The article deals with a method of diagnosing GMA welding process, based on the analysis of fused infrared and vision images of a welding arc. The proposed method consists of a few fundamental steps such as acquisition, pre-processing, fusion, post-processing, analysis and recognition of acquired images. The recognition phase is based on diagnostic signals being waveforms of relevant features extracted from the fused images. The authors proposed original approaches to the analysis of fused images. The article describes each step of the proposed method in the context of the authors’ research, performed on the basis of sequences of infrared and vision images gathered during a series of laboratory experiments where different welding instabilities were simulated.