Shepard's distortion #2352
Replies: 1 comment 3 replies
-
|
Hi @tourtiere, That's very cool! I didn't realize Shepard was so straightforward. You can make your loop very slightly faster with: for p1, p2 in couples:
diff = index - p2
distance = (diff[0]**2 + diff[1]**2)
distance = distance.ifthenelse(distance, 0.1)
weight = 1.0 / distance
delta = [(p1[0] - p2[0]), (p1[1] - p2[1])] * weight
weights.append(weight)
deltas.append(delta)
# add, normalize
index += pyvips.Image.sum(deltas) / pyvips.Image.sum(weights)Though it doesn't make a lot of difference. I added a (1/x) shortcut to pow in git master libvips which helps a little too. |
Beta Was this translation helpful? Give feedback.
-
|
1/x patch: dd4b110 |
Beta Was this translation helpful? Give feedback.
-
|
I added your nice code to the pyvips examples area, I hope that's OK. |
Beta Was this translation helpful? Give feedback.
-
|
It's perfect! Thank you |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Hi,
Just a quick note to say that I previously used ImageMagick and its shepard command to distort large orthophoto mosaics, but it was using too much RAM. I found libvips and its mapim command so I came up with this script and it really improved my project: https://github.com/tourtiere/light-distortion. Feel free to use it as you want in the examples.
Libvips is very powerful and fun to use. Thank you John and other contributors for this incredible work, have a nice day to you all
Beta Was this translation helpful? Give feedback.
All reactions