US 12,236,558 B2
Device and method for synthesizing image capable of improving image quality
Gyeong Su Chae, Seoul (KR); and Guem Buel Hwang, Seoul (KR)
Assigned to DEEPBRAIN AI INC., Seoul (KR)
Appl. No. 17/763,941
Filed by DEEPBRAIN AI INC., Seoul (KR)
PCT Filed Jun. 8, 2021, PCT No. PCT/KR2021/007145
§ 371(c)(1), (2) Date Mar. 25, 2022,
PCT Pub. No. WO2022/169036, PCT Pub. Date Aug. 11, 2022.
Claims priority of application No. 10-2021-0014560 (KR), filed on Feb. 2, 2021.
Prior Publication US 2023/0177664 A1, Jun. 8, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06T 5/50 (2006.01); G06T 5/00 (2024.01); G06V 10/77 (2022.01); G06V 10/774 (2022.01); G06V 10/82 (2022.01); G10L 15/16 (2006.01); G10L 15/22 (2006.01); G10L 25/57 (2013.01)
CPC G06T 5/50 (2013.01) [G06T 5/00 (2013.01); G06V 10/7715 (2022.01); G06V 10/774 (2022.01); G06V 10/82 (2022.01); G10L 15/16 (2013.01); G10L 15/22 (2013.01); G10L 25/57 (2013.01); G06T 2207/10016 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/20212 (2013.01); G06T 2207/30196 (2013.01)] 10 Claims
OG exemplary drawing
 
1. An image synthesis device having one or more processors and a memory which stores one or more programs executed by the one or more processors, the device comprising:
a first artificial neural network provided to learn each of a first task of using a damaged image as an input to output a restored image and a second task of using an original image as an input to output a reconstructed image; and
a second artificial neural network connected to an output layer of the first artificial neural network, and trained to use the reconstructed image output from the first artificial neural network as an input and improve the image quality of the reconstructed image,
wherein, in the first artificial neural network, an objective function Lsynth for learning the first task is represented by Equation 1 below:
Lsynth=∥X−Xsynth∥  [Equation 1]
where X: Original image;
Xsynth: Restored image output from first artificial neural network;
∥X−Xsynth∥: Function for obtaining difference between X and Xsynth; and
in the first artificial neural network, an objective function Lrecon for learning the second task is represented by Equation 2 below:
Lrecon=∥X−Xrecon∥  [Equation 2]
where Xrecon: Reconstructed image output from first artificial neural network; and
∥X−Xrecon∥: Function for obtaining difference between X and Xrecon.