A performance impact of Andrew’s Sine threshold for a robust regularized SRR based on ML framework

Published date
2016
Resource type
Publisher
ISBN
ISSN
DOI
Call no.
Other identifier(s)
Edition
Copyrighted date
Language
eng
File type
application/pdf
Extent
16 pages
Other title(s)
Advisor
Other Contributor(s)
Citation
Rangsit Journal of Arts and Sciences, Rangsit University, Vol. 6 No. 1, January-June 2016, 1-16
Degree name
Degree level
Degree discipline
Degree department
Degree grantor
Abstract
One of the most successful Digital Image Reconstruction (DIR) techniques for increasing image resolution and improving image quality is the Super-Resolution Reconstruction (SRR), which is the procedure of integrating a collection of aliased low-resolution low-quality images to form a single high-resolution high-quality image. However, the mainstream SRR algorithms are too delicate to noisy environments because these mainstream SRR algorithms are often comprised by the ML (L1 or L2) estimation techniques thereby the new robust SRR algorithm, which is comprised by Andrew’s Sine norm, has been proposed for dealing with noisy environments. Because the performance of the new SRR algorithm heavily relies on this Andrew’s Sine norm soft-threshold parameter, resultantly, this paper aims to investigate the impact characteristic of this norm constant parameter on the novel SRR algorithm. In addition, multitudinous experiments (which are applied on two standard images: Lena image and Susie image) are simulated to make the extensive results under five noise models: noise free, additive Gaussian noise, multiplicative Gaussian noise, Poisson noise and Impulsive noise with several noise powers for demonstrating the relationship between the SRR performance (in PSNR) and Andrew’s Sine norm soft-threshold parameter under each noisy cases.
Table of contents
Description
punsarn.dc.description.sponsorship
Spatial Coverage
Subject(s)
Rights
Access rights
Rights holder(s)
Location
View External Resources
Collections