Page 1 of 1

Rotating images and methods

Posted: April 4th, 2015, 12:53 am
by tomczak
There are at least 4 transformations that could rotate an image by an acute angle:

- Level
- Warp
- Crop add border
- Composite

If only rotation is performed by any of the above, do they use similar resampling algorithms? Which one would be the best (the least blurry, or with the least amount of artifacts, for example)?

Cheers!

Re: Rotating images and methods

Posted: April 4th, 2015, 3:12 pm
by jsachs
They all use the same resampling code.

Re: Rotating images and methods

Posted: December 6th, 2015, 11:51 am
by tomczak
It looks like rotating in Crop/Add Border uses some sharpening routine at the end. Do they all do?

Re: Rotating images and methods

Posted: December 6th, 2015, 12:41 pm
by tomczak
Also, Level seems to execute the rotation part much faster than Crop/Add Border. I'm not sure how different the results are - I can't detect much difference.

Re: Rotating images and methods

Posted: December 6th, 2015, 6:40 pm
by jsachs
Level is faster because Crop rotates the entire image at full resolution to speed up subsequent operations at the same rotation angle -- not be cause of any difference in resampling.

Re: Rotating images and methods

Posted: December 6th, 2015, 10:36 pm
by tomczak
And the sharpening part of Crop rotate - is that common to all, just happening silently?

Re: Rotating images and methods

Posted: December 7th, 2015, 5:14 am
by jsachs
Am currently on vacation without access to the code.

Re: Rotating images and methods

Posted: December 8th, 2015, 10:04 pm
by jsachs
Only Crop automatically sharpens after rotating. It does the equivalent of the basic Sharpen function (with method = Sharpen) which does light sharpening to offset the slight blurring from the bilinear interpolation used to rotate the image. You can get the same effect in Level, Warp or Composite by sharpening the output images. Otherwise the effects should be more or less identical.

Re: Rotating images and methods

Posted: December 10th, 2015, 11:13 am
by tomczak
Cheers!