First build with automatic dewarping

[jimboh]

The commit called "Fix a bug in vertical bounds detection" at Sun, 30 Jan 2011 10:51:59 +0000 resolved the problem described in my immediately preceding post. Both the grid and the output are now perfect.

[Tulon]

Here comes beta 9 where you can adjust the auto-detected grid. This build also contains new features introduced by new developers, though I don't guarantee all of them will make it into 1.0

[emmerick]

Here comes beta 9 where you can adjust the auto-detected grid. This build also contains new features introduced by new developers, though I don't guarantee all of them will make it into 1.0

Thanks man. Great job

[JonEP]

Is it? Can it be? YES! The manual select content control is in ST. Huge cheers emerge from under the snow.

Thanks Tulon + crew.

Jon

[knk]

Hi,

first of all, thank you for this great piece of software. But there is a bug in beta 9 (a major one for me).
I scan 2 pages a once with a single camera setup (120°) and my plan was to correct the keystone effect using the new dewarping feature.
The dewraping algorithm works excellent, but the aspect ratio gets messed up (or is not corrected).

Here is a picture of what I mean:




Should be relatively simple to fix, I hope you guys can manage to do that.

Sorry for my bad english (its not my first language)

[Anonymous1]

I think ST's dewarping feature is for more subtle warps. What you are looking at is the Keystone Effect: http://en.wikipedia.org/wiki/Keystone_effect.

The other problem is that the book's width is not known from the original picture. If it were, you could just scale the image appropriately to adjust (in this case, just scale the image on the X-axis by a factor of 25/20 = 1.25).

So for to batch process using ImageMagick, run this:

Code: Select all

mogrify -resize 125%x100% *.tif

Quality will degrade slightly due to the resizing and scaling, though.

[Tulon]

Scan Tailor can actually handle any kind of dekeystoning, provided the camera had enough megapixels to get acceptable quality.

As for the aspect ratio, it's actually quite hard to calculate from a perspectively projected rectangle. Think about it: are you going to divide a horizontal edge by a vertical one? Which vertical edge are you going to take? They are quite different. Anyway, I may take a look at this problem when I am done with more important stuff.

[Anonymous1]

Does ST have some hidden de-keystoning features besides the super-epic dewarping? The only thing I could think of was if you delete all but 4 vertices on the grid, you basically get de-keystoning...

[Tulon]

The only thing I could think of was if you delete all but 4 vertices on the grid, you basically get de-keystoning...

Right. BTW, you don't even have to delete midpoints - as long as they are on the same line. The control points don't matter by themselves - they only help the user to control the shape of the top and bottom curves. If these curves are straight lines, you get dekeystoning without dewarping.

[zamacam]

The only thing I could think of was if you delete all but 4 vertices on the grid, you basically get de-keystoning...

Right. BTW, you don't even have to delete midpoints - as long as they are on the same line. The control points don't matter by themselves - they only help the user to control the shape of the top and bottom curves. If these curves are straight lines, you get dekeystoning without dewarping.

I'm not sure to understand correctly. Does this mean that software like qprc (http://diybookscanner.org/forum/viewtopic.php?f=3&t=784), PhotoShop or DxO are useless and that is better to do de-keystonne in ScanTailor?