Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
407067 | Neurocomputing | 2013 | 8 Pages |
•New feature extraction method based on QWT is proposed.•The GGD is used to describe the feature of the transformed coefficients.•New banknote classification system is proposed based on QWT.
In order to improve the performance of the banknote classification, this paper presents a new feature extraction method based on quaternion wavelet transform (QWT). The QWT yields one shift invariant magnitude and three phases based on quaternion algebra. The generalized Gaussian density (GGD) is applied to capture the statistical characteristics of QWT coefficients. The neural network is used as classifier in the framework of banknote classification. Experimental results demonstrate its effectiveness and the proposed method obtains a higher recognition rate in the banknote classification.