Abstract

In conventional pattern recognition studies, individual components are classified by their spatial features, such as shape, size, spatial statistics, and spatial frequencies. For the cases in which the components could not be classified by the spatial features because of their mutual dependence, we have developed a distinct method based on the optimization theory, using the image data of a scene sensed at a number of different wavelengths (colors).¹,² In this paper, we present a novel idea of unknown-component reconstruction from unknown mixture scenes. The sample is excited (illuminated) at two different wavelengths separately, and the fluorescent images are sensed at different wavelengths. If the absorption spectra of components (which are equivalent to excitation spectra) are independent of each other, the component patterns can be uniquely found, because we have doubled the equations for fluorescent-emission spectra and spatial patterns of components, by two-wavelength excitation.

© 1988 Optical Society of America