Abstract: In order to address the problem of intense arc light interfering with welding imaging during the tungsten inert gas (TIG) welding process, an optical imaging system based on passive vision technology was proposed. The system effectively reduced the negative impact of strong arc light on the imaging of the welding process through spatiotemporal composite filtering technology, while retaining key welding feature information such as the edge contour of the weld pool and weld seam texture. The optical filtering system including an adjustable neutral density scrim, a wide bandpass filter, and a special high-pass spatial filter was designed. By selecting a suitable imaging band and filtering the arc light spectrum information in the frequency domain, the light intensity gradient in the imaging region was weakened to enhance the imaging quality of welding detail information. The imaging application effect of the system in TIG welding of narrow-gap steel plates was experimentally investigated, which significantly improved the clarity of welding imaging and the recognition of information features, and provided powerful support for online monitoring of welding quality.