During our experiments with thermal imaging we created a database containing facial portriats taken in visible and thermal spectra.
The photographic stand used for taking portraits is a low-cost installation consisting of:
1. Infra-Red (IR) camera mounted on a 108 cm tall tripod with tubular and bull’s eye spirit levels
2. Digital Single Lens Reflex Camera (DSLR) mounted on a 116 cm tall W803Light Stand
3. CY-25WT fluorescent lamp mounted on a 116 cm tall W806 Light Stand (2 total)
4. fluorescent tube reflector lamp mounted on a 118 cm tall W806 Light Stand (2 total)
5. focus marker (5 total)
6. approx. 2m long piece of string
7. a 50 cm tall chair with back support
8. calibration board mounted on a 152 cm tall W803 Light Stand
9. a 275 cm 200 cm greenscreen
In our case, the DSLR is Canon EOS 600D, with a 14-bit CMOS sensor registering 5184 3456 pixels with 74:2 27:5 eld of view (FOV), interfaced by USB 2.0 (see Fig. 2), and the IR camera is FLIR SC325, with 16-bit sensor of 320 240 pixels working at 60 Hz, having 25 18:8 FOV, interfaced by Ethernet.
WIZUT Face Databases is just a simple folder structure, where main folder’s name is ‘database’ and within it, numerated subfolders are created, each representing a dierent subject (class). Every class folder stores raw images captured from IR and DSLR cameras, naming them accordingly: #imageNumber deviceID.leExtension (e.g. #002 ir.PNG). Subfolders within the class folder contain absolute temperature values (abs temp, saved in Matlab array), images after warping (warp), eye coordinates on both images (eyeCP,
saved in Matlab structure), images after cropping (crop), and images fused together from cropped faces, thermal and visual (fusion).
Sample images from our database are presented below. The WIZUT Face Database contains over 100 classes (with 5 image pairs per class, e.g. thermal and visible).