Brain forms a significant part of the human body but has a complex structure which makes it difficult to examine the brain abnormalities. Brain tumor is one of the significant diseases and various technologies are used to detect and diagnose brain tumor by non-invasive method. Magnetic Resonance Images being one of the best is directly being analyzed by the doctors for the diagnosis of tumor which is time-consuming. Automatic detection and classification of tumor makes diagnosis easy for the doctors and can provide a valuable outlook of earlier methods. The proposed method has been implemented in two different approaches, for the analysis of MR images to identify the tumor grade (Low/High). The first approach gave results of 92.5% accuracy which was found to be better than the second approach of 87.5% accuracy. This design is easy to use and the results ensure that it is efficient, and satisfying for quick detection of the grade of tumor. © 2017 IEEE.