US 11,983,849 B2
Image filling method and apparatus, device, and storage medium
Chao Li, Beijing (CN); Dongliang He, Beijing (CN); Fu Li, Beijing (CN); and Hao Sun, Beijing (CN)
Assigned to BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD., Beijing (CN)
Filed by BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD., Beijing (CN)
Filed on Mar. 16, 2021, as Appl. No. 17/203,437.
Claims priority of application No. 202010610906.0 (CN), filed on Jun. 30, 2020.
Prior Publication US 2021/0201448 A1, Jul. 1, 2021
Int. Cl. G06T 9/00 (2006.01); G06F 18/214 (2023.01); G06T 3/4046 (2024.01); G06T 5/00 (2006.01); G06T 11/40 (2006.01); G06T 11/60 (2006.01); G06V 10/44 (2022.01); G06V 10/774 (2022.01); G06V 10/82 (2022.01)
CPC G06T 5/005 (2013.01) [G06F 18/214 (2023.01); G06T 3/4046 (2013.01); G06T 9/00 (2013.01); G06T 11/40 (2013.01); G06T 11/60 (2013.01); G06V 10/454 (2022.01); G06V 10/774 (2022.01); G06V 10/82 (2022.01)] 18 Claims
OG exemplary drawing
 
1. An image filling method, comprising:
performing multilevel encoding processing on features of an image to be filled to generate multilevel encoded feature layers, sizes of the multilevel encoded feature layers being reduced layer by layer, wherein the image to be filled has a missing region;
performing layer-by-layer decoding processing on the multilevel encoded feature layers to obtain multilevel decoded feature layers and a first image, there being no missing region in the first image, wherein the layer-by-layer decoding processing comprises a concatenation operation on a decoded feature layer of the multilevel decoded feature layers and an encoded feature layer of the multilevel encoded feature layers with a same size as the decoded feature layer; and
performing up-sampling processing on the first image to obtain first multilevel up-sampled feature layers and a second image optimized by the up-sampling processing, the up-sampling processing comprising a concatenation operation on an up-sampled feature layer of the first multilevel up-sampled feature layers and a decoded feature layer of the multilevel decoded feature layers with a same size as the up-sampled feature layer, wherein the first image is taken as an input to concatenate a first up-sampled feature layer of the first multilevel up-sampled feature layers obtained by up-sampling the first image and a decoded feature layer of the multilevel decoded feature layers with the same size as the first up-sampled feature layer, a concatenated feature layer is taken as an input of next up-sampling, and up-sampling processes is performed for multiple times to obtain the second image.