Revision history [back]

maybe you could look at opencv's bag of words implementation

you could do it like this:

   Mat histogram(1, num_clusters, CV_32F, 0.0f); 
   // * extract brisk features from your image, then
   // * match them to your kmedian clusters
   for (each match)
       histogram(matched_cluster_id) += 1;
   normalize(histogram, histogram);

maybe you could look at opencv's bag of words implementation

you could do it like this:

   // this will be one row of your traindata.
   Mat histogram(1, num_clusters, CV_32F, 0.0f); 
   // * extract brisk features from your image, then
   // * match them to your kmedian clusters
   for (each match)
       histogram(matched_cluster_id) += 1;
   normalize(histogram, histogram);