Skip to main content
. 2022 Mar 3;16:837208. doi: 10.3389/fnbot.2022.837208

Table 3.

Performance of the following low-light enhancement methods calculated on the DICM, LIME, MEF, and NPE public datasets using the no-reference metrics NIQE and BRISQUE.

NIQE BRISQUE
Method DICM LIME MEF NPE DICM LIME MEF NPE
MBLLEN 3.707 4.427 4.549 3.902 24.96 26.089 31.807 28.348
Retinex-Net 4.488 4.869 4.41 4.461 31.289 32.088 20.607 30.537
RRM 3.953 4.902 4.918 4.591 32.044 32.07 32.644 33.328
FFM 3.23 3.99 3.665 3.46 22.117 18.388 18.204 23.05
IB 3.211 4.191 3.735 3.961 26.909 22.496 17.895 33.871
KinD 3.564 4.628 3.877 3.443 25.238 23.681 27.505 23.336
DLN 3.009 4.054 3.356 3.571 20.19 18.656 17.165 25.754
SDD 3.554 4.302 4.268 3.54 28.267 24.163 28.095 26.238
MIRNet 3.133 4.008 3.77 3.701 17.926 21.634 25.275 23.431
Zero-DCE 2.899 3.997 3.697 3.527 19.215 19.473 16.604 26.581
Ours 2.847 3.938 3.291 3.422 17.228 17.74 14.243 22.492

The best results are marked in bold red.