Solving the Problem of Automated 2D Images Compositional Characteristics Evaluation
Gazimzyanov, F. F., Al Akkad, M. A.
Article language: English
Abstract. This paper contains an overview of the progress achieved by researchers in solving the problem of evaluating 2D images compositional characteristics. A review of methods potentially suitable for solving such a problem is given, the selected method is justified, and the adaptation of the selected method to a specific task is given. A mathematical model adapted to work with existing models using a new method called Adjusting the Structural Skeleton Coefficients is presented. The structure of the training samples and the special aspects of data collection are described. Data analysis and sorting is performed using a developed genetic algorithm, and the choice of the method is justified. The ob-tained results are analyzed, and visualization of the compositional parameters of simple scenes, for different groups of respondents identified during data sorting and analysis, is introduced. Finally, the overall results of the research are presented, concluding that they coincide with the suggestion of Arnheim about perception.
Keywords: structural skeleton, image analysis, computer vision, perception, genetic algorithms
Pages: 33–39Total pages: 7
Funding, support: This research is funded by Kalashnikov Izhevsk State Technical University grant 27.06.01/18BCB.
Year of publication: 2019