Digital restoration of damaged color documents based on hyperspectral imaging and lattice associative memories

This paper introduces an autonomous hybrid technique designed for the digital restoration of the missing parts and occluding artifacts in damaged historical or artistic color documents. For this purpose, a hyperspectral imaging device is used to acquire sets of images in the visible and near infrared ranges. Assuming the presence of linearly mixed pixels registered from the spectral images, our technique uses two lattice auto-associative memories to extract the set of pure pigments spectra. Fractional abundance maps indicating the distributions of each pigment along the image are obtained by spectral linear unmixing. These maps are then used to locate holes and cracks in the document under study. The restoration process is performed by the application of a modified morphological region filling algorithm, followed by a vectorial linear interpolation scheme to restore the original color appearance of the filled areas. For illustration purposes, our procedure has been applied successfully to the restoration of superimposed scripts and an art painting.