Device and method for detecting the presence of a logo in a picture

Patent metadata

Field	Value
Publication number	US-9311533-B2
Application number	US-201314028425-A
Country	US
Kind code	B2
Filing date	Sep 16, 2013
Priority date	Sep 17, 2012
Publication date	Apr 12, 2016
Grant date	Apr 12, 2016

Abstract

A method for detecting the presence of a logo in a current picture of a video is disclosed. The method comprises:
determining a current candidate logo bounding box in the current picture and a preceding candidate logo bounding box in a picture preceding the current picture; detecting the presence of a logo in the current picture if at least one of the following conditions is fulfilled: the distance between the centers of the current and the preceding candidate logo bounding boxes is below a first threshold or an overlap between the current and the preceding candidate logo bounding boxes is above a second threshold; or an overlap between the current candidate logo bounding box and a saliency mask of the current picture is above a third, wherein the saliency mask identifies regions around saliency peaks.

First claim

What is claimed is:

1. A method for detecting a presence of a logo in a current picture of a video comprising:
determining a current candidate logo bounding shape in said current picture and a preceding candidate logo bounding shape in a picture preceding said current picture;
calculating a first score value based at least partially on whether a distance between a center of said current and said preceding candidate logo bounding shapes is below a first threshold or whether an overlap between said current and said preceding candidate logo bounding shapes is above a second threshold;
calculating a second score value based at least partially on whether an overlap between the current candidate logo bounding shape and a saliency mask of the current picture is above a third threshold, wherein the saliency mask identifies regions around saliency peaks; and
detecting the presence of a logo in the current picture based at least partially on a comparison of the first score value and the second score value.

2. The method according to claim 1 , wherein said candidate logo bounding shape is a candidate logo bounding box.

3. The method according to claim 2 , further comprising,
increasing a confidence level by N each time the presence of a logo is detected for a current picture of said video; and
decreasing the confidence level by N each time the presence of a logo is not detected for a current picture, the presence of a logo being finally detected for the current picture when said confidence level is positive.

4. The method according to claim 3 , wherein N equals 1.

5. The method according to claim 3 , further comprising, increasing a value of N after several consecutive detections or non-detections are made.

6. The method according to claim 2 , further comprising:
checking whether the current candidate logo bounding box comprises a peak of saliency, when the presence of a logo is detected in the current picture; and
adding said current candidate logo bounding box to a list of logo candidate bounding boxes, when the current candidate logo bounding box comprises the peak of saliency, the logo in said current picture being a bounding box computed from the list of candidate logo bounding boxes when said list comprises M candidate logo bounding boxes and the logo in said current picture being the current candidate logo bounding box otherwise.

7. The method according to claim 6 , wherein M equals 10.

8. The method according to claim 2 , wherein determining the current candidate logo bounding box comprises:
computing an edge map for each picture of a detection window in at least one search area, said detection window comprising the current picture and P pictures preceding said current picture;
summing edge maps over the pictures into an accumulated edge map and binarizing the accumulated edge map;
filtering the binarized accumulated edge map into a filtered edge map;
binarizing the filtered edge map to into a map of blobs;
selecting, among the blobs containing pixels whose value in the filtered edge map is the highest, the blob containing the highest sum of pixel values in the filtered edge map ; and
determining a box bounding the selected blob, said candidate bounding box being said bounding box.

9. The method according to claim 8 , further comprising
sliding said detection window by k pictures; and
detecting a logo of a next picture, upon sliding said detection window.

10. The method according to claim 8 , wherein said at least one search area comprises four corners of the current picture and top and bottom horizontal banners.

11. A device for detecting a presence of a logo in a current picture of a video comprising:
a memory;
at least one processor configured to:
determine a current candidate logo shape in said current picture and a preceding candidate logo shape in a picture preceding said current picture;
calculate a first score value based at least partially on whether a distance between a center of said current and said preceding candidate logo bounding shapes is below a first threshold or whether an overlap between said current and said preceding candidate logo bounding shapes is above a second threshold;
calculate a second score value based at least partially on whether an overlap between the current candidate logo bounding shape and a saliency mask of the current picture is above a third threshold, wherein the saliency mask identifies regions around saliency peaks; and
detect the presence of a logo in the current picture based at least partially on a comparison of the first score value and the second score value.

12. The device according to claim 11 , wherein said candidate logo bounding shape is a candidate logo bounding box.

13. The device according to claim 12 , wherein the at least one processor is further configured to increase a confidence level by N each time the presence of a logo is detected for a current picture of said video and to decrease said confidence level by N otherwise, the presence of a logo being finally detected for the current picture when said confidence level is positive.

14. The device according to claim 13 , wherein N equals 1.

15. The device according to claim 13 , wherein the at least one processor is further configured to increase a value of N after several consecutive detections or non-detections are made.

16. The device according to claim 12 , wherein the at least one processor is further configured to:
determine a logo in said current picture when the presence of said logo is detected;
determine whether the current candidate logo bounding box comprises a peak of saliency; and
add said current candidate logo bounding box to a list of logo candidate bounding boxes if the current candidate logo bounding box comprises the peak of saliency, the logo in said current picture being a bounding box computed from the list of candidate logo bounding boxes when said list comprises M candidate logo bounding boxes and the logo in said current picture being the current candidate logo bounding box otherwise.

17. The device according to claim 16 , wherein M equals 10.

18. A non-transitory computer readable medium with instructions stored therein which, upon execution, instruct at least one processor to:
determine a current candidate logo bounding shape in a current picture and a preceding candidate logo bounding shape in a picture preceding said current picture;
calculate a first score value based at least partially on whether a distance between center of said current and said preceding candidate logo bounding shapes is below a first threshold or whether an overlap between said current and said preceding candidate logo bounding shapes is above a second threshold;
calculate a second score value based at least partially on whether an overlap between the current candidate logo bounding shape and a saliency mask of the current picture is above a third threshold, wherein the saliency mask identifies regions around saliency peaks; and
detect a presence of a loqo in the current picture based at least partially on a comparison of the first score value and the second score value.

Assignees

• Thomson Licensing

Inventors

• Huynh-Thu Thai Q
• Urban Fabrice

Classifications

• G06F2218/10

  by analysing the shape of a waveform, e.g. extracting parameters relating to peaks · CPC title

• G06V20/40 Primary

  in video content (extracting overlay text G06V20/62; video retrieval G06F16/70; processing of video elementary streams in video servers H04N21/234; processing of video elementary streams in video clients H04N21/44) · CPC title

• G06K2209/25

  Physics · mapped topic

• G06K9/0053 Primary

  Physics · mapped topic

• G06K9/00711

  Physics · mapped topic