Hi,
I shoot Alt/Az and have field rotation. That new feathering option is outstanding! I tested with feather=0, feather=10, and feather=20 and with 20, it eliminates that majority of border artifacts.
That made me wonder, is there a guideline on what to set it to, or what values may be too large or in the danger zone? I’m going to run more tests with 40 and 80 just to see, but I thought I would ask from the experts…
Great feature!
Hi Steven,
from the docs, in case you have not read it:
Borders feathering applies a feathering mask on each image during stacking. The smoothing is applied over X pixels distance, X between the value you set with this interface. The masks are cached in the./cachesubfolder.
So basically, the mask assigns a weight for each pixel in the image. All pixels further away than X pixels from the border are assigned a weight of 1. And then, the closer you get to the border, the smaller the weight, down to 0 with a nice smooth ramped shape.
When we then assemble a pixel stack (the value to assign to a given pixel in the stacked image), we make a weighted sum, i.e. sum over i of Weight_i * pixel_i / sum over i of Weight_i
There’s no good value because it would depend a lot on your data. I would say that for you use case anything in the range 10 to 100 that give no artifact should do. The larger the value, the more you are “at risk” to lessen the contribution of perfectly valid pixels to the overall results. This may give a bit more noise. To have a more deterministic value, try using the Profiling tool: on an image with mostly dark sky, plot a cut on a border and look at which distance from the border the curve is “smooth” (lest the noise of course) and represents correctly the background level. That should be a good measure to input in the feathering tool.
Now, if you were to assemble mosaics tiles instead, much larger values can (and probably should) be used because only one or 2 (usually max 4) pixels contribute to the final pixel stack.
One concluding remark, the masks are cached, just like we cache normalization statistics… so when you try out different feathering values, your second trial, and all subsequent trials, should run much faster.
Cheers,
Cecile
I noticed the second stacking with a different feathering option was faster… so that’s the caching! You sneaky little speed demons! Well done.