Textures classification based on DWT

摘要

Nowadays classification of textures is very actual. Classification means separation of textures into classes. The main task of the classification is to find such a descriptor which is unique for the given class. This article deals with the classification based on energies distributed into DWT subimages. Description vectors are created on the base of subimages energies. The classification is done by comparison of the Euclidian or the Mahalnobis distance between the description vector of the input texture and the description vectors of classes. The efficiency of the classification is evaluated with regard to level of DWT as well. For purposes of experimental verification of this classification method a database is created. This database consist of 21 classes and each class has at least three textures.