US 12,266,173 B2
Image recognition method, video playback method, related devices
Fasheng Chen, Shenzhen (CN)
Assigned to TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED, Shenzhen (CN)
Filed by TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED, Shenzhen (CN)
Filed on Mar. 30, 2022, as Appl. No. 17/708,419.
Application 17/708,419 is a continuation of application No. PCT/CN2021/085799, filed on Apr. 7, 2021.
Claims priority of application No. 202010390453.5 (CN), filed on May 11, 2020.
Prior Publication US 2022/0270365 A1, Aug. 25, 2022
Int. Cl. G06V 20/40 (2022.01); G06V 10/30 (2022.01); G06V 10/98 (2022.01); H04N 5/14 (2006.01); H04N 7/01 (2006.01); H04N 19/137 (2014.01); H04N 19/14 (2014.01); H04N 19/142 (2014.01); H04N 19/577 (2014.01)
CPC G06V 20/41 (2022.01) [G06V 10/98 (2022.01); G06V 20/46 (2022.01)] 18 Claims
OG exemplary drawing
 
1. An image recognition method, performed by a server, the method comprising:
obtaining a first frame image of which a frame type is known, and a second frame image and a third frame image of which frame types are unknown from an image frame sequence,
the second frame image being located between the first frame image and the third frame image in the image frame sequence,
the frame type comprising at least one of an effective frame, an overlapped frame, or a repeated frame,
the effective frame being a non-repeated and non-overlapped frame in the image frame sequence,
the repeated frame is a frame with an identical image with a previous frame in the image frame sequence, and
the overlapped frame is a frame obtained by respectively multiplying pixel values of respective adjacent image frames by weight coefficients of the respective adjacent image frames and add the multiplied pixel values of the respective adjacent image frames together;
constructing a first pixel sequence using pixel values of pixels in the first frame image, and constructing a third pixel sequence using pixel values of pixels in the third frame image;
obtaining a first image overlapping ratio between the second frame image and the first frame image and a second image overlapping ratio between the second frame image and the third frame image; and
performing weighted sum on the first pixel sequence and the third pixel sequence using the first image overlapping ratio and the second image overlapping ratio, to obtain predicted pixel sequence of the second frame image, the predicted pixel sequence comprising prediction values of pixels in the second frame image;
calculating a pixel error according to a second pixel sequence of the second frame image and the predicted pixel sequence, the second pixel sequence comprising pixel values of the pixels in the second frame image, the pixel error indicating a difference between the second pixel sequence and the predicted pixel sequence;
determining, in response to the pixel error meeting an error condition, that the frame type of the second frame image is the effective frame; and
determining, in response to the pixel error failing to meet the error condition, the frame type of the second frame image or the third frame image according to the first image overlapping ratio and the second image overlapping ratio.