Niall took us through an overview of why Image Enhancement was necessary for OCR use and what the current state of the art is as practiced by the IMPACT partners.
Of course, this was largely about image enhancement for OCR and therefore the main issues have been to do with warping and how to de-warp images to allow the OCR engine to work properly.
The first tool that Niall showed us was the online version of the NCSR border removal tool, which was very impressive.
The second was a group of tools that undertook ‘binarisation’, ‘border-removal’ and ‘de-warping’….again very impressive.
Indicative Results
Assessment is still under way and so far it has not been possible to really scientifically test the tools, but early testing has suggested that the tools can out-perform commercial competition such as BookRestorer.
We then had quite a few questions from the audience regarding how this image enhancement is recorded and whether this optimisation should or could be included within the masterfile. Niall was sure that it would be both possible and a good idea to record this metadata with the image so both an original and an optimised version of the image could be delivered if required.
or see his slides here: