Abstract
Document image enhancement is one of the prime researches in the area of optical character recognition and computer vision. Preprocessing procedure for a document depends on document layout, aging and document material type. This paper proposes a preprocessing technique for the enhancement of ancient and degraded document images. Initially the degraded patches from the document image is collected and used for learning through a variational de-noising autoencoder followed by document image enhancement. Ground truth images of the degraded patches are trained with the help of an adamax optimizer. A deep learning architecture comprised of five levels of convolution is devised for encoding and decoding process. Down sampling is initially performed in the encoding stage after each level of convolution. Further up sampling is conducted in the decoding stage. Experimentations are conducted on DIBCO (2016, 2013, 2012, 2011, 2010 and 2009) datasets and the results of enhancement are found to be promising with an average RMSE of 0.106 for batch size 1 and 24 epochs.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.