2

I have scanned image (2000 * 2500) and the original image of it. I want algorithm or open source API to develop a function which would give me the probability of matching these two images. I am developing this in C#. some of the issues are:

Do I need to work around on the image rotation part ? You never know what will come from scanner.

Gerhard
  • 6,850
  • 8
  • 51
  • 81
Piyush Patel
  • 139
  • 2
  • 16

2 Answers2

3

The scanned image might be distorted in many ways, to improve the recognition you should try to undo / ignore all of them:

  • When printed, the output might be scaled up or down to fill the page
  • The printed copy is rotated by a fraction of a degree during scan
  • Printer or Scanner hardware is unreliable and hardware movement is flakey, parts of the image are squeezed

So it is not simple to match the source and the scan, but there are some possiblities using image alignment. CodeProject has something that looks like a fit and there are some open source tools like Hugin that contain image alignment as part of panorama stitching.

Stephan B
  • 3,671
  • 20
  • 33
1

Look at the following questions for similar problems:

Community
  • 1
  • 1
Gerhard
  • 6,850
  • 8
  • 51
  • 81