目前已有的点云超分辨率方法只是利用点云的几何信息对点云的坐标进行重建，没有考虑到与几何结构相关的颜色属性并对颜色信息进行超分辨率。本文联合使用点云的几何信息和颜色信息，通过双流的图卷积网络同时重建出点云的坐标和颜色，并且使用基于图卷积的判别网络来提高重建点云的质量。在图卷积网络中，本文使用多属性联合的图卷积，将多属性特征相似的点构成局部图，增强了局部图中节点的关联性，扩展局部图卷积操作的上下文信息范围。本文还提出了基于结构的局部-全局几何约束和基于几何信息的点云颜色约束，有效地提高了点云坐标和颜色的重建质量。

The existing point cloud super-resolution methods only used the geometric information to reconstruct the coordinates of the point cloud, without considering the color attributes associated with the geometric structure and super-resolving the color information. In this paper, the geometric and color information of the point cloud are jointly used to reconstruct the coordinates and color of the point cloud simultaneously by a dual-stream graph convolution network, and a discriminative network based on the graph convolution is used to improve the quality of the reconstructed point cloud. In the graph convolution network, we use multi-attribute joint graph convolution to form a local graph of points with similar multi-attribute features, thus enhancing the correlation of nodes in the local graphs and extending the scope of contextual information for local graph convolution operations. We also propose multi-level geometric constraints and geometric information-based point cloud color constraints, which effectively improve the reconstruction quality of coordinates and colors.