Determination of geometric distortions when digitizing stitched books

Abstract

At present, the topic of creating a digital heritage of cultural objects is one of the most popular and relevant in Russia. Particular importance is attached to methods of non-destructive digitization while ensuring high quality digital copy. This article describes an approach to eliminating geometric distortions of a page when scanning a stitched book. It is proposed to use a modification of the parallax method for determining page deformation. For this, a design and device for creating structured illumination by laser light sources have been developed. The scanning process is adjusted according to the image of the structure on the scanner lodgment. Each page is scanned twice. The first time - with structured illumination to detect geometric distortion. The second time - without illumination to form a digital copy. The coordinates of the nodes and lines of the structured illumination image are determined using cognitive algorithms implemented in a convolutional neural network.

Tests have shown high accuracy in determining geometric distortions. The proposed approach and cognitive algorithms can be generalized and used to create digital copies of various classes of cultural objects.