With rising attention to machine learning and artificial intelligence, image classification problems have been investigated thoroughly for the past years. The present thesis studies especially texture classification in images. Descriptors like graph indices allow textural information encoding and have not jet been investigated closely in context with machine learning methods. Thus, I describe how respective descriptors are formed and encoded and investigate their performance, their applicability to image segmentation and significant parameters in more detail. Known machine learning models like Support Vector Machines (SVMs) or Convolutional Neural Networks (CNNs) are used.