Abstract: Because the welding image has too much gray level and information and serious interference from the arc, splash and others which reduces welding image contrast and blurs some image details, a newly-improved Otsu algorithm was proposedto accurately extract weld seam region from welding image.By replacing the mean value with variance, the algorithm was foundedto improve the image segmentation quality and reduce processing time, and the interclass variance and intraclass variance were also taken into considerations. The experimental results show that the proposed algorithm achieves good welding image segmentation effect with the less time. Compared with the other algorithms, it has obvious advantage and it is an effective and efficient algorithmof welding image segmentation.