US 12,125,171 B2
Video denoising method and apparatus, and storage medium
Zhengzhi Duan, Shenzhen (CN)
Assigned to TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED, Shenzhen (CN)
Filed by Tencent Technology (Shenzhen) Company Limited, Shenzhen (CN)
Filed on Nov. 5, 2021, as Appl. No. 17/520,169.
Application 17/520,169 is a continuation of application No. PCT/CN2020/119910, filed on Oct. 9, 2020.
Claims priority of application No. 201911251050.6 (CN), filed on Dec. 9, 2019.
Prior Publication US 2022/0058775 A1, Feb. 24, 2022
Int. Cl. G06K 9/00 (2022.01); G06T 5/50 (2006.01); G06T 5/70 (2024.01); H04N 19/105 (2014.01); H04N 19/159 (2014.01); H04N 19/176 (2014.01); H04N 19/85 (2014.01)
CPC G06T 5/70 (2024.01) [G06T 5/50 (2013.01); H04N 19/105 (2014.11); H04N 19/159 (2014.11); H04N 19/176 (2014.11); H04N 19/85 (2014.11); G06T 2207/10016 (2013.01)] 11 Claims
OG exemplary drawing
 
1. A video denoising method performed by a computing device, the method comprising:
acquiring a current frame of a target video;
in accordance with a determination that the current frame is a P frame or a B frame in the target video:
determining a plurality of reference frames of the current frame from the target video according to a pre-existing time domain reference relationship between the current frame and each of the plurality of reference frames that is established by an encoder and represented by a respective motion vector;
determining a plurality of reference blocks in the plurality of reference frames corresponding to a current block in the current frame, wherein a degree of matching between each reference block in a respective reference frame and the current block is greater than a matching threshold; and
performing denoising on the current block according to the plurality of reference blocks, further including:
performing pre-denoising on the current block by using pixels of one or more spatially neighboring blocks in the current frame and a two-dimensional (2D) denoising algorithm;
calculating a weight matrix of the plurality of reference blocks based on a three-dimensional (3D) denoising algorithm according to the respective degrees of matching between the plurality of reference blocks and the current block; and
performing weighted summation by using pixels of the current block, pixels of the plurality of reference blocks, and the corresponding weight matrix to obtain outputted pixels of the current block after denoising.