Parameter estimation of spectral information on a subspace model

A set of spectral intensity data of interference color can be considered as a high-dimensional pattern vector which is varied, depending on the optical path difference as a parameter. In this paper, variation of a pattern depending on parameter change is generally treated on a model in a subspace. Unknown parameter can be estimated inversely from the projection of a test pattern on the subspace. A subspace to expand interference color was constructed by the learning process in which uniform colors of differential interference microscope were detected by the system. Two-dimensional distribution of optical path differences was successfully analysed.