The count of tumor patients is increasing day by day. Brain tumor, whose main cause is the uncontrolled division of the cells, if detected at an early stage, will help a lot in curing it. Various detection techniques are available for identifying the abnormality in the brain, but, MRI is a better technique in comparison to others. This paper presents a method for distinguishing the tumor affected images from the normal ones from the database. The procedure involves preprocessing, segmentation, feature extraction and classification. Ensemble based classification is done using neural network. The area of abnormal portion is also computed.