Granted Patent B2

US 12,524,849 · App. 18/472,001 · Granted Jan 13, 2026

Inventors: Song Yang (Beijing, CN); Yulong Liu (Beijing, CN)

Assignee: BEIJING ZITIAO NETWORK TECHNOLOGY CO., LTD.

G06T5/70G06T7/73G06T2207/10016

Abstract

Claims (66)

1 . A method for video anti-shake processing, comprising:

determining a moving track of a shooting position of a video by performing feature point tracking between different image frames in the video, wherein the moving track indicates shooting positions of the different image frames in the video;

performing smoothing processing on a shooting position of each image frame of the different image frames in the moving track, to obtain a smooth track; and

deforming the video based on a difference between the smooth track and the moving track, to obtain an anti-shake processed video,

wherein the smooth track is dynamically determined for the different image frames in the video, and

said performing smoothing processing on the shooting position of each image frame of the different image frames in the moving track to obtain the smooth track comprises:

performing smoothing processing on the shooting position of each image frame of the different image frames in the moving track based on a smooth radius corresponding to the image frame in the video to obtain the smooth track, wherein the smooth radius corresponding to each image frame of the different image frames in the video is determined based on a number interval into which the number of the feature points used for tracking in the image frame in the video falls, and a correspondence between number intervals and smooth radii.

2 . The method according to claim 1 , wherein said performing smoothing processing on the shooting position of each image frame of the different image frames in the moving track based on the smooth radius corresponding to the image frame in the video to obtain the smooth track comprises:

determining, based on the smooth radius corresponding to each frame of image in the video, a preset number of frames of images involved in each smoothing processing from the video; and

performing, based on the shooting positions of the preset number of frames of images involved in each smoothing processing in the moving track, smoothing processing on the shooting position of each frame of image in the moving track, to obtain the smooth track.

3 . The method according to claim 2 , wherein said determining, based on the smoothing radius corresponding to each frame of image in the video, the preset number of frames of images involved in each smoothing processing from the video comprises:

based on the smooth radius corresponding to each frame of image in the video, determining a first preset number of previous frames of images before the frame of image from the video; and determining each frame of image and the first preset number of previous frames of images as the preset number of frames of images involved in each smoothing processing from the video; or

based on the smooth radius corresponding to each frame of image in the video, determining a second preset number of previous frames of images before the frame of image from the video, and determining the second preset number of subsequent frames of images after the frame of image from the video; and determining each frame of image, the second preset number of previous frames of images and the second preset number of subsequent frames of images as the preset number of frames of images involved in each smoothing processing from the video.

4 . The method according to claim 2 , wherein said performing, based on the shooting positions of the preset number of frames of images involved in each smoothing processing in the moving track, smoothing processing on the shooting position of each frame of image in the moving track to obtain the smooth track comprises:

performing, based on the shooting positions of the preset number of frames of images involved in each smoothing processing in the moving track, weighted summation calculation on the shooting positions, to obtain a smooth position for each frame of image; and

obtaining the smooth track based on the smooth position for each frame of image.

5 . The method according to claim 1 , wherein the moving track is represented by transformation matrices, and different transformation matrices in the moving track respectively represent the shooting positions of the different image frames in the video, and

said determining the moving track of the shooting position of the video by performing feature point tracking between the different image frames in the video comprises:

determining a transformation matrix between the shooting positions of the different image frames in the video by performing feature point tracking between the different image frames in the video, and determining the moving track of the shooting position of the video based on the transformation matrix between the shooting positions of the different image frames in the video.

6 . The method according to claim 1 , wherein said deforming the video based on the difference between the smooth track and the moving track to obtain the anti-shake processed video comprises:

determining an adjustment parameter based on the difference between the smooth track and the moving track; and

deforming the video based on the adjustment parameter to obtain the anti-shake processed video.

7 . The method according to claim 1 , wherein the moving track is dynamically determined for the different image frames in the video, and

said determining the moving track of the shooting position of the video by performing feature point tracking between the different image frames in the video comprises:

determining an initial change amount between the shooting positions of the different image frames in the video based on an initial transformation mode by performing feature point tracking between the different image frames in the video;

determining a target change amount between the shooting positions of the different image frames in the video based on a target transformation mode matching a fitting error corresponding to the initial change amount; and

forming the moving track of the shooting position of the video based on the target change amount between the shooting positions of the different image frames in the video, wherein the moving track indicates the shooting positions of the different image frames in the video.

8 . The method according to claim 7 , further comprising:

performing coordinate transformation on a feature point on a previous frame of image among the different image frames based on the initial transformation mode, to obtain a transformation coordinate of the feature point on the previous frame of image; and

calculating the fitting error corresponding to the initial change amount between the shooting positions of the different image frames based on an image coordinate of the feature point on a subsequent frame of image among the different image frames and the transformation coordinate of the feature point on the previous frame of image.

9 . The method according to claim 8 , wherein said calculating the fitting error corresponding to the initial change amount between the shooting positions of the different image frames based on the image coordinate of the feature point on the subsequent frame of image among the different image frames and the transformation coordinate of the feature point on the previous frame of image comprises:

calculating an accumulated error corresponding to the initial change amount between the shooting positions of the different image frames, based on image coordinates of feature points on the subsequent frame of image among the different image frames and transformation coordinates of feature points on the previous frame of image; and

calculating the fitting error corresponding to the initial change amount between the shooting positions of the different image frames, based on the accumulated error and the number of the feature points on the previous frame of image.

10 . The method according to claim 7 , wherein said determining the target change amount between the shooting positions of the different image frames in the video based on the target transformation mode matching the fitting error corresponding to the initial change amount comprises:

when the fitting error corresponding to the initial change amount is smaller than a first error threshold, determining the initial transformation mode as the target transformation mode matching the fitting error, and determining the target change amount between the shooting positions of the different image frames in the video based on the target transformation mode; or

when the fitting error corresponding to the initial change amount is greater than or equal to the first error threshold and smaller than a second error threshold, determining a first transformation mode with a smaller degree of freedom than the initial transformation mode as the target transformation mode matching the fitting error, and determining the target change amount between the shooting positions of the different image frames in the video based on the target transformation mode; or

when the fitting error corresponding to the initial change amount is greater than or equal to the second error threshold, determining a second transformation mode with a smaller degree of freedom than the first transformation mode as the target transformation mode matching the fitting error, and determining the target change amount between the shooting position of the different image frames in the video based on the target transformation mode.

11 . The method according to claim 10 , wherein the initial transformation mode comprises a homography transformation mode, the first transformation mode comprises an affine transformation mode, and the second transformation mode comprises a similarity transformation mode.

12 . The method according to claim 7 , wherein the initial change amount or the target change amount is represented by a transformation matrix, and

said forming the moving track of the shooting position of the video based on the target change amount between the shooting positions of the different image frames in the video comprises:

determining a transformation matrix of each frame of image in the video relative to a first frame of image based on the target transformation matrix between shooting positions of the different image frames in the video; and

forming the moving track of the shooting position of the video based on the transformation matrix of each frame of image in the video relative to the first frame of image.

13 . An electronic device, comprising a memory and a processor, wherein the memory has a computer program stored thereon, the computer program, when executed by the processor, causes the electronic device to implement a method for video anti-shake progressing, the method comprising:

determining a moving track of a shooting position of a video by performing feature point tracking between different image frames in the video, wherein the moving track indicates shooting positions of the different image frames in the video;

performing smoothing processing on a shooting position of each image frame of the different image frames in the moving track, to obtain a smooth track; and

deforming the video based on a difference between the smooth track and the moving track, to obtain an anti-shake processed video,

wherein the smooth track is dynamically determined for the different image frames in the video, and

said performing smoothing processing on the shooting position of each image frame of the different image frames in the moving track to obtain the smooth track comprises:

performing smoothing processing on the shooting position of each image frame of the different image frames in the moving track based on a smooth radius corresponding to the image frame in the video to obtain the smooth track, wherein the smooth radius corresponding to each image frame of the different image frames in the video is determined based on a number interval into which the number of the feature points used for tracking in the image frame in the video falls, and a correspondence between number intervals and smooth radii.

14 . The electronic device according to claim 13 , wherein the moving track is dynamically determined for the different image frames in the video, and

said determining the moving track of the shooting position of the video by performing feature point tracking between the different image frames in the video comprises:

determining an initial change amount between the shooting positions of the different image frames in the video based on an initial transformation mode by performing feature point tracking between the different image frames in the video;

determining a target change amount between the shooting positions of the different image frames in the video based on a target transformation mode matching a fitting error based on the fitting error corresponding to the initial change amount; and

forming the moving track of the shooting position of the video based on the target change amount between the shooting positions of the different image frames in the video, wherein the moving track indicates the shooting positions of the different image frames in the video.

15 . A non-volatile computer readable storage medium, having a computer program stored thereon, the computer program, when executed by a computing device, causes the computing device to implement a method for video anti-shake progressing, the method comprising:

determining a moving track of a shooting position of a video by performing feature point tracking between different image frames in the video, wherein the moving track indicates shooting positions of the different image frames in the video;

performing smoothing processing on a shooting position of each image frame of the different image frames in the moving track, to obtain a smooth track; and

deforming the video based on a difference between the smooth track and the moving track, to obtain an anti-shake processed video,

wherein the smooth track is dynamically determined for the different image frames in the video, and

said performing smoothing processing on the shooting position of each image frame of the different image frames in the moving track to obtain the smooth track comprises:

performing smoothing processing on the shooting position of each image frame of the different image frames in the moving track based on a smooth radius corresponding to the image frame in the video to obtain the smooth track, wherein the smooth radius corresponding to each image frame of the different image frames in the video is determined based on a number interval into which the number of the feature points used for tracking in the image frame in the video falls, and a correspondence between number intervals and smooth radii.

16 . The non-volatile computer readable storage medium according to claim 15 , wherein the moving track is dynamically determined for the different image frames in the video, and

said determining the moving track of the shooting position of the video by performing feature point tracking between the different image frames in the video comprises:

determining an initial change amount between the shooting positions of the different image frames in the video based on an initial transformation mode by performing feature point tracking between the different image frames in the video;

determining a target change amount between the shooting positions of the different image frames in the video based on a target transformation mode matching a fitting error based on the fitting error corresponding to the initial change amount; and

forming the moving track of the shooting position of the video based on the target change amount between the shooting positions of the different image frames in the video, wherein the moving track indicates the shooting positions of the different image frames in the video.

Assignments (3)

Priority Claims (2)

Continuity (3)

References Cited (46)