eduzhai > Applied Sciences > Engineering >

Improved Deep Point Cloud Geometry Compression

  • Save

... pages left unread,continue reading

Document pages: 6 pages

Abstract: Point clouds have been recognized as a crucial data structure for 3D contentand are essential in a number of applications such as virtual and mixedreality, autonomous driving, cultural heritage, etc. In this paper, we proposea set of contributions to improve deep point cloud compression, i.e.: using ascale hyperprior model for entropy coding; employing deeper transforms; adifferent balancing weight in the focal loss; optimal thresholding fordecoding; and sequential model training. In addition, we present an extensiveablation study on the impact of each of these factors, in order to provide abetter understanding about why they improve RD performance. An optimalcombination of the proposed improvements achieves BD-PSNR gains over G-PCCtrisoup and octree of 5.50 (6.48) dB and 6.84 (5.95) dB, respectively, whenusing the point-to-point (point-to-plane) metric. Code is available atthis https URL .

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×