摘要

A method of restoring scratches on old paintings is proposed, and the corresponding high-accuracy output workflow is also developed. Firstly the scanner is selected as an input device to get the RGB (red, green, blue) image of the painting, and for the purpose of capturing high-quality image, scanner characterization is done by using neural network. And then the scratches on the RGB image are restored with the technology of digital inpainting, while the inpainting algorithm is mainly based on gradient vector and fast marching method. Finally the restored image is output with a printer, which is calibrated by using the high order polynomial regression method. In experiment the new replicated painting is well restored in the scratched areas, as well as keeps high resemblance with the original painting.

  • 出版日期2010

全文