The geometric appearance of weld is very important for the evaluations of welded joint performance and weld quality during oscillating laser welding (OLW). To represent the characteristics of weld appearance more accurately, the surface reconstruction method is often adopted to obtain the complete geometric appearance of weld. The efficiency of surface reconstruction method is crucial for practical application, especially for processing large-scale point cloud. An improved octree-based surface reconstruction method is proposed to enhance the computational efficiency for complex weld appearance based on the point cloud. The large-scale point cloud of weld which is transformed from numerical simulation of OLW is preprocessed by octree structure to reduce the number of points and the Poisson surface reconstruction (PSR) method is utilized to reconstruct the three-dimensional profile of weld appearance. The reconstruction accuracy is quantified by contrasting the cloud-to-mesh distances between the raw point cloud and the reconstructed mesh model. Furthermore, the reconstruction efficiency and accuracy of the proposed method are compared with those of PSR method and the reconstruction quality of proposed method with octree structure is compared with those of methods with other sampling strategies. The results demonstrate that the reconstruction efficiency of the proposed method is significantly increased with excellent reconstruction accuracy. The improved octree-based surface reconstruction method is of great importance for analyzing weld appearance characteristics and evaluating weld quality.