I am using Document Understanding framework for extraction. I was trying to extract tables from few images/pdf files. Few images were slightly tilted, due to which tables were also tilted, and this resulted in wrong or no extraction.
In general OCR engines pre-process images before text extraction. I used omnipage OCR during Digitization, but it look it is not enhancing or doing the necessary modifications to image before OCRing.
Any solution to this issue, as you can understand practically we will never get ideal images as are always taken in account in tutorials and videos .