omparison of image processing techniques for crack detection in oil paintings

Abstract

In the present study, three image processing techniques were applied to digitized oil paintings to detect their crackle and represent it in binary maps, which are subsequently compared. The three techniques applied were: oriented elongated filter, top-hat morphological filter, and K-SVD algorithm. The first technique showed the ability to detect the continuity of cracks but had problems discarding isolated pixels. The second technique detected continuity and represented the width of the fissures correctly. However, it did not work correctly in softened areas and with less lighting; and, the third technique quickly identified cracking in dark and softened areas. Thus, the results showed that the techniques are complementary to each other.