In-place correction
Short answer: In-place correction is a way of changing a scan that alters only the target text and leaves all other pixels as they are. It is the opposite of recognizing the whole page as text and laying it out again.
Steps
- Locate the target text with OCR.
- Work out the stroke area from the color difference between ink and background.
- Blend in clean background taken from nearby, or fill with an inpainting algorithm, and add back the paper grain.
- Fit color, blur and slant to the original ink, then write the new text.
- Add an invisible text layer so the replaced content is searchable.
Good and poor fits
Good for small, precise corrections, batch replacement and putting translations back into scans. Not suited to rewriting long passages, or to text that runs across lines, vertical text (except vertical Japanese and Chinese), or photos with strong perspective.