1990 Volume 30 Issue 7 Pages 529-534
X-ray mappings on very large samples are used for the studies of segregations in steel. This has some consequences in terms of electron beam defocusing, calibration and time of acquisition.
To minimize this time, it is possible to increase the signal to noise ratio by using optimal linear filters. Traditionally these filtes are used in a univariate case, but as well study many elements simultaneously, the calculation of cross-correlations between elements allows a generalization of these filters to the multivariate case. The model is described with the assumption of a locally stationary phenomenon.
It can give the principal components of the spatial structure (i.e., synthetic images at different scales). This facilitates the next step which is to characterize the segregations by classical image analysis of the X-ray mappings. Application to an industrial program of reduction of segregation is presented.