Abstract:

In this book chapter, we present how graphs and graph signal processing contribute from low-level to high-level point cloud processing and understanding. The processing of point clouds addresses denoising and enhancement, as well as resampling techniques. The analysis of point clouds considers segmentation, classification, and recognition (e.g., skeleton-based action recognition). These are just some of the main areas of work related to GSP-based point cloud processing, which are unique relative to the use of GSP techniques for other data in terms of the signal model and processing methods. In practice, GSP has already demonstrated great benefits for processing point clouds and the main purpose of this chapter is to bring the underlying concepts together within a common framework.