查看文章

Feature plays an important role in the processing of medical images. The different features of an image include color, texture, shape or domain specific features. Texture is considered as one of the important feature of an image. In this paper, the global and local texture feature extraction is done using different algorithms. The global texture features for an image such as energy, entropy, homogeneity, correlation, contrast, dissimilarity, maximum probability are computed using Gray level co-occurrence matrix (GLCM). The local texture features for an image is extracted using a texture feature descriptor named Speeded Up Robust Features (SURF). The performance of feature extraction is based on the classification results. The process of classification is done using Support vector machine (SVM) and KNN classifier. The performance is evaluated on the basis of different metrics namely sensitivity, specificity, accuracy, precision and F1 score. The experimental result shows that the local texture feature extracted using SURF performs best when compared to global feature extraction (GLCM) and also with other descriptors such as Scale Invariant Feature Transform (SIFT). The SURF local feature descriptor along with SVM-RBF classifier provides better classification accuracy.