Generating useful labels to help identify events in photo collections

GENERATING USEFUL LABELS TO HELP IDENTIFY EVENTS IN PHOTO COLLECTIONS Disclosed is a method of labelling a cluster of images. Each image in the cluster is 5 associated with location data identifying a capture location at which the image was captured. The method determines a representative location (centroid Cp, or D in Fig. 11) for the image cluster (PI, P2, ...) based on at least the location data associated with the images in the cluster, and determines a set of geographic points (D) within a first threshold distance of the representative location. A first geographic point is selected from the set of 10 geographic points based on a distance between the selected first geographic point and the representative location. The method removes from the set of geographic points those geographic points (e.g. D6, D7) beyond a determined radius (250, 350) from the representative location so as to determine a reduced set of geographic points, the determined radius preferably being a sum of a predetermined distance and a calculated 15 distance. The method then identifies one or more geographic points in the reduced set of geographic points that are photographically significant and applies a first geographic label to the cluster of images where the reduced set of geographic points consists of one remaining photographically significant geographic point, the geographic label being associated with the remaining geographic point. The method then applies an alternate 20 geographic label to the cluster of images where the further reduced set of geographic points comprises two or more photographically significant geographic points. 4939185_1 991852_specilodge Identify data points forming a cluster Str| Calculate Representative -502 501 Location (RL) of the cluster Find all Landmarks within (100m) of RL andrnarks No Expand Sac 504 yes 550 Find closest Landmark 510 Is the closest Use the closest Landmark closer by 50m or Landmark for more than all other Landmarks Irepresentative within 1 00m of the Yes text (label) cluster? No Form a shortlist of closest Landmarks and 515 all Landmarks within 1 00m of the cluster Remove all Landmarks which are not of 516 photographic significance from the shortlist Replace Number of Landmarks Landmarks in in shortlist Yes shortlist = 0? 525 No ~ 517 Number of Use the one data 520 Landmarks in point in current hortlist = 1? Yes shortlist for representative text 519 NoII Ran Lndmrk inshrtlstin ers o p--530 ->~~~ R an La d a k in sh rls int r s ofp