As a pivotal component in modern technological domains such as Autonomous drivingautonomous driving and Virtual realityvirtual reality, point cloud data necessitate enhancement to ensure that the quality meets the demands of downstream tasks. Point cloud enhancement methods delve into postprocessing techniques, including upsampling, frame interpolation, completion, and the removal of compression artifacts, addressing prevalent issues in traditional point cloud intelligent systems. This chapter highlights several deep-learning-based point cloud upsampling methods, such as the pioneering PUNet, the progressive point cloud upsampling network, the GAN-based upsampling method PUGAN, and Semantic Point Cloud Upsampling (SPU). These methods leverage the power of deep neural networks to restore geometric information and generate dense point clouds from sparse data. The SPU method, in particular, emphasizes alignment with downstream tasks by using classification networks to supervise the learning process. Additionally, this chapter explores point cloud frame interpolation techniques, which are crucial for generating intermediate point cloud frames along the temporal dimension, thereby enhancing the frame rate of LiDAR point clouds for applications like autonomous vehicles. Finally, a discussion on the challenges and future directions of point cloud upsampling and point cloud frame interpolation is provided.