This paper deals with the realization of a visual monitoring system for the real time detection of spatters in laser beam welding (LBW). Spatters deteriorate the corrosion resistance and the aesthetics of the welding result. Therefore, the real time detection of spatters allows providing on-line quality information about the process, thus reducing material waste in production chains. The proposed Cellular Neural Network (CNN) based algorithm has been implemented in the Eye-RIS vision system (VS). Monitoring rates up to 15 kHz have been reached, allowing the integration of the spatter detection with the evaluation of additional image features, e.g. the full penetration hole (FPH).