Automatic detection of logos in video sequences
Particular embodiments automatically identify and track a logo that appears in video content. For example, particular embodiments can track a branding logo's position and size without any prior knowledge about the logo, such as the position, type, structure, and content of the logo. In one embodiment, a heat map is used that accumulates a frequency of short-term logos that are detected in the video content. The heat map is then used to identify a branding logo in the video content.
1. A method comprising:
generating an image-based set of accumulators, the image-based set of accumulators indicating a frequency and a position of each of a set of short-term logos determined in a first set of frames in a video sequence of video content;
detecting whether a second frame includes a first short-term logo of the set of short-term logos, wherein the second frame is in the video sequence and is not included in the first set of frames;
based on a result of the detecting, determining whether a position of the first short-term logo for the second frame corresponds to a matching position indicated in the image-based set of accumulators;
based on a result of the detecting, determining whether a frequency of the first short-term logo in the image-based set of accumulators is above a threshold; and
if the frequency of the first short-term logo is above the threshold:
identifying the first short-term logo as a branding logo shown more frequently than others of the set of short-term logos, and
outputting information for the image-based set of accumulators based on the matching position, wherein an encoding system uses the information in an encoding process.
2. The method of claim 1 , wherein:
the image-based set of accumulators comprises a first boundary, the first short-term logo for the second frame comprises a second boundary, and
determining whether the short-term logo for the second frame matches the image-based set of accumulators by comparing the first boundary to the second boundary.
3. The method of claim 2 , wherein comparing the first boundary to the second boundary comprises determining when the first boundary overlaps the second boundary by an overlap threshold.
4. The method of claim 1 , wherein detecting the first short-term logo for the second frame comprises:
performing a first edge detection on each of the set of short-term logos for the first set of frames;
performing a second edge detection on the second frame; and
comparing the first edge detection with the second edge detection to detect the first short-term logo for the second frame.
5. The method of claim 1 , further comprising:
updating the frequency of the image-based set of accumulators based on detecting the first short-term logo in the second frame.
6. The method of claim 1 , wherein outputting the information comprises outputting the matching position in the image-based set of accumulators to the encoding system along with a frame identifier for the second frame.
7. The method of claim 1 , wherein outputting the information comprises outputting a size of the image-based set of accumulators.
8. The method of claim 1 , wherein the encoding system uses the information to encode the matching position in the image-based set of accumulators differently from video information at a different position in the second frame.
9. The method of claim 1 , wherein detecting the first short-term logo for the second frame comprises:
determining a short-term logo template including a first set of short-term logos detected in the video sequence: and
comparing the second frame to the first set of short-term logos in the short-term logo template to detect the first short-term logo in the second frame.
10. A non-transitory computer-readable storage medium comprising instructions that, when executed, control the one or more computer processors to be configured for:
generating an image-based set of accumulators, the image-based set of accumulators indicating a frequency and a position of each of a set of short-term logos determined in a first set of frames in a video sequence of video content;
detecting whether a second frame includes a first short-term logo of the set of short-term logos, wherein the second frame is in the video sequence and is not included in the first set of frames;
based on a result of the detecting, determining whether a position of the first short-term logo for the second frame corresponds to a matching position indicated in the image-based set of accumulators; and
based on a result of the detecting, determining whether a frequency of the first short-term logo in the image-based set of accumulators is above a threshold; and
if the frequency of the first short-term logo is above the threshold:
identifying the first short-term logo as a branding logo shown more frequently than others of the set of short-term logos, and
outputting information for the image-based set of accumulators based on the matching position, wherein an encoding system uses the information in an encoding process.
11. The non-transitory computer-readable storage medium of claim 10 , wherein:
the image-based set of accumulators comprises a first boundary, the first short-term logo for the second frame comprises a second boundary, and
determining whether the first short-term logo for the second frame matches the image-based set of accumulators by comparing the first boundary to the second boundary.
12. The non-transitory computer-readable storage medium of claim 10 , wherein detecting the short-term logo for the second frame comprises:
performing a first edge detection on each of the set of short-term logos for the first set of frames;
performing a second edge detection on the second frame; and
comparing the first edge detection with the second edge detection to detect the first short-term logo for the second frame.
13. The non-transitory computer-readable storage medium of claim 10 , further comprising:
updating the frequency of the image-based set of accumulators based on detecting the first short-term logo in the second frame.
14. The non-transitory computer-readable storage medium of claim 10 , wherein outputting the information comprises outputting the matching position in the image-based set of accumulators to the encoding system along with a frame identifier for the second frame.
15. The non-transitory computer-readable storage medium of claim 10 , wherein detecting the short-term logo for the second frame comprises:
determining a short-term logo template including a first set of short-term logos detected in the video sequence; and
comparing the second frame to the first set of short-term logos in the short-term logo template to detect the first short-term logo in the second frame.
16. A method comprising:
generating a map of image-based set of accumulators based on a first set of short-term logos in a first set of frames in a video sequence, each of the image-based set of accumulators indicating a frequency and a position of respective short-term logos detected in the first set of frames in the video sequence;
determining a short-term logo template including the first set of short-term logos in the first set of frames in the video sequence;
determining a current frame in the video sequence;
comparing the current frame to the first set of short-term logos in the short-term logo template to determine a second set of short-term logos in the current frame;
updating a map of image-based set of accumulators based on the second set of short-term logos;
based on a result of the comparing, determining whether a position of a first short-term logo in the first set of short-term logos corresponds to a matching position indicated in the first image-based set of accumulators; and
based on a result of the comparing, determining whether a frequency of the first short-term logo in the first image-based set of accumulators is above a threshold; and
if the frequency of the first short-term logo is above the threshold:
identifying the first short-term logo as a branding logo shown more frequently than others of the first set of short-term logos, and
outputting information for the first image-based set of accumulators in the map of image-based set of accumulators when a match occurs, wherein an encoding system uses the position of the first image-based set of accumulators in an encoding process.
17. The method of claim 16 , wherein comparing the current frame to the first set of short-term logos comprises:
performing a first edge detection on each of the first set of short-term logos for the first set of frames;
performing a second edge detection on the current frame; and
comparing the first edge detection with the second edge detection to detect the second set of short-term logos for the current frame.
18. The method of claim 16 , wherein updating the map of image-based set of accumulators comprises:
updating the frequency of the image-based set of accumulators at positions where the short-term logos are detected in the current frame.
19. The method of claim 16 , wherein outputting the information comprises outputting the matching position in the first image-based set of accumulators to the encoding system along with a frame identifier for the current frame.
20. The method of claim 16 , wherein outputting the information comprises outputting a size of the first image-based set of accumulators.