An integrated methodology for the detection andremoval of cracks on digitized paintings is presented in thispaper. The cracks are detected by thresholding the output of themorphological top-hat transform. Afterward, the thin dark brushstrokes which have been misidentified as cracks are removed usingeither a median radial basis function neural network on hue andsaturation data or a semi-automatic procedure based on regiongrowing. Finally, crack filling using order statistics filters or con-trolled anisotropic diffusion is performed. The methodology hasbeen shown to perform very well on digitized paintings sufferingfrom cracks. Digital Image Processing Techniques for the Detection and Removal of Cracks in Digitized Paint
MANY paintings, especially old ones, suffer from breaksin the substrate, the paint, or the varnish. These patternsare usually called cracks or craquelure and can be caused byaging, drying, and mechanical factors. Age cracks can resultfrom nonuniform contraction in the canvas or wood-panel sup-port of the painting, which stresses the layers of the painting.Drying cracks are usually caused by the evaporation of volatilepaint components and the consequent shrinkage of the paint. Fi-nally, mechanical cracks result from painting deformations dueto external causes, e.g., vibrations and impacts.
The appearance of cracks on paintings deteriorates the per-ceived image quality. However, one can use digital image pro-cessing techniques to detect and eliminate the cracks on digi-tized paintings. Such a “virtual” restoration can provide clues toart historians, museum curators and the general public on howthe painting would look like in its initial state, i.e., without thecracks. Furthermore, it can be used as a nondestructive tool forthe planning of the actual restoration. A system that is capableof tracking and interpolating cracks is presented in [1]. The usershould manually select a point on each crack to be restored. Amethod for the detection of cracks using multioriented Gabor fil-ters is presented in. Crack detection and removal bears cer-tain similarities with methods proposed for the detection andremoval of scratches and other artifacts from motion picture.