Author:
Vincent Barra
Affiliation:
LIMOS, UMR 6158, France
Keyword(s):
Local binary pattern, Multispectral image, Segmentation, Texture, Total orderings.
Related
Ontology
Subjects/Areas/Topics:
Computer Vision, Visualization and Computer Graphics
;
Feature Extraction
;
Features Extraction
;
Image and Video Analysis
;
Informatics in Control, Automation and Robotics
;
Segmentation and Grouping
;
Signal Processing, Sensors, Systems Modeling and Control
Abstract:
Texture is an important feature when considering image segmentation. Since more and more image segmentation problems involve multi- and hyperspectral data, including color images, it becomes necessary to define multispectral texture features. In this article, we propose LMBP, an extension of the classical Local Binary Pattern (LBP) operator to the case of multispectral images. The LMBP operator is based on the definition of total orderings in the image space and on an extension of the standard univariate LBP. It allows the computation of both a multispectral texture structure coefficient and a multispectral contrast parameter for each spatial location, that serve as an input to an unsupervised clustering algorithm. Results are demonstrated in the case of the segmentation of brain tissues from multispectral MR images, and compared to other multispectral texture features.