In the last decades laser beam welding (LBW) has outclassed older welding techniques in several manufacturing processes ranging from automobile production to precision mechanics. The huge improvement in welding technologies has been followed by the introduction of on-line monitoring systems based on photodiodes and/or CCD/CMOS cameras. Nevertheless, sophisticated equipment for the real time control of LBW processes is not commercially available yet, because photodiodes and camera based systems do not provide the required spatial and temporal resolution. In order to follow the high dynamics of LBW and to reach a sufficient robustness of the feedback system against physical fluctuations, controlling rates in the multi kilo Hertz range are necessary. This work has been focused on the implementation of high-speed Cellular Neural Network (CNN) based algorithms for feature extraction in coaxial process images. In particular, two image features have been considered, i.e. the full penetration hole (FPH) for the real time control of the laser power, and the occurrence of spatters for the on-line monitoring of the welding quality. CNN are networks of regularly spaced processing units called cells, which interact directly with neighbouring units and indirectly with the other cells because of propagation effects of the continuous-time dynamics of the network. The algorithms presented in this work have been implemented on a CNN based chip called QEye, which is included in the Anafocus’ Eye-RIS vision system (VS). The main advantage of this chip lies in performing simultaneously image sensing and processing. Such property combined with the strong real-time signal processing capabilities of CNN based computing structures makes the Eye-RIS VS a powerful development platform. The latter has been integrated in a closed-loop control system which led to the first real-time control of LBW processes at rates up to 14 kHz. In this work, besides the description of the control algorithms, experimental results obtained by the adoption of this control system in real-life LBW processes will be presented and discussed.