Glaucoma is a major cause of irreversible blindness globally, underscoring the need for early and accurate diagnosis to prevent vision loss. This paper presents an automated glaucoma detection framework using colored fundus images, designed to overcome limitations such as uneven illumination, high computational cost, and overfitting found in earlier approaches. The preprocessing phase enhances retinal structures using Contrast Limited Adaptive Histogram Equalization (CLAHE) and green channel extraction. Texture features are extracted via the Gray-Level Co-occurrence Matrix (GLCM), and the most relevant features are selected using the Relief algorithm. The selected features are classified using three machine learning algorithms: Support Vector Machine (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost). Performance is evaluated using accuracy, precision, recall, and F1-score. SVM achieved an accuracy of 83.15%, precision of 80.17%, recall of 87.17%, and F1-score of 83.52%. RF obtained 84.38% accuracy, 78.77% precision, 93.25% recall, and 85.40% F1-score. XGBoost outperformed both, achieving the highest accuracy of 85.03%, along with 79.93% precision, 92.72% recall, and 85.85% F1-score. These results demonstrate that while SVM and RF excel in certain metrics, XGBoost provides the most balanced and robust performance, making it a promising tool for reliable glaucoma detection.