<p>Age-related macular degeneration (AMD) is a leading cause of vision loss in older adults. Detecting AMD early can prevent the irreversible damage caused in later stages. Most existing methods detect drusens as a preliminary step for AMD detection, which is quite challenging as the drusens are present alongside other exudates in the retina. This paper presents a system for early diagnosis of dry AMD that does not rely on detecting drusens and integrates handcrafted features with machine learning and image processing techniques. The system performs several image processing tasks, including pre-processing a color fundus photograph of the retina, macula detection, Region-of-Interest (RoI) selection around the macula, and feature extraction from the RoI. Several texture and color features of the macula region are extracted and analyzed using t-test and ReliefF feature selection algorithms. Supervised classification techniques such as Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Naïve Bayes Classifier (NBC), and Multi-Layer Perceptron (MLP) are trained on the selected features with stratified 10-fold cross-validation to classify retinal images as normal or AMD. The system’s performance is evaluated based on accuracy, error, recall, specificity, precision, and F1-score. Classifiers are trained and tested on three feature sets: texture, color, and a combination of both. The proposed system achieves excellent results with texture features and the SVM classifier, attaining accuracies of 98.89% and 95.43% on the STARE and ODIR datasets, respectively.</p>