Abstarct: Autism is a brain disorder in children from different parts of the brain in people who are problems communicating with each other. The main causes of the disease are not completely defined and there is no cure for that. However, early treatment of the disease, its symptoms and improved social skills and verbal child is raised or increased severity of symptoms can be avoided. By doing research in this area suggests a place that does not cure autism, but it is timely rehabilitation actions of these patients by doing research in this area suggests a place that does not cure autism, but it is timely rehabilitation actions of these patients some of them can go to regular schools and even of higher education. Diagnosis the symptoms of autism at birth only a very difficult and even impossible. So, for better diagnosis of the brain data is used. A type of brain images, magnetic resonance imaging (MRI). This data allows us to examine all the details of the brain. Using these data, we can hope that clinical symptoms of autism better than only the prescribed discerning. In this thesis, a general frxamework for classifying images in both healthy subjects and patients autistic using MRI data is presented. In this context, after pre-processing images MRI, using algorithms PCA, in addition to data transfer to a new space, they also reduce the number of dimensions. Output from applying PCA on the data, are used as features. Then, to identify the characteristics of the higher resolution of the statistical T-test method used. After selecting features high resolution, using the clustering methods data in one of the two healthy or ill put. Other way to classify the MRI images in addition to data processing, image segmentation operation is performed. This action image of the brain is divided into two parts white and gray matter. In continuation of the gray matter data used for extraction and feature selection and classification. In this study of used multiple classifier to classify different data. The classification accuracy using the first method (using the whole picture) 76% and in the second method (using the gray matter) 94%.