Classification of brain MR images using wavelets texture features and k-Means classfier

In this paper we deal with the problem of classification of brain MR images as normal or abnormal to assist in clinical diagnosis. The proposed method use wavelets to decompose the input image into the approximate and detailed components and extracts of texture features using gray level co-occurrence matrix at three levels of image resolution. Euclidean distance is measured between the feature vectors of test MR image and reference MR image. These distances are further fed to k-Means classifier to classify the MR images as normal and abnormal images.