V.P. Kozyrev, A.K. Troitsky, N.G. Volchenkov, A.A. Khramov

The main part of scientific and technical information is stored in paper form. It is needed to translate it to computer form for automatic processing. Images can contain text and graphics information. That is why the task to solve segmentation problem arises to separate text and graphics regions on images. Image segmentation method based on texture analysis theory is suggested in this paper. Spatial neighborhood matrix was chosen as an information characteristic for image region evaluation. It allows to separate images with different texture in feature area. The texture characteristics of images with text and graphics information were analyzed. Standard forms of spatial neighborhood matrix for both image types were identified. Image regions and boundaries classification algorithm based on these standard forms was suggested. This algorithm may be used for automatic input image systems.