LogCubic

class randd.model.LogCubic(r, d, d_measure, ndim)

1D log cubic rate-distortion function estimator.

Bitrate is converted into log scale according to the reference below. Extrapolation is automatically enabled, but not reliable. Used in estimation of BD-PSNR and BD-Rate.

Parameters

• r (NDArray) – Encoding representations.

• d (NDArray) – Corresponding distortions.

• d_measure (str) – Name of the distortion measure.

• ndim (int) – Number of dimensions of the RD function domain.

References

G. Bjøntegaard, “Calculation of average PSNR differences between rdcurves, Austin, TX, USA, Tech. Rep. VCEG-M33, ITU-T SG 16/Q6, 13th VCEG Meeting, Apr. 2001. G. Bjøntegaard, “Improvements of the BD-PSNR model,” Berlin, Germany, Tech. Rep. VCEG-AI11, ITU-T SG 16/Q6, 35th VCEG Meeting, Jul. 2008.

__call__(r)

Predict the distortion at the given representation.

Parameters

r (NDArray) – Input encoding representation.

Returns

Predicted distortion.

Return type

NDArray