WestminsterResearch

Multidimensional image selection and classification system based on visual feature extraction and scaling

Mancusi, Francesco and Triantaphillidou, Sophie and Allen, Elizabeth (2010) Multidimensional image selection and classification system based on visual feature extraction and scaling. In: Farnand, Susan P. and Gaykema, Frans, (eds.) Image quality and system performance VII : 18-19 January 2010, San Jose, California, United States. Proceedings of SPIE (7529). SPIE, Bellingham, Wash., A1-A11. ISBN 9780819479228

[img]
Preview
PDF
870Kb
[img]
Preview
PDF (Copyright notice - READ FIRST)
23Kb

Official URL: http://dx.doi.org/10.1117/12.838734

Abstract

Sorting and searching operations used for the selection of test images strongly affect the results of image quality investigations and require a high level of versatility. This paper describes the way that inherent image properties, which are known to have a visual impact on the observer, can be used to provide support and an innovative answer to image selection and classification. The selected image properties are intended to be comprehensive and to correlate with our perception. Results from this work aim to lead to the definition of a set of universal scales of perceived image properties that are relevant to image quality assessments. The initial prototype built towards these objectives relies on global analysis of low-level image features. A multidimensional system is built, based upon the global image features of: lightness, contrast, colorfulness, color contrast, dominant hue(s) and busyness. The resulting feature metric values are compared against outcomes from relevant psychophysical investigations to evaluate the success of the employed algorithms in deriving image features that affect the perceived impression of the images.

Item Type:Book Section
Research Community:University of Westminster > Electronics and Computer Science, School of
ID Code:8956
Deposited On:16 Dec 2010 12:42
Last Modified:16 Dec 2010 12:42

Repository Staff Only: item control page