US 10,893,283 B2
Real-time adaptive video denoiser with moving object detection
Jing Chen, Mountain View, CA (US); Marco Paniconi, Campbell, CA (US); and Michael Horowitz, Manhattan Beach, CA (US)
Assigned to Google LLC, Mountain View, CA (US)
Filed by Google LLC, Mountain View, CA (US)
Filed on Nov. 26, 2019, as Appl. No. 16/695,911.
Application 16/695,911 is a continuation of application No. 15/254,723, filed on Sep. 1, 2016, granted, now 10,511,846.
Prior Publication US 2020/0099944 A1, Mar. 26, 2020
This patent is subject to a terminal disclaimer.
Int. Cl. H04N 19/36 (2014.01); H04N 19/137 (2014.01); H04N 19/117 (2014.01); H04L 12/801 (2013.01); H04N 19/14 (2014.01); H04N 19/176 (2014.01); H04N 19/172 (2014.01); H04N 19/182 (2014.01)
CPC H04N 19/36 (2014.11) [H04L 47/29 (2013.01); H04N 19/117 (2014.11); H04N 19/137 (2014.11); H04N 19/14 (2014.11); H04N 19/172 (2014.11); H04N 19/176 (2014.11); H04N 19/182 (2014.11)] 20 Claims
OG exemplary drawing
 
1. A method for adaptive noise filtering of a source video, the method comprising:
receiving a source frame of the source video from a video capturing device;
removing noise from the source frame by:
dividing the source frame into source blocks;
performing moving object detection to identify first source blocks that are moving blocks;
performing a noise estimation for the source frame;
identifying one or more second source blocks that are static blocks;
calculating a noise value for the source frame based on a sum of weighted variances for a number of consecutive static blocks;
calculating an average noise value over a fixed running window; and
responsive to the average noise value satisfying a noise value threshold, performing at least one of lowering a variance threshold for the moving object detection or lowering a difference threshold for noise filtering;
generating an output frame based on the moving object detection and the noise estimation; and
encoding the output frame.