Compression performance¶
Reproduce our current best results¶
We provide already encoded files as bitstreams in results/<configuration>/<dataset_name>/
<configuration>
can beimage
<dataset_name>
can bekodak, clic20-pro-valid, jvet
.
These bitstreams allow to reproduce the results stated above qnd reflect the current best rate-distortion results achievable with Cool-chic.
For each dataset, a script decodes all the bitstreams.
(venv) python results/decode_one_dataset.py <configuration> <dataset_name> # Can take a few minutes
The file results/<configuration>/<dataset_name>/results.tsv
provides the results that should be obtained.
Image compression¶
Image compression performance are presented on the kodak dataset, clic20-pro-valid and jvet datasets.
The following table sums up the compression performance obtained by Cool-chic 3.4 against different anchors.
BD-rate of Cool-chic 3.4 vs. [%] | Avg. decoder complexity | |||||||
---|---|---|---|---|---|---|---|---|
Cheng | ELIC | Cool-chic 3.3 | C3 | HEVC (HM 16) | VVC (VTM 19) | MAC / pixel | CPU Time [ms] | |
kodak | -4.2 % | +7.5 % | -0.9 % | -4.3 % | -17.2 % | +3.4 % | 1303 | 74 |
clic20-pro-valid | -13.2 % | -0.2 % | -0.3 % | -1.3 % | -25.1 % | -2.3 % |
1357 | 354 |
jvet (BCDEF) | / | / | -0.2 % | / | -18.3 % | +18.6 % | 1249 | 143 |
jvet (class B) | / | / | +0.9% | / | -9.9 % | +20.7 % | 1300 | 282 |
jvet (class C) | / | / | -1.3% | / | -16.1 % | +9.2 % | 1289 | 69 |
jvet (class D) | / | / | -1.4% | / | -12.4 % | +9.6 % | 948 | 18 |
jvet (class E) | / | / | -3.8% | / | -6.2 % | +27.8 % | 1347 | 125 |
jvet (class F) | / | / | +0.2% | / | -31.8 % | +20.6 % | 1249 | 182 |
Decoding time are obtained on a single CPU core of an an AMD EPYC 7282 16-Core Processor, PSNR is computed in the RGB domain for kodak and CLIC20, in the YUV420 domain for jvet
We also present the performance-complexity continuum on Kodak. The different configurations (VLOP, LOP, MOP, HOP) are obtained using the configuration files detailed in the decoder documentation. For the MIX configuration, we pick the best configuration out of the 4 available ones for each image.