infotech

 

 

Feedback and inquiries

Center for Machine Vision Research
Department of Computer Science and Engineering P.O.Box 4500
FIN-90014 University of Oulu
Finland
Tel. +358-8-553 2525
contact-cse (at) ee.oulu.fi

University of Oulu
Outex Texture Database

Image acquisition

Surface textures were captured using the setup shown in Fig. 1a, which included a Macbeth Spec-traLight II Luminare light source and a Sony DXC-755P three chip CCD camera attached to a GMFanuc S-10, a 6-axis industrial robot. A workstation controlled the light source for the purpose of switching on the desired illuminant, the camera for the purpose of selecting desired zoom dictating the spatial resolution, the robot arm for the purpose of rotating the camera into the desired rotation angle and a frame grabber for capturing the images. The relative positions of a texture sample, illuminant and camera are illustrated in Fig. 1b.

Each texture available at the site is captured using three different simulated illuminants pro-vided in the light source: 2300K horizon sunlight denoted as 'horizon', 2856K incandescent CIE A denoted as 'inca', and 4000K fluorescent TL84 denoted as 'TL84'. The spectra of the illuminants are shown in Fig. 2. Text files containing discrete representations of the spectra (400-700nm, 10nm sample step) and the sensitivites of the camera (obtained from Sony) can be downloaded here. The camera was calibrated using the 'inca' illuminant. It should be noted that despite of the diffuse plate the imaging geometry is different for each illuminant, due to their different physical location in the light source. Each texture is captured using six spatial resolutions (100, 120, 300, 360, 500 and 600 dpi) and nine rotation angles (0o, 5o, 10o, 15o, 30o , 45o, 60o, 75o and 90o ) .

The frame grabber produced 24-bit RGB images of size 516 (height) x 716 (width) pixels. The aspect ratio of the pixels was roughly 1.04 (height/width). The aspect ratio was corrected by stretching the images in horizontal direction to size 538x746 using the imresize command with bilinear interpolation provided in Matlab's Image Processing Toolbox. Bilinear interpolation was employed instead of bicubic, because the latter may introduce halos or extra noise around edges or in areas of high contrast, which would be harmful to texture analysis. Horizontal stretching was used instead of vertical size reduction, because sampling images captured by an interline transfer camera along scan lines produces less noise and digital artifacts than sampling across the scan lines. Given three illuminants, six spatial resolutions and nine rotation angles, 162 RGB images are captured from each texture sample. Eight bit intensity images are generated from the RGB images using the standard formula:

I = 0.299R 0.587G 0.114B

Fig. 1 a) Imaging setup. b) Relative positions of texture sample, illuminant and camera.


Fig. 2 Spectra of the illuminants.

Last modified: 2007-08-27