Hi all

I have a group of jpeg images(frames obtained from a video).I need to cluster the similar images based on color similarity using agglomerative hierarchical clustering using euclidean distance as distance metric.can anyone provide me with some idea as how to proceed.

thanks in advance

kokila

It is a very interesting subject. As you may know, cluster analysis has strong data-dependency.
Off the top of my head, first (step1), I will create a distance matrix between all images. Secondly (step2), a pair having the smallest distance will be merged as a cluster node and the center of newly assigned cluster will be updated. Thirdly (step3), recreate a distance matrix from updated cluster node and go to step 1. This may be called Forgy's algorithm.

Hi

Thanks Seung Yeon

kindly bare with me

Please can you explain me as to how it can be done practically based on color and euclidean distance.