Texture Classification Based on Multi-scale Wavelet

Authors

Liao Ning, 1. Shewai School, Central South University of forestry and Technology, Changsha 410004, China;;
Lisha Xu, 1. Shewai School, Central South University of forestry and Technology, Changsha 410004, China;;
Xiaoshan Qian, 2. Physical Science and Technology College, Yichun University, Yichun 336000, China;

Abstract

Abstract: Texture analysis is quite sensitive to rotations. An efficient approach, called Invariant Contourlet-Fourier Descriptor, was proposed to achieve rotation invariance in texture analysis by extracting a set of Shannon entropy in contourlet domain. Discrete Fourier Transform analysis was applied to entropy vectors of each scale to form rotation invariant feature vectors, the dimensionality of which was reduced further due to the symmetry of DFT magnitude spectrum. Two classifiers, including the well-known Euclidean distance and Support Vector Machine, were studied to measure the distance between the known and unknown features. Experimental results on 1500 texture images show that contourlet is an efficient tool to represent directional texture, and the rotation invariant texture features were effective to achieve accurate classification with low computational complexity.

Recommended Citation

Ning, Liao; Xu, Lisha; and Qian, Xiaoshan (2020) "Texture Classification Based on Multi-scale Wavelet," Journal of System Simulation: Vol. 27: Iss. 9, Article 4.
Available at: https://dc-china-simulation.researchcommons.org/journal/vol27/iss9/4

First Page

1951

Revised Date

2015-07-22

DOI Link

https://doi.org/

Last Page

1959

CLC

TP18

Recommended Citation

Liao Ning, Xu Lisha, Qian Xiaoshan. Texture Classification Based on Multi-scale Wavelet[J]. Journal of System Simulation, 2015, 27(9): 1951-1959.