It has been presented the issues concerning the application of images fusion registered in visible and infrared radiation for diagnostics of MIG/MAG processes. The sequences of visible and infrared images visualizing the arc burning during performing the weld were examined. Recorded during experiments image sequences were subjected to fusion and next to analysis in order to obtain diagnostic signals. On the basis of selected features of diagnostic signals it was carried out the classification of the welding process state and its results were compared with those obtained for diagnostic signal features determined independently for visible and infrared images. The results show that the application of images fusion enables to identify effectively various welding imperfections forming in welding processes.