Image noise severity metric

We propose in this paper an image noise severity measurement method that correlates well with human's quality perception on the presence of noise in images. In our approach, a 32x32 pixels mask is used to compute the differences between the original and noise-degraded images in terms of the sta...

Full description

Bibliographic Details
Main Authors: Bong, D.B.L, Khoo, B.E
Format: Article
Published: Proceedings of the SPIE 2012
Subjects:
Online Access:http://ir.unimas.my/id/eprint/3013/
_version_ 1848835124439810048
author Bong, D.B.L
Khoo, B.E
author_facet Bong, D.B.L
Khoo, B.E
author_sort Bong, D.B.L
building UNIMAS Institutional Repository
collection Online Access
description We propose in this paper an image noise severity measurement method that correlates well with human's quality perception on the presence of noise in images. In our approach, a 32x32 pixels mask is used to compute the differences between the original and noise-degraded images in terms of the statistical means and outlier values. These differences are formulated and then compared to the quality scores from the subjective evaluations. The degraded images were distorted by two common types of random noise for images - Gaussian white noise and impulse noise. Experiment results showed that this approach obtained higher correlation compare to classical Peak Signal to Noise Ratio (PSNR) method.
first_indexed 2025-11-15T06:02:53Z
format Article
id unimas-3013
institution Universiti Malaysia Sarawak
institution_category Local University
last_indexed 2025-11-15T06:02:53Z
publishDate 2012
publisher Proceedings of the SPIE
recordtype eprints
repository_type Digital Repository
spelling unimas-30132015-03-24T00:27:27Z http://ir.unimas.my/id/eprint/3013/ Image noise severity metric Bong, D.B.L Khoo, B.E TK Electrical engineering. Electronics Nuclear engineering We propose in this paper an image noise severity measurement method that correlates well with human's quality perception on the presence of noise in images. In our approach, a 32x32 pixels mask is used to compute the differences between the original and noise-degraded images in terms of the statistical means and outlier values. These differences are formulated and then compared to the quality scores from the subjective evaluations. The degraded images were distorted by two common types of random noise for images - Gaussian white noise and impulse noise. Experiment results showed that this approach obtained higher correlation compare to classical Peak Signal to Noise Ratio (PSNR) method. Proceedings of the SPIE 2012 Article NonPeerReviewed Bong, D.B.L and Khoo, B.E (2012) Image noise severity metric. Proceedings of the SPIE, 8334. 83343I-83343I. http://adsabs.harvard.edu/abs/2012SPIE.8334E.124B
spellingShingle TK Electrical engineering. Electronics Nuclear engineering
Bong, D.B.L
Khoo, B.E
Image noise severity metric
title Image noise severity metric
title_full Image noise severity metric
title_fullStr Image noise severity metric
title_full_unstemmed Image noise severity metric
title_short Image noise severity metric
title_sort image noise severity metric
topic TK Electrical engineering. Electronics Nuclear engineering
url http://ir.unimas.my/id/eprint/3013/
http://ir.unimas.my/id/eprint/3013/