Texture Classification Using Modulus Extremum of Wavelet Frame Representation
Yulong Qiao, Sheng-He Sun · 2006
The wavelet transform modulus extremum is considered as one of the most meaningful characteristics of a signal. This paper proposes a feature based on the density of modulus extrema of the wavelet frame representation for texture classification. It is compared with existing features by using three representative classifiers, k-nearest neighbor classifier, learning vector quantization and support vector machines. The experimental results on two well-known databases indicate that our proposed feature is superior to other features. The same conclusion can be drawn after feature selection