Application of various mathematical methods of extracting background and text elements from images

This paper presents the results of a research of the possible applications of different mathematical methods for the identification of duplicate lecture note images belonging to Dubna State University students. The analysis revealed that the duplicate images are exact copies or photographs of the same note taken under varying environmental conditions, necessitating background removal for subsequent comparison.For the purpose of background removal, a multi-step procedure (with usage of U-Net architecture) was compared with computer vision methods. The comparison revealed that the multi-step procedure showed better results. Furthermore, algorithms for subsequent analysis are proposed, including quick image comparison and Siamese neural network architecture.