Schizophrenia (SZ) is a brain disorder with high financial and social burden worldwide. The majority of morphometric analyses have focused on volumetric measurements derived from brain MR images. However, there is significant variation in the measurements due to heterogeneity of SZ. There is a need to study the links between MR image textures in Schizophrenic and normal images. In this work, the specific pattern changes in Schizophrenic MR images are better represented by texture features such as Hu moments, gray level co-occurrence matrix (GLCM), Zernike moments and structure tensor. The SZ subjects are differentiated from healthy subjects using binary particle swam optimization (BPSO) based fuzzy SVM (FSVM) classifier with mutual information quotient as an objective function for feature selection. Initially the MR brain images are skull-stripped using non- parametric region based active contour. The efficacy of the proposed work is analyzed with different objective functions and compared with BPSO based SVM classifier. This work is evaluated on National Alliance for Medical Image Computing (NAMIC) database. The results show that the proposed method is able to skull-strip the brain region with better similarity values compared to different tool based methods. BPSO-FSVM along with Hu moments, GLCM and structure tensor could classify the normal and SZ better with an accuracy of 90% compared to BPSO-SVM. The AUC is 0.9 for BPSO-FSVM. Thus BPSO optimized features highlights significant texture pattern change in Schizophrenia. Hence this frame work could be used to aid disease prognosis, treatment and study the neuropsychiatric disorder.