Hello,
while working on a script to detect and correct dust and scratches from slides, I got to the point where I use a threshold to define a mask identifying the regions to be inpainted.
I found that a low threshold gives me more regions, but too many, while higher values correctly identify only dust and scratches, but especially thin and long regions (hair, scratches, …) get significantly shorter and thinner.
For example, here three images:
The hair along the top side or the dust agglomerate in the centre are present in both images, but the size of the marked region differ significantly.
While reading the docs and the tutorial I noticed there is a function “-label” that marks connected regions.
I wonder if there is a way to use it to mark the (thinner and smaller. but correct) regions in the image with the higher threshold and then to find the same regions in the other image, where they are larger.
This would give me only the actual dust and scratches from the high threshold image, but with the greater area they show in the lower threshold image.
However, I’m not sure how to do this, if it is possible in the first place.
Thanks!