Elaboration of structural representation of regions of scanned document images for MRC model

• Main Authors: Alesya Ishchenko, Marina Polyakova, Varvara Kuvaieva, Alexandr Nesteryuk
• Format: Article
• Language: English
• Published: PC Technology Center 2018-11-01
• Series: Eastern-European Journal of Enterprise Technologies
• Subjects: scanned document image; mixed raster content model; text extraction; image layer
• Online Access: http://journals.uran.ua/eejet/article/view/147671

The Mixed Raster Content (MRC) model is a common form of representation of a scanned document image. The further development of this model, which consists in the structural representation of homogeneous regions on each layer of the MRC image has been completed. The aim of such representation is to detect regions of interest in the image and identify it to solve the problem of segmentation of scanned document images. The layer containing graphic and photographic images was represented as a union of several regions using a piecewise constant function of the intensity of the image region. For this, the graphic and photographic images were represented in the form of a partition into segments containing pixels of uniform intensity. To determine these regions in order to separate the graphic from the photo images, the edges at the segment boundaries were considered. The layer containing the text was represented as an image of the regions of the structural texture on a uniform background. These regions contained fragments of normal text and heading with the same pixel intensity and differing in the shape and size of symbols, as well as the distance between them. Such a representation of the layer made it possible to take into account the spatial relationships between symbol image pixels and further to extract text from the background. The proposed model for the representation of the scanned document image allows to extracting the image layers that contain homogeneous regions, and reducing the process of segmentation of the entire image to the segmentation of separate layers of the image. This allows increasing the performance speed while maintaining a high quality of image segmentation