METHOD OF GENERATING A COMMON LOSS FUNCTION FOR TRAINING A CONVOLUTIONAL NEURAL NETWORK FOR CONVERTING AN IMAGE INTO AN IMAGE WITH DRAWN PARTS AND A SYSTEM FOR CONVERTING AN IMAGE INTO AN IMAGE WITH DRAWN PARTS Russian patent published in 2019 - IPC G06T5/00 

FIELD: image processing means.

SUBSTANCE: present group of inventions relates to the field of image processing, in particular to a method and a system for converting an image into an image with drawn parts. Method comprises steps of: obtaining a pair of images, including low-quality input P_L and high-quality input fragment P_H; feeding each of the input fragments (P_L, P_H) to the input of the corresponding Siamese convolutional neural subnetwork and processing the input fragments (P_L, P_H) to obtain output fragments (P’_L, P’_H) of the image; calculating a regression difference D(P’_L, P_H); modulated retentional difference D(P’_L, P_L) ⊙D(P’_H, P_H) is calculated; generating a common loss function and training the convolutional neural network based on the generated common loss function.

EFFECT: technical result consists in providing an image, on which small parts of the image are restored and traced in a natural way and integrity of extended objects on the image, having property of self-similarity.

4 cl, 7 dwg