Abstract

A new theory is developed for estimating spatial patterns of unknown components from multispectral images. The component patterns are uniquely found from the image data by using two sequential eigenvector analyses. No constraint on the component patterns and spectra, such as nonnegativity and finite extent, and no criterion of optimization is used in this method. The limitation of the method is that it is applicable only to specific image data, formed by two related physical processes. We describe the theory in the case of fluorescent sample images. The sample is excited (illuminated) at two different spectral bands, and the fluorescent images of the bands are observed at a number of different emission bands for each excitation. The experimental result with a cytochemical specimen verifies the effectiveness of the proposed method.

© 1990 Optical Society of America

Component analysis of spatial and spectral patterns in multispectral images. II. Entropy minimization

K. Sasaki, S. Kawata, and S. Minami
J. Opt. Soc. Am. A 6(1) 73-79 (1989)

Component analysis of spatial and spectral patterns in multispectral images. I. Basis

S. Kawata, K. Sasaki, and S. Minami
J. Opt. Soc. Am. A 4(11) 2101-2106 (1987)

Estimation of component spectral curves from unknown mixture spectra

Keiji Sasaki, Satoshi Kawata, and Shigeo Minami
Appl. Opt. 23(12) 1955-1959 (1984)

Diffraction-limited imaging of unknown objects through fixed unknown aberrations using interferometry

Claude Roddier and François Roddier
J. Opt. Soc. Am. A 7(10) 1824-1833 (1990)

Method of evaluating image-recovery algorithms based on task performance

K. M. Hanson
J. Opt. Soc. Am. A 7(7) 1294-1304 (1990)