The article discusses the influence of ANN topology on its efficiency in solving the problem of increasing the image resolution. An empirical approach is used to establish the fact and determine the nature of the influence. At the beginning of the article, the most commonly used topological techniques for constructing an ANN, which are used to solve the problem of increasing the image resolution, are described. Then, the process of creating an ANN is described based on the above topologies. After that, the learning process of ANN is described. A supervised learning algorithm was used to train the networks, and a set of 7 000 images was used as training data. At the end of the article, an assessment of the efficiency of the trained ANN is carried out, through which the effectiveness of topological solutions is determined. To assess the performance of an artificial neural network, a validation dataset of 100 image is used. Two algorithms are used to assess the quality of enhanced images: SSIM and PSNR. The interpretation of the results obtained is also given.

The article is devoted to an artificial neural network that increases image resolution. According to the meaning of the article, two parts can be distinguished. The first part of the article discusses the process of designing an artificial neural network using DAGNet technology, during which its main parameters and structure are described. The second part is devoted to assessing the quality of functioning of the developed artificial neural network. To evaluate the performance of ANNs a comparative analysis with the bicubic interpolation method was used.

No abstract

The article discusses the influence of the choice of a color model for representing a raster image on the efficiency of increasing its resolution using an artificial neural network. The article begins with the essence of the research. After that, the process of obtaining for training networks with an identical topology of a dataset using different color models of image representation, namely RGB, HSV, L * a * b *, NTSC and YCbCr, is described. After that, an assessment of the effectiveness of trained neural networks on the previously described datasets is given. To assess the performance of an artificial neural network, two algorithms are used: SSIM and PSNR. As a result of the assessment, the networks using RGB and YCbCr color models showed the highest results. At the end of the article, there are reflections on the reasons for this result.

No abstract