Peculiarities of the raster images of hieroglyphic texts digitalization process organization

Abstract

 An analysis of modern techniques for the use of software and neural network algorithms in solving the task of automating the digitization and subsequent restoration of text data, which include the Chinese characters "Hanzi", was carried out. It is noted that when restoring printed hieroglyphic texts, it is possible to effectively use procedures based on morphological operators, the adaptation of which to the given task makes it possible to reduce the load on the computing resource of the hardware and software platform of the machine analysis complex, as well as to reduce the delay when restoring text data presented in the form raster image. It is indicated that the efficiency of hieroglyphic text recovery significantly increases when the text block is divided into constituent elements and groups of elements according to the multi-level classification structure, such as the "sentence-phrase-word" structure. Similarly, effective restoration of a single hieroglyph is possible when it is divided into radicals, and radicals further into separate components, such as strokes and combinations of strokes of a fixed slope and thickness according to the font. A method of restoration of the hieroglyphic text presented in the form of raster images is proposed, based on the application of morphological procedures of erosion and dilation, which are used to remove defects of the image matrix and restore the structure of individual hieroglyphs. The specified approach consists in choosing the matrix of the graphic primitive of the morphological algorithm procedure and rotating the image matrix according to the thickness and angle of inclination of individual lines as elementary components of hieroglyphs. At the same time, within the framework of the study, the effectiveness of the recovery of the hieroglyphic text can also be carried out in an automated mode according to the target indicators of the accuracy of the reconstruction process. Thus, the task of optimizing the data conversion algorithm can be carried out by determining the maximum of the objective function, the arguments of which are the size and shape of the graphic primitive, the rotation angle of the image matrix, as well as the sequence of the erosion and dilation procedures.