< Terug naar vorige pagina

Publicatie

Mesostructure Acquisition With Planar Illuminants

Boek - Dissertatie

In this dissertation we propose several techniques for acquiring local surface orientation and reflectance properties of small-scale surface details, or the surface mesostructure. Our primary focus is on increasing the applicability by providing an efficient, easy to implement and execute approach, employing solely off-the-shelf hardware components. Basically, our setup consists of a regular digital still camera and a computer screen functioning as a planar illuminant. Light patterns are displayed on the screen, illuminating the surface to be scanned, and reflections are captured by the camera. The recorded images are then processed, yielding a digital representation of the relief and reflectance of the scanned surface. In order to correctly process the acquired input images, geometric information of the position and orientation of the screen with respect to the camera has to be available. Therefore, we propose two different calibration methods employing a spherical mirror in order to make the screen visible to the camera. The first approach is efficient in terms of the number of calibration images, whereas the second method is more accurate and more efficient in terms of the number of manual sphere displacements. When the setup is calibrated, the material placed in front of the screen-camera setup can be illuminated by specific light patterns. We propose the use of Gray code patterns to efficiently scan the surface orientation as well as the reflectance of specular materials. Gradient patterns are used to obtain relief information of both specular and diffuse surfaces. We point out that Gray code patterns perform especially well for highly specular materials, whereas gradient patterns are more suited for glossy and diffuse surfaces. In practice, materials are often a combination of both a specular and a diffuse reflection component. We propose the use of an LCD screen to handle such cases, since specular and diffuse separation can robustly be achieved by cross polarization of the linearly polarized light emitted by the LCD. Only a simple polarizing filter has to be placed in front of the camera, blocking and unblocking specular reflections by rotating the filter a 90 degrees. The separated components can then be processed by the appropriate algorithm. As both the geometric relief and the reflectance properties strongly influence a surface’s appearance, we propose a method to also analyze the gloss level of the specular reflection component in addition to the geometric relief. This is achieved by a straightforward extension of the Gray code based normal acquisition method. The presented results show that a simple and inexpensive computer screen and digital camera can be transformed into a mesostructure acquisition system, yielding high quality scans. We believe that our system has possible applications in several areas, such as computer games, computer aided design and industrial inspection.
Aantal pagina's: 138
Jaar van publicatie:2010
Toegankelijkheid:Open