UMP Institutional Repository

Multi-Scale Colour Completed Local Binary Patterns for Scene and Event Sport Image Categorisation

Rassem, Taha H. and Bee, Ee Khoo and Makbol, Nasrin M. and Al-Sewari, Abdul Rahman Ahmed Mohammed (2017) Multi-Scale Colour Completed Local Binary Patterns for Scene and Event Sport Image Categorisation. IAENG International Journal of Computer Science, , 44 (2). pp. 197-211. ISSN 1819-9224

[img]
Preview
PDF
IJCS_44_2_09.pdf

Download (4MB) | Preview

Abstract

The Local Binary Pattern (LBP) texture descriptor and some of its variant descriptors have been successfully used for texture classification and for a few other tasks such as face recognition, facial expression, and texture segmentation. However, these descriptors have been barely used for image categorisation because their calculations are based on the gray image and they are only invariant to monotonic light variations on the gray level. These descriptors ignore colour information despite their key role in distinguishing the objects and the natural scenes. In this paper, we enhance the Completed Local Binary Pattern (CLBP), an LBP variant with an impressive performance on texture classification. We propose five multiscale colour CLBP (CCLBP) descriptors by incorporating five different colour information into the original CLBP. By using the Oliva and Torralba (OT8) and Event sport datasets, our results attest to the superiority of the proposed CCLBP descriptors over the original CLBP in terms of image categorisation.

Item Type: Article
Uncontrolled Keywords: Local Binary Pattern (LBP), Texture Descriptors, Completed Local Binary Pattern (CLBP), colour CLBP (CCLBP), Image Categorisation
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Faculty/Division: Faculty of Computer System And Software Engineering
Depositing User: Dr. Taha Hussein Alaaldeen Rassem
Date Deposited: 29 May 2017 05:59
Last Modified: 29 Mar 2018 07:56
URI: http://umpir.ump.edu.my/id/eprint/17765
Download Statistic: View Download Statistics

Actions (login required)

View Item View Item