Issue No. 02 - February (1994 vol. 16)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/34.273732
<p>In many remote sensing and machine vision applications, the shape of a specular surface such as water, glass, or polished metal must be determined instantaneously and under natural lighting conditions. Most image analysis techniques, however, assume surface reflectance properties or lighting conditions that are incompatible with these situations. To retrieve the shape of smooth specular surfaces, a technique known as specular surface stereo was developed. The method analyzes multiple images of a surface and finds a surface shape that results in a set of synthetic images that match the observed ones. An image synthesis model is used to predict image irradiance values as a function of the shape and reflectance properties of the surface, camera geometry, and radiance distribution of the illumination. The specular surface stereo technique was tested by processing four numerical simulations-a water surface illuminated by a low- and high-contrast extended light source, and a mirrored surface illuminated by a low- and high-contrast extended light source. Under these controlled circumstances, the recovered surface shape showed good agreement with the known input.</p>
stereo image processing; shape information retrieval; multiple images; remote sensing; machine vision; reflective surface; image analysis; specular surface stereo; image synthesis; image irradiance; camera geometry; radiance distribution; mirrored surface; water surface; shape from shading
H. Schultz, "Retrieving Shape Information from Multiple Images of a Specular Surface," in IEEE Transactions on Pattern Analysis & Machine Intelligence, vol. 16, no. , pp. 195-201, 1994.