k-medoids

On matrix modification of clarans clustering method in large video surveillance databases

Clustering algorithms for Very Large Data Bases (VLDB) are observed in application with image and video processing. Such a specific case requires initial data presentation as multidimen-sional vectors. That is why matrix modifications of traditional k-medoids, Partitioning Around Medoids, Clustering LARge Applications and CLARA based on RANdomized Search methods are proposed. Benefits and drawbacks of them all are examined.