This paper presents a methodology to detect the central region of weld beads on petroleum pipelines in double wall double image (DWDI) radiographic images. The method is based on three steps: pre-processing (to isolate selected regions), optimization (to define the ellipse that best fits in selected region), and decision (to choose the best region). Results show that the Particle Swarm Optimization (PSO) algorithm converges satisfactorily to the selection of the region that is most similar to the central region of the weld on the optimization and decision steps (to balance the weights of the classifier). The scientific contribution of this research is the improvement of the method applied in the search of candidate regions through ellipses' attributes analysis.