High Fidelity Digital Acquisition For Large Area Ancient Painting

Cultural heritage has great values in history, science and art. But aging and scattering causes trouble in protection and research. On the contrary, digital heritage (DH) can be preserved eternally. Thus, digitization is a necessary and useful supplement to the current protection techniques. Moreover, digital heritage can be disseminated across the world, which will promote the research of cultural heritage greatly. Based on the above objection, digital acquisition of cultural heritage must faithfully reflect the characteristics of the original. To be more specific, it should be clear in detail as well as accurate in shape and color.Aiming at large area ancient painting digitization, a complete solution covering all stages including photographing, correction and stitching to achieve high fidelity is proposed in this dissertation as follows:1) An overview of the current digitization devices and techniques for paintingFirst, after an overview of the current digital acquisition devices and techniques for painting, their defects in the technical framework are pointed out. Then, the disadvantages of using ISO 12233 and MTF in testing the image shaipness of a camera are analyzed. After that, the geometrical and luminance correction is also overviewed. Finally, the digital image stitching techniques are also discussed and a conclusion that all the stitching algorithms are based on affine parameters estimation is drawn.2) Techniques for high fidelity digital acquisition of ancient paintingsThe definition of high fidelity is proposed. Then, techniques to guarantee geometry and luminance accuracy along with high definition, which can make the photos exactly reflect the original paintings before stitching, are proposed.3) Quantitative analysis of definition of acquired paintingsFirst, the accuracy of focusing methods is analyzed and compared. Then by integrating focusing, aperture and depth factors, an equation describing the theoretical maximum optical definition in painting digitization is proposed. Thus, the waste of computation and storage caused by scanning beyond optical limit can be prevented. A way to compute the theoretical optical aperture is also presented. Finally, because the theoretical optical aperture can not exactly reflect the practical performance, a clear region in a photo of ancient paintings by measuring the astignation is proposed and in this way, the practical optical aperture can also be determined.4) Image registration algorithm analysis and improvement for ancient painting imageCurrently, all the image registration algorithms are based on affine transformation parameters estimation. Hence, the correctness in registering photos of the planar paintings is proved mathematically. Then, an error formula depicting the affine transformation of images of quasi-planar paintings is proposed. It shows that the error is proportional to the ruggedness and resolution and inverse proportional to the focal length. Thus, when a quasi-planar object is acquired by a regular lens at high resolution, the image registration algorithms based on affine transformation parameters estimation will cause unacceptable error, which must be eliminated. Finally, an algorithm which can utilize the depth information from 3D scanning to correct the error is proposed. The algorithm includes three steps:optimal re-projection plane fitting by least square, transformation matrix computation between the optimal re-projection coordinate system and world coordinate system and affine transformation error correction for image of rugged surface.Last but not least, based on the above techniques, a high fidelity painting acquisition and storage system (PASS) for ancient painting is realized and tested in ancient painting documentation. The results of the acquired Dunhuang and other famous paintings shows the research in this dissertation has solved the problem of high fidelity digitalized acquisition of large area ancient painting effectively.