We extracted or segmented weld defects from the digital image(X-ray) through image processing techniques. My question is, what are the features of weld defects can play important role in the weld defect classification module? I mean that, which features can be utilised in the classification for better result. I am using X-radiography digital image. Please suggest me any real time case studies or articles.

Note: Feature in the sense of pixel values, mean, standard deviation etc. of region of interest (weld defect).