In this paper, we present a new strategy for automatic classification of weld defects in radiographs based on Gentle Adaboost algorithm. Radiographic images were segmented and moment-based features were extracted and given as input to Gentle Adaboost classifier. The performance of our classification system is evaluated using hundreds of radiographic images. The classifier is trained to classify each defect pattern into one of four classes: Crack, Lack of penetration, Porosity, and Solid inclusion. The experimental results show that the Gentle Adaboost classifier is an efficient automatic weld defect classification algorithm and can achieve high accuracy and is faster than support vector machine (SVM) algorithm, for the tested data.