In the field of medicine, Tumor Segmentation from MR images is a laborious and time-consuming process conducted by medical experts. Based on image registration technique and classification approaches, this paper offers an fully automatic and efficient technique for identifying as well as Segmenting brain tumors. As a common procedure to alter the characteristics of MRI images before feature extraction, we perform intensity normalization and contourlet transform as part of our feature extraction process. In this paper we propose Pyramid transforms in place of traditional Laplacian pyramids as part of our modified contourlet transformation to enhance the effectiveness of the proposed approach. Using genetic algorithm, the extracted features are optimized, and a neuro fuzzy inference scheme classification algorithm is applied to brain MR images for classification of feature for tumor region identification. Quantitative analysis is carried out on the proposed approach to calculate parameters such as specificity, segmentation accuracy, sensitivity, precision and dice similarity coefficient.