the techniques employed and yet even when done
adequately interpretation errors occur whether it be the
nondetection of a defect present or the incorrect classifi-
cation of a detected defect.
At present there are several research centers around the
world working on the development of automatized systems
of inspection and analysis of radiographs using compu-
tational methods, aiming mainly at the classification of
defects. Over the last 15/20 years research has produced
various publications dealing in this matter, due to the
undeniable relevance and importance in this area. Some of
these publications will be discussed during this work and at
times compared to the work proposed here.
The automatic analysis system of radiographs for weld
beads follows the same steps as for casting piece radio-
graphs, that is the film digitalization stage, or direct digital
data acquisition (forgoing the film), pre-processing of
images, identification and classification of defects found
in the weld bead; a post-processing stage of the results
[1 –3] may also be carried out. The pattern classification
stage has been and is one of the most sort after goals in terms