Point Cloud Completion Method Assisted by Projected Image

Image-guided point cloud completion task aims to utilize image information to address the uncertainties in point cloud completion inference. Although acquiring 2D image data is relatively simpler than 3D data, it is still ineffective in scenarios with occlusions where image data cannot be reliably obtained as a reference. Therefore, we propose a point cloud completion model assisted by projected image data, which addresses the limitations of acquiring 2D images by constructing projected images of the point cloud. Extensive experiments demonstrate that our proposed method enhances the quality of point cloud completion and outperforms other advanced methods.