US 11,798,132 B2
Image inpainting method and apparatus, computer device, and storage medium
Yi Wang, Shenzhen (CN); Xin Tao, Shenzhen (CN); and Jia Ya Jia, Shenzhen (CN)
Assigned to TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED, Shenzhen (CN)
Filed by TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED, Guangdong (CN)
Filed on Mar. 1, 2021, as Appl. No. 17/188,449.
Application 17/188,449 is a continuation of application No. PCT/CN2019/011958, filed on Nov. 20, 2019.
Claims priority of application No. 201811442428.6 (CN), filed on Nov. 29, 2018.
Prior Publication US 2021/0183022 A1, Jun. 17, 2021
Int. Cl. G06K 9/00 (2022.01); G06T 3/40 (2006.01); G06N 3/08 (2023.01); G06T 5/00 (2006.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01)
CPC G06T 3/4053 (2013.01) [G06N 3/08 (2013.01); G06T 3/4007 (2013.01); G06T 5/005 (2013.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01); G06T 2207/20076 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] 18 Claims
OG exemplary drawing
 
1. An image inpainting method, performed by at least one processor of a computer device, the method comprising:
determining, from a target image, a first region to be inpainted and a second region that is not to be inpainted;
performing feature extraction on the second region based on different receptive fields and spatial resolutions, to obtain feature information of a plurality of scales;
generating a texture of the first region based on the feature information of the plurality of scales; and
filling the first region in the target image with the generated texture, to obtain an inpainted image,
wherein the performing the feature extraction comprises:
obtaining a trained multi-column convolutional neural network, the trained multi-column convolutional neural network comprising a plurality of subnetworks connected in parallel, the different receptive fields and spatial resolutions being set for the plurality of subnetworks; and
respectively performing the feature extraction on the second region by using the plurality of subnetworks, to obtain feature information corresponding to the plurality of subnetworks as the feature information of the plurality of scales.