Abstract
Recognition of incomplete duplicates of images and tables is considered. In order to recognize graphical data (for image classification and compression), wavelet analysis is used with a set of classic characteristic functions: Morlet and Haar wavelets, Mexican hat wavelet, etc. Special types of filters are also used, which are based on the so-called ridgelet, curvlet and beamlet transformations. The main classical methods of image collection clustering that can be used to find incomplete duplicates in the graphic data of electronic documents are considered. The Harris method is analyzed, which allows to determine the reference points of the images by measuring the intensity of the brightness of the image. SIFT (scale-invariant feature transformation) technology, which is a powerful tool for forming a system of invariant structural features, is also analyzed, another class of methods is considered, which are easy to implement and use to detect incomplete duplicate images – hash methods. It is described that there are three such signals for RGB images: brightness in Red, Green and Blue channels. In signal processing and related branches of Fourier transform, decomposition of the signal into frequencies and amplitudes is usually considered. A method for identifying context-sensitive values and indexing textual data is considered, which helps to find incomplete duplicates in tables based on textual and numerical representation of data. Similarly, the described method can be used to index data of numerical and text types, if they are not placed in a table, but inside the content of an electronic document. The results of the research are used in combination with the system of detection of incomplete duplicates in scientific documents, in particular dissertations for the degree.