM) is used for feature classification. The SVM classifier was trained using 468 samples consist of samples from skin areas (smooth texture) and non-smooth area (eye and nose) of 78 test images. The performance of the classifier was evaluated using k-fold cross validation with k range from 2 to 10. The performance was excellent with the average accuracy for each value of k above 95%. The performance was also very consistent across different set of test images with standard deviation range from 1% ~ 4%.

UR - http://www.scopus.com/inward/record.url?scp=84962552242&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84962552242&partnerID=8YFLogxK

M3 - Article

VL - 85

SP - 215

EP - 220

JO - Journal of Theoretical and Applied Information Technology

JF - Journal of Theoretical and Applied Information Technology

SN - 1992-8645

IS - 2

ER -