How to achieve the best results with Auto Dewarping in ST

[gbroadhead]

First of all, I want to say that I absolutely love Scan Tailor, and thanks to Tulon for creating it for us.

I have started building a Standard Scanner, but thought I would first try to my hand at just scanning a textbook laying flat on the ground and then processing with Scan Tailor and use the auto dewarping feature to achieve a flat image.

I have had good results using manual dewarping, but when using auto dewarping, I don't always get the best results. Often, the horizontal boundary is not recognized correctly, which forces me to go back and make a manual adjustment.

Does anyone have any suggestions on how to achieve better results with auto dewarping?

[daniel_reetz]

Do you have some sample images you can share? Scan Tailor's algorithms for deciding what edge is what depend on the way the image is set up. Most likely you can set up your images in a way that ST will really like.

[gbroadhead]

Here is my source image. I know that the angle of the book is not a flat 180 degrees. Could that be the problem?



And this is how Scan tailor reads the image. As you can see, the bottom line of the page is not recognized correctly.



Any tips/suggestions would be much appreciated!

[daniel_reetz]

I can't see your images (and I've tried on two computers). If your images won't attach to the post, feel free to try a free image host like imgur:

http://imgur.com/

[gbroadhead]

Yeah sorry about that: Lets try this

Here is my source image. I know that the angle of the book is not a flat 180 degrees. Could that be the problem?

http://imageshack.us/photo/my-images/35 ... esized.jpg

And this is how Scan tailor reads the image. As you can see, the bottom line of the page is not recognized correctly.

http://imageshack.us/photo/my-images/85 ... shotxc.png

Any tips/suggestions would be much appreciated!

[Tulon]

The problem is caused by the headlines reaching too far from the main content block. When I tested my algorithms, I didn't have examples like that. I am going to fix that as part of my work on moving dewarping to the Deskew stage. It's not going to be quick though.

[gbroadhead]

The problem is caused by the headlines reaching too far from the main content block. When I tested my algorithms, I didn't have examples like that. I am going to fix that as part of my work on moving dewarping to the Deskew stage. It's not going to be quick though.

Thank you! I appreciate the feedback.