MULTIWAVELENGTH MICROSCOPIC IMAGE-ANALYSIS OF A PIECE OF PAINTED CHINAWARE - CLASSIFICATION AND REGRESSION

A multivariate microscopic study of a piece of painted china poreclain is undertaken by image analysis. The object has archeological and art istic value and therefore a detailed study may be worthwhile. The use of multivariate image analysis intends the exploration of the artefact (artistic object) in more spectral and spatial detail than by just vi sual inspection. The study is used as a means of introducing and furth er exploring the different aspects of multivariate image analysis. The goal of the paper is twofold: (1) showing how the multivariate image is constructed and analyzed and (2) using some of the obtained results to introduce and further expand some of the techniques of multivariat e image analysis. The example image has a size of 6 x 512 x 512. Class ifications by feature space segmentation and by regression are shown t o be useful and objective methods of acquiring insight in measurement errors and in the artistic detail of the painting. Some new concepts f or multivariate analysis are introduced. One of them is related to the comparison of regression models.