A new error metric for geometric shape distortion using depth values from orthographic projections

  • Authors:
  • Maja Krivokuća;Burkhard C. Wünsche;Waleed Abdulla

  • Affiliations:
  • The University of Auckland, New Zealand;The University of Auckland, New Zealand;The University of Auckland, New Zealand

  • Venue:
  • Proceedings of the 27th Conference on Image and Vision Computing New Zealand
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

Euclidean-based distance metrics are commonly used for measuring geometric shape distortions of 3D models, but have several drawbacks. They have strict requirements regarding model representation and usually necessitate expensive surface sampling or point correspondence matching. Furthermore, many distortion metrics have been designed to capture relatively minor shape changes, rather than large-scale (global) shape distortions which can occur during compression and image-based reconstruction. This paper presents a new geometric distortion metric for offline quality assessment of 3D models. The new metric is largely independent of object representation and does not require any expensive surface sampling or point matching operations. It can capture well both minor and severe shape distortions, on a large and small scale. In the context of measuring the rate-distortion performance of a lossy mesh compression algorithm, the new metric provides a more reliable measure of overall shape distortion than the commonly used Hausdorff distance and a more relevant measure of surface error than the RMSE. Visual distortion maps for the new metric are also created, which indicate that the metric additionally captures well the perceived shape error between two objects.