An Improved Texture Feature Extraction
Method for Tyre Tread Patterns
Ying Liu, Zong Li, Zi-Ming Gao
( School of Communication and Information Engineering,
Xi’an University of Posts and Telecommunications, Xi’an 710061,China)
xupt2011@126.com
Abstract: Tamura features have been found to be effective in describing image
textures and contrast is one of the Tamura features popularly used. As a global
variable, contrast can well describe the statistical distribution of the brightness
in the entire image, but cannot reflect the local brightness information of the
image. To solve this problem, this paper proposes an improved texture feature
extraction method which makes use of the statistical moments of intensity
histogram to extract more information from the image. Tested on a tyre tread
pattern dataset, the proposed method is found to be able to provide better
retrieval performance than other existing methods.
Keywords: Image Retrieval; Tamura feature; Contrast; Statistical moments
1 Introduction
With the rapid popularization of the Internet and multimedia technology, the size of
image collection is increasing rapidly. In order for intelligent and efficient image
database management, content-based image retrieval (CBIR) have been developed
since 1980's. CBIR performs image retrieval based on the similarity measure of image
features such as color feature, texture feature, shape feature and spatial location.
These features are extracted from image content[1].
Texture features describes the homogeneity of image surface and the spatial
distribution of different elements not depending on the color and brightness
information. It reflects the global and local structure of images and has been widely
used for image retrieval. There are many texture feature extraction methods designed
for CBIR, including wavelet-based texture feature, Tamura feature, feature
Gray-Level Co-occurrence Matrix and so on[1]. Tamura texture features have six
dimensions including coarseness, contrast, directionality, line-likeness, regularity and
roughness, which are designed based on human visual system. The first three have
been commonly used for image retrieval [1,2].
To further improve the performance of Tamura features for image retrieval, the
author in [3] modified the definitions of coarseness and directionality. This work
intends to find effective texture extraction method for type tread patterns as a special
type of data for our project in public security area. This paper proposes a new method
to calculate the contrast feature by making use of the statistical moments of intensity
histogram to extract more information from the image. Intensive experiments on a
tyre tread pattern dataset were carried out to test the performance of the proposed
method by comparing its retrieval performance with other existing texture feature
extraction methods. The simulation results prove the effectiveness of the proposed