US 12,190,474 B2
Super-resolution reconstruction method and related apparatus
Ru Wang, Shenzhen (CN); Yaqing Li, Shenzhen (CN); Chengjie Tu, Shenzhen (CN); Shiyao Xiong, Shenzhen (CN); Linyan Jiang, Shenzhen (CN); and Longtao Peng, Shenzhen (CN)
Assigned to TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED, Palo Alto, CA (US)
Filed by Tencent Technology (Shenzhen) Company Limited, Shenzhen (CN)
Filed on May 6, 2022, as Appl. No. 17/739,004.
Application 17/739,004 is a continuation of application No. PCT/CN2021/086584, filed on Apr. 12, 2021.
Claims priority of application No. 202010435082.8 (CN), filed on May 21, 2020.
Prior Publication US 2022/0261960 A1, Aug. 18, 2022
Int. Cl. G06T 3/4053 (2024.01); G06T 3/4046 (2024.01)
CPC G06T 3/4053 (2013.01) [G06T 3/4046 (2013.01); G06T 2207/10016 (2013.01); G06T 2207/20081 (2013.01)] 20 Claims
OG exemplary drawing
 
1. An artificial intelligence-based super-resolution reconstruction method performed by a terminal device, the method comprising:
obtaining a to-be-processed video frame sequence of a first resolution, the to-be-processed video frame sequence including a plurality of to-be-processed video frames;
for each to-be-processed video frame in the to-be-processed video frame sequence:
performing resolution reconstruction according to a resolution reconstruction model so as to obtain an initial reconstructed video frame of a second resolution, wherein the second resolution is greater than the first resolution;
determining a contour region and a flat region in the initial reconstructed video frame according to a residual separation method; and
performing contour enhancement on the contour region and denoising on the flat region to obtain a target reconstructed video frame; and
generating a reconstructed video frame sequence of the second resolution according to a plurality of target reconstructed video frames corresponding to the plurality of to-be-processed video frames.