I try to recognize multiple documents containing tables with black, red and yellow text.
While red and black text is recognized, yellow text is not recognized in almost all cases.
I tried adding TextBlocks at the cells position. However, the text is still only recognized ~25% of the time.
I had greater success when I extract the yellow text as a Bitmap into a new ImageDocument. This, however, costs an additional page for each ImageDocument. This is not feasible.
I am guessing that the yellow text is lost during the binarization for the OCR process.
Any Ideas how to overcome this problem?
コメント
3件のコメント
Hi,
Thank you for using our products!
ABBYY FineReader Engine has a lot of options for improving recognition accuracy. I would advise you to try the following:
• Set a new resolution through PrepareImageMode::XResolutionToOverwrite and PrepareImageMode::YResolutionToOverwrite properties. 300 is a suitable value in most cases. But don’t be afraid to specify different values. Also you need to set PrepareImageMode::OverwriteResolution to TRUE before that.
• Specify a suitable profile. I guess in your case it should be DocumentConversion_Accuracy.
• Try to change properties of ObjectsExtractionParams. Visit Help → API Reference → Parameter Objects → ObjectsExtractionParams to see all available properties in details.
• Try preprocessing methods listed here: Help → Guided Tour → Advanced Techniques → Working with Images. Especially pay attention on the EnhanceLocalContrast method.
These recommendations are quite common, but I hope they will be useful for you.
Unfortunately, none of these suggestions worked.
I even tried adding InvertRegions to the affected cells which worked for two out of 40 cells... but that's not really useful.
I’ll be glad to help you and to give a more detailed answer, if you provide us with some examples of these images with yellow text.
You can send them to our email SDK_Support@abbyy.com.
サインインしてコメントを残してください。