I lead a research team on the computer vision (CV) dedicated to nondestructive testing by radiography (RT). It consists here to detect and identify defects and discontinuities in the welded joints using digital image processing and analysis, statistical pattern recognition, artificial intelligence techniques, etc. Besides an important scientific impact thanks to the use of advanced and leading techniques in the CV area such as mixture models, active contours, Bayesian networks, SVM, ANN, CBIR and integral transforms, this project presents also professional, technological and economic interests. In fact, some tools employed will help the radiogram interpreter in the decision-making on the quality of the executed welds (taking, of course, into account the standards and norms which govern RT) because the human vision based interpretation may be erroneous due to weak sizes of some defects, bad contrast, noise, etc. In addition, for human interpretation, there are some subjective factors related among others to the interpreter’s experience and skill. Consequently, the developed computer vision techniques will permit unavoidably to the weld examination to be reproducible, fast and reliable.