This study presents a novel approach for classifying lung diseases from chest X-ray (CXR) images, employing a two-phase methodology: feature extraction and classification. The feature extraction phase utilizes a comprehensive set of features, including texture, geometric, and gradient features, to capture the diverse characteristics of lung images. For classification, an ensemble of Support Vector Machines (SVM) and Extreme Learning Machine (ELM) is used, leveraging the SVM’s capability for robust separation in high-dimensional spaces and the ELM’s rapid learning and generalization strengths. SVM classifies the input images into two classes they are Normal and diseased and the ELM again classifies the diseased image into thee classes namely COVID-19, Tuberculosis, and Pneumonia. Simulation results on a voluntarily collected dataset demonstrate the superior performance of this approach, effectively classifying various diseases such as Tuberculosis, Pneumonia, COVID-19, as well as normal lung conditions. The results highlight the method's effectiveness in improving diagnostic precision and reliability in lung disease classification from CXR images.