With the rapid development of 3D technology, the digitization of cultural relics has emerged as a highly scientific preservation measure. By capturing three-dimensional surface information of cultural artifacts, it provides essential data support for artifact restoration, reconstruction, virtual exhibition, and innovation. Given the intricate structures and the delicate, untouchable nature of these artifacts, existing methods can no longer meet the diverse needs of cultural relic digitization. Further refinement and enhancement of point cloud data processing techniques for artifacts with varying structures and materials are required. The construction of a 3D scanning platform enabled the collection of data from various samples. An algorithm for multi-scale point cloud fusion and denoising, tailored for cultural relic digitization, was introduced. This algorithm categorizes noise point clouds into large-scale and small-scale based on distance, utilizes a combination of multiple algorithms to remove large-scale noise, and employs improved bilateral filtering with normal vector correction to eliminate small-scale noise. Experimental validation was carried out using the bunny model.