International Journal of Artificial Intelligence and Neural Networks
Author(s) : ENTESAR B. TALAL , KHAWLAH H. ALI
this paper explores statistical features of texture based image descriptors that make use of the spatial gray level of bag of visual words model to efficiently improve classification performance for two types of spine MRI images, which is normal and abnormal(which is may be a cancer). At first, texture is characterized through second order statistical measurements based on the gray-level co-occurrence matrix introduced by Haralick. By this method it is possible to compute, four features which are designed to perform texture: contrast, correlation, homogeneity, and energy for spine MRI images, and then construct a bag of visual word (BoW) to encode feature vector into visual words. Features of these types are used to classify two categories of spine MRI image: normal and abnormal, then, classify them by using SVM, which is works efficiently. Experiment results on spine MRI show significant improvement of classification.