Improving Full-Reference Image Quality Assesment Using Machine Learning

Dimitrievski, Martin and Ivanovski, Zoran (2011) Improving Full-Reference Image Quality Assesment Using Machine Learning. In: Proceedings of the Eighth Conference on Informatics and Information Technology. Institute of Informatics, Faculty of Natural Sciences and Mathematics, Ss. Cyril and Methodius University in Skopje, Macedonia, Skopje, Macedonia, pp. 88-92. ISBN 978-9989-668-90-6

[img]
Preview
Text
978-9989-668-90-6_pp88-92.pdf

Download (442kB) | Preview
Official URL: http://ciit.finki.ukim.mk

Abstract

In this work we comprehensively analyze three standard image quality measures, measure their performance on gray scale images with different qualities against opinions from human viewers, and propose a novel image quality measure. Furthermore we inspect the behavior of several types of image degradations and the ability of each metric to detect the quantity of perceived loss of image quality for each degradation type. It was found that these measures based on first and second order statistics, although computationally simple, do not correlate very well to mean opinion scores. By means of machine learning we find an optimal regression model that combines these statistical metrics into a novel full reference image quality metric which correlates better to human scores for all tested degradation types.

Item Type: Book Section
Subjects: International Conference on Informatics and Information Technologies > Optimization
International Conference on Informatics and Information Technologies > Signal Processing
Depositing User: Vangel Ajanovski
Date Deposited: 28 Oct 2016 00:15
Last Modified: 28 Oct 2016 00:15
URI: http://eprints.finki.ukim.mk/id/eprint/11201

Actions (login required)

View Item View Item