Identificação de regiões de lacunas numa pintura retabular: análise comparativa de métodos de classificação em ambiente SIG

In this text some GIS-based spatial analysis methods are presented, enabling the quantification of areas with white lacunae in a 16th century wood painting. With an imageproduced during a conservation/restoration intervention, and applying a principal component analysis (PCA) as well as several classificators, thematic maps of the lacunae were created. Four methods were used: automatic division of histogram in intervals (level slicing); alimiarization (thresholding); a supervised classification method based in the productionof training areas, a spectral signature and a maximum likelihood classification; and anunsupervised classification method with a cell aggregation algorithm (iso cluster). In the conditions under which essays were made, best results in the identification of lacunae were obtained with the limiarization (thresholding) and the supervised classification methods.