Image Texture Classification Using Combined Grey Level Co-occurrence Probabilities

In this paper, we present an image texture classification method combining Grey Level Co-occurrence Probabilities (GLCP) with Support Vector Machines. Texture represents properties characterizing an object's surface or structure, defined as patterns composed of interrelated elements. The primary focus of this research is texture segmentation and classification of digital texture images. We employ the GLCP method for feature extraction from texture images, which calculates probability distributions of pixel intensity pairs at specified spatial relationships (typically implemented using MATLAB's graycomatrix function with parameters like distance and orientation). The extracted features are then classified using Gaussian Support Vector Machines (GSVM), where the kernel function maps features into higher-dimensional space for optimal hyperplane separation (implementable through scikit-learn's SVC class with RBF kernel). For validation, we selected the widely-used Brodatz texture album to test performance. Our experiments demonstrate that the GLCP-GSVM combination achieves significantly better classification accuracy compared to using GLCP features with conventional classifiers. To further enhance accuracy, subsequent research could explore alternative feature extraction methods (such as Local Binary Patterns or Gabor filters) and advanced classification algorithms (like ensemble methods or deep learning architectures), potentially integrated through Python's OpenCV and TensorFlow libraries for optimized implementation.