VIDEO THUMBNAIL SELECTION

• US 20110064318A1
• Filed: 09/17/2009
• Published: 03/17/2011
• Est. Priority Date: 09/17/2009
• Status: Active Grant

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Xiao, Jun, Gao, Yuli, Zhang, Tong
• Granted Patent: US 8,571,330 B2
• Time in Patent Office: Days
• US Class Current: 382/224
• CPC Class Codes: G06F 16/70 of video data
• CPC Class Codes: G06F 16/73 Querying

Abstract

A method for selecting a video thumbnail includes generating a visual theme model for a sample set of images that are representative of textual information corresponding to a video file. Each of a set of candidate key frames is distinguished according to similarities shared between the candidate key frames and the visual theme model. A display is caused of a selected one of the distinguished candidate key frames as a video thumbnail for the video file.

Claims

1. A method for selecting a video thumbnail, comprising:

generating a visual theme model for a sample set of images that are representative of textual information corresponding to a video file;

distinguishing each of a set of candidate key frames according to similarities shared between that candidate key frame and the visual theme model; and

causing a display a selected one of the distinguished candidate key frames as a video thumbnail for the video file.

2. The method of claim 1, further comprising obtaining the textual information, the textual information being obtained automatically from one or more of:

a filename

a title of the video file;

a tag associated with the video file;

an extraction of close caption text in the video file; and

a conversion of audio content contained in the video file.

3. The method of claim 1, further comprising:

causing a network image search utilizing the textual information as a search term or terms;

identifying a ranked set of images corresponding to results of the image search; and

selecting, according to rank, a plurality of images from the ranked set as the sample set of images.

4. The method of claim 1, wherein generating a visual theme model for the sample set of images comprises, for a given image feature type, identifying a weighted value for each of a set of features shared among the sample set of images.

5. The method of claim 4, wherein distinguishing each of a set of candidate key frames comprises:

evaluating each of the candidate key frames against the weighted feature values; and

ranking each of the candidate key frames according to similarities shared between that candidate key frame and the weighted feature values.

6. The method of claim 1, wherein generating a visual theme model for the sample set of images comprises, for each of a plurality of image feature types, identifying a weighted value for each of a set of features shared among the sample set of images.

7. The method of claim 6, wherein distinguishing each of a set of candidate key frames comprises:

identifying an image feature type of the plurality of image feature types having a greatest disparity in feature value weights;

evaluating each of the candidate key frames against the weighted feature values of the identified feature type; and

ranking each of the candidate key frames according to similarities shared between that candidate key frame and the weighted feature values of the identified feature type.

8. A computer readable medium storing computer executable instructions that when executed implement a method for selecting a video thumbnail, the method comprising:

generating a visual theme model for a sample set of images that are representative of textual information corresponding to a video file;

distinguishing each of a set of candidate key frames according to similarities shared between that candidate key frame and the visual theme model; and

causing a display a selected one of the distinguished candidate key frames as a video thumbnail for the video file.

9. The medium of claim 8, wherein the method further comprises obtaining the textual information, the textual information being obtained automatically from one or more of:

a file name of the video file;

a title of the video file;

a tag associated with the video file;

an extraction of close caption text in the video file; and

a conversion of audio content contained in the video file.

10. The medium of claim 8, wherein the method further comprises:

causing a network image search utilizing the textual information as a search term or terms;

identifying a ranked set of images corresponding to results of the image search; and

selecting, according to rank, a plurality of images from the ranked set as the sample set of images.

11. The medium of claim 8, wherein generating a visual theme model for the sample set of images comprises, for a given image feature type, identifying a weighted value for each of a set of features shared among the sample set of images.

12. The medium of claim 11, wherein distinguishing each of a set of candidate key frames comprises:

evaluating each of the candidate key frames against the weighted feature values; and

ranking each of the candidate key frames according to similarities shared between that candidate key frame and the weighted feature values.

13. The medium of claim 8. wherein generating a visual theme model for the sample set of images comprises, for each of a plurality of image feature types, identifying a weighted value for each of a set of features shared among the sample set of images.

14. The medium of claim 13, wherein distinguishing each of a set of candidate key frames comprises:

identifying an image feature type of the plurality of image feature types having a greatest disparity in feature value weights;

evaluating each of the candidate key frames against the weighted feature values of the identified feature type; and

ranking each of the candidate key frames according to similarities shared between that candidate key frame and the weighted feature values of the identified feature type.

15. A system for selecting a video thumbnail, comprising:

a modeling engine operable to generate a visual theme model for the sample set of images that are representative of textual information corresponding to a video file;

an evaluation engine operable to distinguish each of a set of candidate key frames according to similarities shared between that candidate key frame and the visual theme model; and

a display engine operable to cause a display a selected one of the ranked candidate key frames as a video thumbnail for the video file.

16. The system of claim 15, wherein the modeling engine is operable to obtain the textual information from one or more of:

a file name of the video file;

a title of the video file;

a tag associated with the video file;

an extraction of close caption text in the video file; and

a conversion of audio content contained in the video file.

17. The method of claim 15, wherein the modeling engine is operable to:

cause a network image search utilizing the textual information as a search term or terms;

identify a ranked set of images corresponding to results of the image search; and

select, according to rank, a plurality of images from the ranked set as the sample set of images.

18. The system of claim 15, wherein the modeling engine is operable to generate a visual theme model for the sample set of images by, for a given image feature type, identifying a weighted value for each of a set of features shared among the sample set of images.

19. The system of claim 18, wherein the evaluation engine is operable to distinguishing each of a set of candidate key frames by:

evaluating each of the candidate key frames against the weighted feature values; and

ranking each of the candidate key frames according to similarities shared between that candidate key frame and the weighted feature values.

20. The system of claim 19, wherein the modeling engine is operable to generate a visual theme model for the sample set of images by, for each of a plurality of image feature types, identifying a weighted value for each of a set of features shared among the sample set of images and wherein the evaluation engine is operable to distinguish each of a set of candidate key frames by:

identifying an image feature type of the plurality of image feature types having a greatest disparity in feature value weights;

evaluating each of the candidate key frames against the weighted feature values of the identified feature type; and

ranking each of the candidate key frames according to similarities shared between that candidate key frame and weighted feature values of the identified feature type.