engleski

Subjective and Objective Testing in Support of the JPEG Pleno Point Cloud Compression Activity

Point clouds have many applications in today’s society ranging from entertainment to autonomous driving. With these new applications comes the need to compress the growing volume of point cloud data in a manner that is both suitable for human visualization and machine processing applications. The JPEG Pleno Point Cloud activity has been working toward a learning-based coding standard for point clouds, offering a single-stream, compact compressed domain representation, supporting advanced flexible data access functionalities targeting both in-teractive human visualization, and effective performance for 3D processing and machine-related computer vision tasks. As part of this activity, the JPEG Committee has been performing a number of exploration studies to evaluate existing coding standards as well set up baseline anchors and examine objective metrics against which new learning-based solutions may be compared. This article provides an overview of the JPEG Pleno Point Cloud activity and discusses challenges and solutions to the problem of evaluating and comparing cloud coding solutions. Experimental results will be presented demonstrating methodologies used by the JPEG Committee for point cloud compression assessment as well as outlining the performance of current state of the art compression standards on point clouds as well as the sensitivity of the objective metrics used for this activity to various adjustable parameters.

JPEG ; point cloud ; compression ; machine learning

nije evidentirano