Restoration of Curved Document Images Through 3D Shape Modeling
Abstract
In this paper, we address the problem of discovering the 3D shape of a book surface from the shading information in a scanned document image. This shape-from-shading problem is characterized in real world environments by a proximal and a moving light source, Lambertian reflection and a non-uniform albedo distribution. By considering all these factors, we first build the practical model (consists of geometric model and optical model) to reconstruct the 3D shape of book surface. We next restore the scanned image using this shape based on two models, namely de-shading and dewarping models. Finally, we compare the OCR results on the original and restored document image. The experiments show that the geometric and photometric distortions are mostly removed and the OCR results are improved markedly.
Cite
Text
Zhang et al. "Restoration of Curved Document Images Through 3D Shape Modeling." IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2004. doi:10.1109/CVPR.2004.210Markdown
[Zhang et al. "Restoration of Curved Document Images Through 3D Shape Modeling." IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2004.](https://mlanthology.org/cvpr/2004/zhang2004cvpr-restoration/) doi:10.1109/CVPR.2004.210BibTeX
@inproceedings{zhang2004cvpr-restoration,
title = {{Restoration of Curved Document Images Through 3D Shape Modeling}},
author = {Zhang, Zheng and Tan, Chew Lim and Fan, Liying},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition},
year = {2004},
pages = {10-15},
doi = {10.1109/CVPR.2004.210},
url = {https://mlanthology.org/cvpr/2004/zhang2004cvpr-restoration/}
}