Histogram hierarchical algorithm and the reduction of the dimensionality of the spectral features space

This paper proposes the algorithm for dimension reduction of data in the process of hierarchical histogram clustering data of remote sensing of the Earth. Application of the algorithm is illustrated to multispectral data. Clustering large amount of data remote sensing is usually carried out in two ways: by K medium (in advance, you must know the number of clusters K and an approximation of the data distribution), and histogram. Here we propose a hierarchical histogram algorithm, which does not require to specify the number of clusters and is quick. This paper considers the issue of reducing the dimension of own space of features, obtained by hierarchical histogram algorithm. Getting clusters of multispectral image, pay attention to the fact that the different clusters corresponding to different objects on Earth can be characterized by different dimensionality of the data, i.e., the set of spectral channels coming from the satellite, it may be unnecessary for a number of objects. Also, the level of detail of clustering can be different in different clusters.