<<HTML(<meta http-equiv="refresh" content="0;URL=../../CMV/Downloads/Pbfd">)>>

The University of Oulu Physics-Based Face Database

is a unique face database collected at the Machine Vision and Media Processing Unit, University of Oulu which contains color images of faces under different illuminants and camera calibration conditions as well as skin spectral reflectance measurements of each person. In addition, the camera channel response and the spectral power distribution of illuminants used are provided, thus the term physics-based. The database may be of general interest to face recognition researchers and of specific interest to color researchers.

At a glance

Image Acquisition

An image series of one person contains 16 frontal views each of which are taken under different illuminant calibration conditions by a 3CCD Sony DXC-755P camera. One image is 428x569 in 24-bit RGB, stored in BMP-format and is compressed using command 'compress' in UNIX. If the person had glasses another 16 images were taken with the person wearing them. There are minor variations in pose and distance between persons and camera.

Four illuminants provided by Macbeth Luminare were used for image acquisition: Horizon (denoted as h), Incandescent (CIE Illuminant A) (a), TL84 Fluorescent (t), and Daylight (CIE D65) (d).

Images of faces were taken under dark room condition. A gray screen was placed behind the sitting subject and the sides were surrounded by black curtains. Camera was first calibrated under one illuminant then images were taken without changing the camera settings under this and the three other illuminants. The same procedure were done for all illuminants resulting in 16 conditions.

Two letters are used to denote illumination condition: first letter stands for illuminant used in camera white balancing and second for the current illuminant. Thus, 'ah' means that the camera was calibrated under illuminant 'a' and the current illuminant under which the image is taken is 'h'. For faces with glasses on, the letter 'l' is added. For example, an image file ahl94.bmp.Z is a compressed BMP file of person 94 with glasses on taken under condition ah.

Example: person 94

Click on a sample to see the full image.

















Skin Spectral Reflectance

Spectral reflectance of facial skin were measured in three points (right cheek of image (cheek 1 in plot ), forehead and left cheek of image( cheek 2 in plot ) ) from 400nm to 700nm by 10nm steps with a Minolta CM-2002 spectrophotometer. In the database, these measurements are saved as text files and are named personNNN_1, personNNN_2 and personNNN_3, respectively, where NNN denotes person number.

person94_1, _2 and _3.


In addition, spectral sensitivities of the R,G and B channel of the 3CCD Sony DXC-755P and the spectral power distribution of the four illuminants are included from 400nm to 700nm in 10nm steps.

Note: the real camera sensitivities and spectral power distribution of the lamps were not measured in the laboratory.

Ordering and Other Matters

The UOPB Face Database is available for research and verification purposes upon e-mail request to Prof. Matti Pietikäinen. Please indicate your mailing address and the person or institution to whom the invoice will be addressed. The complete database will be sent in 2 CD's for a fixed handling fee of 35 € to cover copying and postage costs (this is non-profit). Invoice will be delivered with the CD. For legal reasons and for the privacy of the database participants, only faces 1, 3, 14, 25, 94 and 111 may be used for presentation or publication. Also, please cite one of the publications below to refer to the database.

  1. Marszalec E, Martinkauppi B, Soriano M & Pietikäinen M (2000)

    • A physics-based face database for color research Journal of Electronic Imaging Vol. 9 No. 1 pp. 32-38.

  2. Soriano M, Marszalec E & Pietikäinen M (1999)

  3. Martinkauppi B (1999)

    • Improving results of simple RGB-model for cameras using estimation. Proc. SPIE Europto Conf. on Polarization and Color Techniques in Industrial Inspection, June 17-18, Munich, Germany, pp. 295-303.

  4. Soriano M, Martinkauppi B, Marszalec E & Pietikäinen M (1999)

    • Making saturated images useful again. Proc. SPIE Europto Conf. on Polarization and Color Techniques in Industrial Inspection, June 17-18, Munich, Germany, pp. 113-121.

Contact Information

Prof. Matti Pietikäinen

Center for Machine Vision Research

Department of Computer Science and Engineering

P.O Box 4500, FI-90014 University of Oulu, Finland

Tel. +358-8-553 2782

Fax. +358-8-553 2612


MVG/Downloads/Pbfd (last edited 2011-11-22 12:20:25 by WebMaster)