Improving the technology for constructing a software tool to determine the similarity of raster graphic images

Abstract

The object of research is the process of searching and analyzing images of raster graphics. In the context of this work, the problem of the lack of an effective and fast procedure for determining the similarity of images was solved. The technology for improving the construction of a software tool for determining the similarity of raster graphics images by devising a procedure for determining the similarity of images using a hash code that corresponds to all variants of the image, regardless of size and aspect ratio, is reported. The image features of raster graphics were systematized. A procedure for determining the similarity of images using a hash code is proposed. This procedure makes it possible to process all possible variants of the image, regardless of the size and aspect ratio. The resulting indicator of the proposed procedure is the value of the hash codes. It is proposed to use the mathematical apparatus of fuzzy logic by introducing linguistic variables to estimate the similarity index. A comparison of the numerical values of similarity, obtained on the basis of the use of information systems, and the linguistic values revealed in the survey process was carried out. Threshold values were obtained that make it possible to assess the degree of similarity of the images. Based on the proposed algorithm, a prototype of the information system for determining the similarity of images of raster graphics has been designed. As a result of the calculation of the numerical characteristics of the improvement of the technology of constructing a software tool for determining the similarity of images of raster graphics, the value of the precision indicators was approximately 0.89 and the completeness was 0.8. The advantage of the proposed technology for determining the similarity of images over known analog technologies is illustrated by the amount of RAM of the developed software, which is 210 MB