摘要 |
PROBLEM TO BE SOLVED: To perform blind unmixing of the spectral images of an observation target with high precision, which are labelled by a plurality of kinds of fluorescent substances. SOLUTION: This spectral image processing method includes: dimension reducing steps (S2-S7) for mapping respective pixels, which constitute the actually measured spectral image of the observation target, to a M-dimensional color space lower than the dimension of the spectral of the spectral image; selection steps (S8-S10) for calculating a hyperplane, wherein the average vector of the position vectors of the respective pixels is set on the M-dimensional color space as a normal line, performing the center projection of the respective pixels toward the hyperplane from the origin, finding M apexes of a hyper-polyhedron showing the distribution region of the respective pixels in the hyperplane and selecting M pixels positioned at these apexes as the optimum sampling pixels; and unmixing steps (S11-S16) for deducing the emission spectra specific to a plurality of kinds of the substances labelling the observation target, on the basis of the respective actually measured spectra of M pixels selected as the optimum sampling pixels and calculating the contribution ratio of a plurality of kinds of the substances to respective positions of the observation target, on the basis of the emission spectra and the actually measured spectral image. COPYRIGHT: (C)2010,JPO&INPIT
|