Rate-distortion curves and Bjøntegaard delta rate
One encode is one point: so many bits, so much quality. Change a setting and both numbers move, usually in the same direction, so two encodes rarely settle which encoder is better. A rate-distortion curve joins several encodes of the same source, and the Bjøntegaard delta rate (BD-rate) reduces the gap between two such curves to one percentage.
Why one point is not enough
The fix is to encode each side several times, at different quantizers or bitrates, and compare the curves at equal quality: how many more or fewer bits does B need to reach the quality A reached?
Compare two encoders over a range
Encoder A codes the cat with intra prediction and rounds each coefficient to the nearest level. Encoder B differs in one thing, which you choose. Each is run at four values of the quantization parameter (QP). Quality is peak signal-to-noise ratio (PSNR), and bits are modeled, not written. The shaded band is the range of quality both curves cover; only that range counts. The two curves lie close together, so a second plot shows the gap between them at each quality of the band. BD-rate is stated for B relative to A: −5% means B needs 5% fewer bits than A, averaged in log rate over the band, and a positive number means more. The two gap readouts are the same comparison at the band's lower and upper edge.
The dashed line averages differences in log rate, so it is not the ordinary average height of the teal curve.
| QP | A bits | A dB | B bits | B dB |
|---|---|---|---|---|
With a dead zone of offset 1/3 at QPs 22 to 37:
- No single point settles it. At every QP encoder B used fewer bits and scored lower: at QP 27, 1918 bits and 36.77 dB against 2080.5 and 37.97.
- Over these four points B needs 0.9% fewer bits. The gap is −3.3% at the lower edge of the band and +0.2% at the upper edge, so the saving is at low quality.
- Other points, another answer. At QPs 28 to 43 the same dead zone is −4.3%, and at 16 to 31 it is −0.5%. Offset 1/6 is +6.3% here and −0.2% at the coarse end, where its two edges read −1.3% and +8.3%.
- Dropping prediction raises modeled bits at nearly the same PSNR. At QP 27: 2297 modeled bits against 2080.5, and 37.98 dB against 37.97. Over the shared range its BD-rate is +11.6%.
How the number is computed
BD-rate, percent = 100 × (10^D − 1)
D = average, over the shared range
of quality, of
log10(rate of B) − log10(rate of A)
The average is taken in log rate, so the result is the geometric mean of the rate ratio, less one. A ratio of 0.8 over half the band and 1.25 over the other half gives 0%, though the two gaps, −20% and +25%, average to +2.5%. Direction matters too: if B against A is −10%, A against B over the same band is +11.1%.
The page's script follows the calculator in Netflix's repository: at least four points a curve, quality that rises with rate, and only the range of quality the two curves share.
The calculator's four steps
The steps below are those of the calculator in the repository of Netflix's Video Multi-Method Assessment Fusion metric. Its comments say it was validated against a document named JCTVC-E137, which was not read for this page.
- At least four points per curve, sorted by rate, duplicates removed.
- Quality must rise with rate. A curve where a higher rate scored the same or lower is rejected, not repaired. Look for a saturated metric or an anomalous encode; do not drop a point only to get a number.
- Each curve becomes log10 of rate as a function of quality, joined between the points by shape-preserving piecewise cubic Hermite interpolation, so the curve does not overshoot;
pchipendin the source sets the end slopes. - Only the shared range is integrated. If the two curves share no quality, there is no result.
Trade-offs
- One number hides where the gap is. The two edge readouts can differ in size and in sign.
- It belongs to the points chosen. Move the QPs and the same pair of encoders gives another figure.
- It belongs to the metric. A BD-rate in PSNR says nothing about another metric, and nothing here measures a viewer.
- It needs curves that rise. Plot the rate that rate control achieved, not the target it was given. After sorting by rate this calculator requires quality to rise strictly, so a metric that saturates, or scores that fall out of order, give no result.
Sources
Read 2026-11-15 and 2026-11-16, at the commits linked.
- Netflix's VMAF repository at commit a6ed0f1: in
bd_rate_calculator.pyunderpython/vmaf/tools,CalcBDRatewith the four-point minimum, the rising-order test, the shared range, the no-overlap case and the result;bdrint, which integrates the cubic pieces; the note on what it was validated against; ininterpolation_utils.pybeside it,computeParamsForSegmentsand the end slopes. The page's script ports the arithmetic and the input checks of these two files, dropping repeated points and rejecting a rate of zero included; it is not a tested replacement for them. Identical curves give 0, a curve with every rate cut by 10% gives −10%, and the demo's twelve settings agree with the calculator, run at this commit on 2026-11-16, to within 1e-12. - x264 at commit c24e06c (GitHub mirror): the closing log line in
encoder/encoder.c, which prints PSNR and bitrate for one encode, one point of such a curve. - Not checked: Bjøntegaard's own document and the document the calculator names (JCTVC-E137), neither of which was read; how other tools interpolate; BD-rate on any real encoder or sequence; any metric other than PSNR.