Dimension Reduction of Texture Features Based upon Independent Component Analysis

Texture features are often employed as supplementary information in order to supplement the inadequacy of spectral information and increase the classification results of high spatial resolution remotely sensed imagery.However,the dimensions of texture features are always large,which obstruct their overall application in pattern recognition.Independent component analysis is applied to reduce the dimensions of texture features.Compared to the conventional one PCA method,ICA can achieve better classification accuracy and show a better compatibility for different texture features such as SWT,SVD and GLCM etc.