Blind feature selection and extraction in a 3D image Cube

As the number of spectral bands of high spectral resolution increases, the ability to detect more detailed classes should also increase, and the classification accuracy should increase as well. As the number of spectral bands becomes large, the limitation on performance imposed by the limited number of observed samples can become severe. A number of techniques for specific feature selection and extraction have been developed to reduce the dimensionality and target detection without the loss of class reparability. This suggests the need for reducing the dimensionality via a preprocessing method that takes into consideration the high dimensional feature space properties. Such reduction should enable the estimation of the feature extraction parameters to be more accurate bu using the technique referred to as the maximum conditional log likelihood estimation. This technique is able to bypass many of the problems of the limitation of small/large numbers of observed samples by making the computations in a lower dimensional space, and optimizing the function called the expectation and maximization. This method leads also to a high dimensional version of the feature selection and extraction algorithm, which requires significantly less computation than the normal procedure. A set of tests with real data evaluates the performance and illustrates the effectiveness of the proposed method. The entire work is done by using MATLAB.