Rough-Neural Image Classification using Wavelet Transform

• Published in: 2007 International Conference on Machine Learning and Cybernetics Vol. 6; pp. 3045 - 3050
• Main Authors: Jun-Hai Zhai, Xi-Zhao Wang, Su-Fang Zhang
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.08.2007
• Subjects: Artificial neural networks; Data mining; Decorrelation; Feature extraction; Frequency; Image classification; Information systems; Neural networks; Rough sets; Statistical features; Wavelet coefficients; Wavelet transform; Wavelet transforms
• ISBN: 1424409721; 9781424409723
• ISSN: 2160-133X
• DOI: 10.1109/ICMLC.2007.4370671

Image classification plays an important role in many tasks, which is still a challenging problem. This paper proposes a hybrid image classification method, which integrates wavelet transform, rough set approach, and artificial neural networks (ANNs). Wavelet transform is employed to decompose the original images into different frequency sub-bands, then a set of statistical features are extracted from the wavelet coefficients, the feature set can be viewed as an information system. Although wavelet transform well decorrelates images, there still exist dependencies between coefficients. Hence the features extracted from the coefficients may be correlated. If the features from one sub-band are dependent on the features from another sub-band, the later one can be removed. Rough set approach is utilized to remove the correlated or redundant features. The reduced information system finally fed into neural network for classification. The performance of the method is evaluated in terms of training accuracy and testing accuracy, the experimental results confirm the effectiveness of the proposed approach.