Restoring a video for improved watermark detection

Patent metadata

Field	Value
Publication number	US-12167105-B2
Application number	US-202217963844-A
Country	US
Kind code	B2
Filing date	Oct 11, 2022
Priority date	Sep 29, 2020
Publication date	Dec 10, 2024
Grant date	Dec 10, 2024

Abstract

A watermark detection method includes identifying, among plural stored videos, a second video having a second video feature matching a first video feature of a first video, the second video comprising watermark information. The method further includes comparing the first video and the second video to obtain a video conversion parameter, the video conversion parameter indicating a difference between the first video and the second video. The method also includes converting the first video based on the video conversion parameter to obtain a third video, and performing watermark detection on the third video to obtain watermark information in the third video. Apparatus and non-transitory computer-readable storage medium counterpart embodiments are also contemplated.

First claim

What is claimed is:

1. A watermark detection method comprising:
identifying, among plural stored videos, a second video having a second video feature matching a first video feature of a first video, the second video comprising watermark information;
traversing video frames in the first video, wherein for each traversed video frame in the first video,
based on a frame feature of the respective traversed video frame and frame features of a plurality of first candidate video frames in the second video, a respective similarity between the respective traversed video frame and each of the plurality of first candidate video frames is determined, and
a benchmark video frame from the plurality of first candidate video frames from the second video is determined, the benchmark video frame being a candidate video frame with a highest similarity to a first video frame in the first video;

determining a plurality of second candidate video frames from the second video, the plurality of second candidate video frames being video frames with a play interval from the benchmark video frame less than a first play duration;
determining a respective similarity between the first video frame and each second candidate video frame based on the frame feature of the first video frame and a frame feature of the respective second candidate video frame;
determining a video frame from the second video with a highest similarity to the first video frame, among the benchmark video frame and the plurality of second candidate video frames, as a second video frame;
comparing the first video frame in the first video and the second video frame in the second video to obtain a video conversion parameter, the video conversion parameter indicating a difference between the first video and the second video;
converting the first video based on the video conversion parameter to obtain a third video; and
performing watermark detection on the third video to obtain watermark information in the third video.

2. The method according to claim 1 , wherein the identifying comprises:
performing feature extraction on the first video to obtain the first video feature;
querying a database based on the first video feature to find the second video feature matching the first video feature; and
determining a video including the second video feature as the second video.

3. The method according to claim 2 , wherein
the database stores a plurality of template feature sets, each template feature set has a corresponding class label and comprises at least one template video feature; and
the querying comprises:
querying a class label corresponding to the first video feature among class labels of the plurality of template feature sets; and
determining the second video feature from a template feature set having the class label that corresponds to the first video feature.

4. The method according to claim 1 , wherein the comparing the first video frame and the second video frame to obtain the video conversion parameter comprises:
determining at least three key point pairs based on the first video frame and the second video frame, each key point pair comprising a first key point in the first video frame and a second key point in the second video frame matching the first key point;
determining a first conversion parameter based on differences between coordinates of each key point in the at least three key point pairs, the first conversion parameter representing a conversion function from the first video frame into the second video frame; and
determining the video conversion parameter based on the determined first conversion parameter.

5. The method according to claim 4 , wherein the determining the first conversion parameter comprises:
determining a plurality of second conversion parameters based on the coordinates of each key point in the at least three key point pairs, the plurality of second conversion parameters representing a conversion function from the first video frame into the second video frame, similarities between the plurality of second conversion parameters and an average conversion parameter corresponding to the plurality of second conversion parameters being all greater than a second threshold; and
determining the average conversion parameter corresponding to the plurality of second conversion parameters as the first conversion parameter.

6. The method according to claim 1 , wherein
the converting comprises:
converting each video frame in the first video based on the video conversion parameter to obtain a corresponding video frame in the third video; and
the performing watermark detection comprises:
performing the watermark detection on at least one video frame in the third video to obtain the watermark information.

7. The method according to claim 6 , wherein the converting each video frame in the first video based on the video conversion parameter to obtain the corresponding video frame in the third video comprises:
determining, based on video frames matching each other in the first video and the second video, a cropped video frame area in at least one video frame in the second video that was cropped out of at least one video frame in the first video;
filling the cropped video frame area in the at least one video frame in the first video; and
converting the at least one filled video frame based on the video conversion parameter to obtain the corresponding video frame in the third video.

8. The method according to claim 6 , wherein the performing the watermark detection on the at least one video frame in the third video to obtain the watermark information comprises:
obtaining a plurality of first frequency domain parameters of the at least one video frame, each first frequency domain parameter corresponding to one frequency band;
determining a second frequency domain parameter in the plurality of first frequency domain parameters, the second frequency domain parameter corresponding to a frequency band in which the watermark information is located; and
converting the second frequency domain parameter to obtain the watermark information.

9. A watermark detection apparatus, the apparatus comprising:
processing circuitry configured to
identify, among plural stored videos, a second video having a second video feature matching a first video feature of a first video, the second video comprising watermark information;
traverse video frames in the first video, wherein for each traversed video frame in the first video,
based on a frame feature of the respective traversed video frame and frame features of a plurality of first candidate video frames in the second video, a respective similarity between the respective traversed video frame and each of the plurality of first candidate video frames is determined, and
a benchmark video frame from the plurality of first candidate video frames from the second video is determined, the benchmark video frame being a candidate video frame with a highest similarity to a first video frame in the first video;

determine a plurality of second candidate video frames from the second video, the plurality of second candidate video frames being video frames with a play interval from the benchmark video frame less than a first play duration;
determine a respective similarity between the first video frame and each second candidate video frame based on the frame feature of the first video frame and a frame feature of the respective second candidate video frame;
determine a video frame from the second video with a highest similarity to the first video frame, among the benchmark video frame and the plurality of second candidate video frames, as a second video frame;
compare the first video frame in the first video and the second v

Assignees

• Tencent Tech Shenzhen Co Ltd

Inventors

• Fang Yang
• Wu Tao
• Xu Xuyuan

Classifications

• H04N21/44029

  for generating different versions · CPC title

• H04N21/44008

  involving operations for analysing video streams, e.g. detecting features or characteristics in the video stream (arrangements characterised by components specially adapted for monitoring, identification or recognition of video in broadcast systems H04H60/59) · CPC title

• H04N21/8358 Primary

  involving watermark {(protecting executable software by watermarking G06F21/16; image watermarking in general G06T1/0021; watermarks inserted in still images for transmission purposes H04N1/32144; inserting watermarks during video coding H04N19/467)} · CPC title

• H04N21/44004 Primary

  involving video buffer management, e.g. video decoder buffer or video display buffer · CPC title