Nasirov F. Methods of low-level analysis of images in application television systems

Object - processes low-level analysis of television pictures. Subject of research - the low-level image processing systems in applied television systems. Methods of research - To develop methods for determining the number of large-scale color segments and identify boundaries on the television pictures were used methods of mathematical statistics, singular spectrum analysis, linear algebra, pattern recognition theory, the theory of algorithms, mathematical modeling and computer simulation. Theoretical and practical results: the application to automatically determine the number of large-scale clusters on the television pictures, by expanding the image on the singular components and analysis of the third eigenvector singular value decomposition. Application for determination of the orientation of the boundary within the mask, by analyzing the mutual position of the center of mass and geometric center of the mask; application for the allocation of boundaries on a television image by analyzing the factorial structure of the vector corresponding to the maximum eigenvalue. For the different orientations of the border using different methods of image vectorization mask; a set of programs for low-level analysis of video images, which performs the following functions: setting the number of segments, determining the orientation and selection borders; Practical results of the thesis introduced in Star Interactive company in a context of image analysis systems and software modules in the company DGTL as software components in a video monitoring system, which is confirmed by relevant documents.