Rob A's (Im)personal Blog.

Another update to my histogram match script. Now with a whole bunch of channel options - In addition to Value and RGB, there is now HSV, HS, LAB, AB, YCrCb, and CrCb. If the images are real close, like two exposures of the same scene, try RGB. To transfer the colours, try AB (preserve L) or CrCb (preserve Y).

EDIT: One more change to the uploaded file, the script not decides if rescaling is needed and does it automatically to an optimal size.

The download can be found here...

For some bizarre reason I needed to transfer the histogram profile from one image to another image. Turning to Gimp, I decided to script it. I used script-fu, as I am still trying to learn python. The basic premise is to sample the histogram of both images, then convert them into an incremental function. Following that, you determine the transfer function to convert back from the one to the other. I realize that sounds complicated, but it isn't that bad. Here is a diagram:

So all I had to do was write a gimp script to do this. No Problem!

On another forum I frequent, someone posted a link to a fake tilt-shift video and asked how it was done.

Final

Original

Previously I had seen still tilt-shift images and googled around to figure out how to create them using Gimp. Lots of references like this one at Gimparoo, or various blogs. One thing in common these had was using a gradient mask to blend a blurred copy with a non-blurred copy, or in the case of fancier efforts that use the focal blur plugin, a gradient to drive that plugin.

To me, these efforts left a bit to be desired, as the gradient would leave odd artifacts of focused bits of objects crossing in front of the focal plane gradient point. I figured a good way to get around this would be to construct a complete mock depth map based on the image. Here is an example of a before image, and the fake tilt shift.

Keep reading for the how-to...