/bjontegaard_metrics

Bjontegaard metric computation in the python language

Primary LanguagePythonMIT LicenseMIT

Bjontegaard metric computation

Introduction

Bjontegaard's metric computes the average gain in PSNR or the average saving in bitrate (in %) between two rate-distortion curves [1][2].

Similar to other packages [3] this function allows to compute the metric for more than 4 RD points, but in this case for the python language.

Usage

Check the test.py for some simple tests.

from bj_delta import *
bd_psnr = bj_delta(Rate1, PSNR1, Rate2, PSNR2, mode=0))
bd_rate = bj_delta(Rate1, PSNR1, Rate2, PSNR2, mode=1))
  • Rate1,PSNR1: RD points for curve 1.
  • Rate2,PSNR2: RD points for curve 2.
  • Mode: 0 for average PSNR difference and 1 for average bitrate savings (%).
  • Returns the calculated Bjontegaard metric (BD-Rate or BD-PSNR).

References

[1] G. Bjontegaard, "Calculation of average PSNR differences between RD-curves", VCEG-M33, Austin, TX, USA, April 2001.
[2] S. Pateux, J. Jung, "An excel add-in for computing Bjontegaard metric and its evolution", VCEG-AE07, Marrakech, MA, January 2007.
[3] G. Valenzise, https://www.mathworks.com/matlabcentral/fileexchange/27798-bjontegaard-metric.