US 11,887,277 B2
Removing compression artifacts from digital images and videos utilizing generative machine-learning models
Ionut Mironica, Bucharest (RO)
Assigned to Adobe Inc., San Jose, CA (US)
Filed by Adobe Inc., San Jose, CA (US)
Filed on Feb. 23, 2021, as Appl. No. 17/182,510.
Prior Publication US 2022/0270209 A1, Aug. 25, 2022
Int. Cl. G06T 5/00 (2006.01); G06T 9/00 (2006.01); G06N 3/088 (2023.01); G06N 3/045 (2023.01)
CPC G06T 5/001 (2013.01) [G06N 3/045 (2023.01); G06N 3/088 (2013.01); G06T 9/00 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A non-transitory computer-readable medium storing instructions which, when executed by a processing device, cause the processing device to perform operations comprising:
determining, utilizing one or more neural network layers, a compression ratio of a compressed digital image;
extracting features from the compressed digital image;
generating a combination of the compression ratio and the features extracted from the compressed digital image;
generating an improved digital image by utilizing a generator neural network comprising a plurality of dilated attention residual neural network layers to remove compression artifacts from the compressed digital image based on the compression ratio by processing the combination of the compression ratio and the features extracted from the compressed digital image; and
providing the improved digital image to a client device for display.