A challenge for tile manufacturers is detecting hard-to-spot defects and blemishes amid the randomly colored patterns on the tiles. Despite the push toward automation in the industry, quality control has remained a manual operation, subject to human error and its accompanying costs. Researchers from the University of the West of England have devised a solution that synthesizes a form of machine vision known as photometric stereo with advanced computer graphics tools. The technique distinguishes two-dimensional defects, such as unacceptable variations in surface patterning, from three-dimensional defects, such as scratches, chips or depressions. Photometric stereo distinguishes the pattern on a surface from the underlying topography. Here, the technique isolates raised features (below) on a ceramic tile. "In essence, a number of separate images, usually three, are captured from a fixed location under a sequence of controlled-illumination conditions," explained Melvyn Smith, a professor of engineering at the university, who developed the approach with his colleague, Lyndon Smith. Separating the 2-D color pattern from the 3-D surface topography enables visualization of the surface in relief, which reveals irregularities obscured by the pattern. Mounting uniform, collimated lights and the camera in an orthogonal configuration is key to the technique. Melvyn Smith said that they had initially used a conventional CCD camera in the system, but that they have since switched to a 1000 x 1000-pixel CMOS camera designed by Alrad Instruments Ltd. in Newbury, UK, for improved resolution and color depth. Frame grabbers manufactured by National Instruments, also of Newbury, and input/output cards that control the lighting system round out the setup. Earlier work with photometric stereo often attempted to recover the 3-D shape of an object, Melvyn Smith said, but the technique proved to be unsatisfactory because of limitations in recoverable surface orientation and cumulative errors in reconstructing the surface. "In the current application," he said, "the nominal underlying surface form is usually known and in many applications is often planar. Any three-dimensional topographic form may be considered to be superimposed on this underlying geometry." The automation of tile inspection, he said, offers economic and social benefits. The elimination of human subjectivity and error will increase manufacturing efficiency and offer the ability to generate statistical process-control data. On the human side, automated inspection will lead to better working conditions and safer work environments.