Magnetic Resonance Images has become a widely used method of high quality medical imaging system, especially in brain imaging, where the soft-tissue contrast and non invasiveness is a clear advantage. Medical image classification is a pattern recognition technique in which different images are categorized into several groups based on some similarity measure. One of the significant applications is the tumor type identification in abnormal MRI brain images. The proposed system comprises feature extraction and classification. In feature extraction, some specific features are extracted using texture as well from intensity using modified Micro Structure Descriptors. The hybrid RBF kernel is designed in the classification stage and applied to training support vector machine to perform automatic detection of tumor in MRI images. The accuracy level (94%) for our proposed approach is proved at detecting the tumors in the brain MRI images. The obtained results depict that the proposed brain tumor detection approach produces better results in terms of k-fold cross validation method.