Automated indexing for distributing event photography
The present method relates to the automated indexing of event images for distribution. The automated indexing can use automated facial recognition to determine which people are in each image. The images indexed in this fashion can be presented in a gallery, ordered by characteristics of the people in the images such as their name or room number, so as to facilitate the selection of the images by the people. The identification of the people in the images can be assisted by security or other information regarding the people that may be available to the event manager. Furthermore, the closeness of the relationships of two people can be inferred from the degree to which the people are in the same images, allowing the people in the images to be placed into groups, which can be hierarchical and/or overlapping, and which can assist in the organization of images being presented to the people, either in a gallery or electronic display format.
1. A method for associating persons, who are represented in a collection of images, into a group, using an automated facial recognition engine, comprising:
Identifying, using the facial recognition engine, a first and a second person in each image in the collection of images using the automated facial recognition engine;
Determining, using the facial recognition engine, a first number comprising the number of images in which the first person is present;
Determining, using the facial recognition engine, a second number comprising the number of images in which both the first person and the second person are present;
Computing, using the facial recognition engine, an association index between a first person and a second person from the ratio of the second number to the first number;
Grouping, using the facial recognition engine, the first person and the second person in a group if the association index between the first and second person exceeds a certain predetermined threshold value.
2. The method of claim 1 wherein the ratio of the first and the second number is raised to a predetermined power.
3. The method of claim 1 wherein computing further comprises using the facial recognition engine to compute a second association index between a first person and a second person from a third number being the number of images in which the second person is represented and from the second number.
4. The method of claim 1 wherein computing further comprises using the facial recognition engine to combine the association index between the first person and a third person with the association index between the third person and the second person.
5. The method of claim 4 wherein combining comprises using the facial recognition engine to multiply a term related to the association index between the first person and the second person with a term related to the association index between the second person and the third person.
6. The method of claim 4 wherein computing further comprises using the facial recognition engine to combine the association indices of multiple chains of association between the first person and the second person.
7. The method of claim 1 further including allowing the same access of images of the first person to the second person as if the images were of the second person, if the first person and the second person are grouped in the same group.
8. The method of claim 1 further including disallowing access of images of the first person to the second person, if the first person and the second person are not grouped in the same group.
9. The method of claim 1 further including clustering a first group comprising the first person with a second group comprising the second person by using the facial recognition engine to increase the threshold value until the threshold value exceeds the association index between a person in the first group and a person in the second group.
10. The method of claim 1 , comprising:
calculating, using the association index between the current members of the group and an additional person; and
admitting the additional person to the group if the association index exceeds a threshold.
11. The method of claim 10 wherein calculating comprises a ratio of a measure of the number of times persons in the group are represented in the same image to a power of the number of persons in the group, wherein the power is greater than zero and less than or equal to 1.
12. The method of claim 10 wherein the threshold is the association of the group without the additional person.