Bag Of Words

Step 1: Extract features of all images and cluster them into k-clusters

 Step 2: Make a histogram for each training image by matching the feature vectors with the k-clusters. This histogram will be the input to the SVM.

StackOverFlow Link - http://stackoverflow.com/questions/11602577/bag-of-words-training-and-testing-opencv-matlab

*Image Source - Jatin(Friend and MS graduate, CSU)