FIELD: computing technology.
SUBSTANCE: group of inventions relates to the field of artificial intelligence (AI) and can be used for forming an output image using a neural network. The method comprises the stages of: obtaining the image data collected by a color filter array (CFA data of the image); and joint debayering and noise elimination on the CFA data of the image using the trained neural network for the purpose of creating an output image, wherein the neural network has a simplified U-Net architecture and is trained on multiple pairs of training images, wherein one image in each pair of training images is obtained with a lower ISO value than the other image in the above pair of training images, and processed by means of a processing algorithm (ISP), wherein the other image in each pair of training images is presented in the format of the CFA data of the image, and the ISP processing comprises separate implementation of noise eliminate and debayering.
EFFECT: ensured joint debayering and elimination of digital noise in images for the purpose of improving image quality.
21 cl, 8 dwg, 1 tbl
