The laser molten pool width detection technology of the closed-loop control system in metal parts laser forming process was researched. A method of applying Kalman filtering for molten pool width detection in laser forming process was presented. The vision sensing system collected the images of the molten pool in laser forming process. The molten pool width calculated by image processing was used as a parameter to establish a state equation and a measurement equation. A Kalman filter recursively computed the solution over least squares approaches to equations which were established based on an estimation of the width and the displacement in width. The Kalman filter optimized the estimation of the molten pool width at the next sampling time and reduced the measurement errors caused by the process and sensor noises. Experiment results show that measurement average errors decrease from 0.028 mm to 0.009 3 mm. Experiments demonstrate the effectiveness and great improvement of precision by using the proposed algorithm in the molten pool width detection.