Wavelet Based feature extraction and Classification of MR Images

Abstract

Feature extraction is very important for the classification of images. There are various feature extraction methods and their applications, but wavelet based feature extraction is the most efficient method. Generally, in classification,features are extracted and then these features are reduced up to a certain level to get maximum information with minimum features. Classification of medical images plays an important role to identify the abnormalities present in the images. Wavelet based feature extraction is very efficient and famous now a days because it extracts useful information and properties from medical images in a better way. To achieve higher classification accuracy,it is important to extract useful content from the image in the original form. Classification of MR Images has been done in this paper. With the help of this algorithm we can differentiate brain MR Images in to classes normal and abnormal. Features are extracted with the help of different wavelets db4, bior5.5 and sym4, total 26 features are extracted with the help of 4-level DWT. Total 120 MR images has been used for this work where, 80 images are used as training images and 40 are used as testing images. Features are reduced with the help of PCA (Principle Component Analysis). Classification is done with the help of SVM with linear kernel. Results are compared in terms of classification accuracy and processing time of three wavelets to identify the best one.