Abstract: Basing on visual sensing method, the electron beam welding molten pool image of titanium alloy was obtained. The binary morphological image processing algorithm was used to process the molten pool image for getting the molten pool edge, moreover, the molten pool width was obtained by using a suite of processing programs for molten pool extraction. The correspondence relationship between the fluctuation of molten pool width and weld appearance is found, i. e. the weld appearance is poor when the fluctuation of molten pool width is relatively large, but the weld appearance is good when the fluctuation of molten pool width is relatively small. Based on analysis, the coefficient of variation CV was proposed as characterization to describe the weld appearance. The results show that the developed image processing algorithms are reliable, and the detected errors are less than 0.1 mm, at the same time the whole extraction process of molten pool width only needs approximately 30 milliseconds, which could meet the needs of real time detection. The coefficient of variation can reflect the quality of weld appearance and it can be used as a parameter to control the weld appearance.