FULLY AUTOMATED APPROACH TO IDENTIFY BRAIN TUMORS USING DYNAMIC REGION MERGING

: Image segmentation is a fundamental task of computer vision application. In this paper address the automatic image segmentation algorithum, which exploit the information obtain from deducting tumor in brain MRI. Many segmentation technic are used to deduct the tumor portion like Thresholding technic, Hybrid technic , Boundary based technic, and Region based technic but in this all technic should deducted the initial tumor tissue but when the secondary tumor present or more than one tumor are present in same brain image it is difficult to deducted the irregular tissue from the regular surrounding tissue to get real identification of involved and non involved area which help the surgeon to distinguish the total effected area covered by (primary & secondary ) tumor portion. The proposed method is DRM(dynamic region growing)to deducted the tumor boundary in region merging style .In this paper two essential issue in a region merging algorithms : order of merging and stopping criteria.in the proposed method this two issued are solved by a novel predicates ,which is defined by a sequential probability ratio test (SPRT)and minimum cost criterion. We also proved that the producedsegmentation satisfied certain global properties