EARI Polarization Demosaicking Codes
This is the source code of our proposed edge-aware residual interpolation (EARI) for monochrome and color polarization demosaicking [Project].
Note on Paper Result Reproduction (March 19, 2021)
We found that we originally uploaded an incorrect file for color demosaicking. Thus, we re-uploaded a correct file that can reproduce our paper results for color polarization demosaicking. We are sorry for any inconvenience this may occur. To reproduce our paper results, please use [0, 1]-normalized MAT file as the input, because the parameters are set based on [0, 1]-range data.
Demo
- sample_download.m
To run our demo codes, please first run sample_download.m code to obtain sample 12-channel color-polarization data to be used for the demo.
- demo_monochrome.m
Input
4-channel monochrome-polarization data (G channel data): I_0, I_45, I_90, I_135
Mosaic pattern
90 45
135 0
Output
Demosaicked monochrome images for each polarization direction: Dem_0, Dem_45, Dem_90, Dem_135
Stokes parameter images derived from the demosaicked images: Dem_S0, Dem_S1, Dem_S2, Dem_DoP, Dem_AoP
CSV file containing PSNR values and angle RMSE for Table 1 in the paper
- demo_color.m
Input
12-channel color-polarization mat data: mat = cat(3, RGB_0, RGB_45, RGB_90, RGB_135)
Mosaic pattern
R_90 R_45 G_90 G_45
R_135 R_0 G_135 G_0
G_90 G_45 B_90 B_45
G_135 G_0 B_135 B_0
Output
Demosaicked RGB images for each polarization direction: Dem_0, Dem_45, Dem_90, Dem_135
Stokes parameter images derived from the demosaicked RGB images: Dem_S0, Dem_S1, Dem_S2, Dem_DoP, Dem_AoP
CSV file containing PSNR values and angle RMSE for Table 2 in the paper
Dataset
Our 12-channel full color-polarization dataset is downloadable from our project page [Project].
Reference
The code is available only for research purpose. If you use this code for publications, please cite the following paper.
"Monochrome and Color Polarization Demosaicking Using Edge-Aware Residual Interpolation"
Miki Morimatsu, Yusuke Monno, Masayuki Tanaka, and Masatoshi Okutomi,
IEEE International Conference on Image Processing (ICIP), pp.2571-2575, October, 2020.