Performance Comparison of Partition Based Clustering Algorithms on Iris Image Preprocessing

Today's life entirely depends on information, and security in information system is an essential term. Now a days various biometric feature like fingerprint, gait, iris, face etc. are used to secure any system and it is more powerful to use biometric feature instead of using other traditional techniques like password, PIN number etc. In automated personal identification system iris recognition technique is the most reliable authentication technique and iris image segmentation step is important to acquire good accuracy in this technique. But noisy image decrease the accuracy and most of the errors occur in non-iris region. So it is better to avoid segmentation errors by excluding non-iris regions from iris image. On the other hand cluster analysis one of the data mining concepts, is very useful for finding similar groups from a data set. So cluster analysis can be used for finding relevant groups from iris image. Basically k-means clustering algorithm used on segmentation step. In this thesis work another two clustering algorithms has been used for iris image preprocessing on segmentation step to cluster most similar objects together so that non-iris region can be reduced and compare their performance with k-means algorithm to find out the better one.