Top-down segmentation of ancient graphical drop caps: lettrines

The restauration and preservation of ancient documents is becoming an interesting application in document image analysis. This paper introduces a top-down approach aimed at segmenting the graphical part in historical heritage called lettrine. The research principle was established on the concept of human visual perception and invariant texture analysis (co-occurrence and run-length matrices, autocorrelation function and wold decomposition). The preliminary results concerning segmentation stages were presented by highligthing difficulties related to the nature of strokes and textures in lettrines. Textured background was extracted although there existed some ambiguities. Nonetheless, the segmented areas of interest are informative enough to serve in an indexing method. The prospective bottom-up approach are mentioned and will be added to gain more precise segmentation.