Improved filter algorithm using inequality fano to select bands for HSI classification

Hyperspectral imagery (HSI) is a remote sensing tool that precisely serves to define the classification of the regions. In fact, the coverage of several images of the ground truth, which provide relevant information, but some of them are influenced by atmospheric noise, and others contain a redundant information. To reduce the dimensionality of Hyperspectral Images, numerous studies using mutual information (MI) also the normalized Mutual information based heuristic to select the appropriate bands for the classification of HSI. Here we expect some methods present a filter strategy based on the measure of (MI), also there is wrapper strategies with error probability, the latter is more efficient than filter strategy, but more expensive. In this paper we will introduce a filter strategy with the error probability measure in order to have more precision in the selections bands with an optimal manner. This method can improve the filter strategy performance. The studies are conducted using HSI AVIRIC92AV3C.