Help Requested w/Gradient Merge Mosaic Problems

[jeffweiss9]

Georg-
   Using its own reference mosaic did the trick.  That worked fairly easily and then I just used StarAlignment to register the merged result to the Luminosity Mosaic. Since that involves an extra interpolation, I would think I'd only use that if there were difficulties merging using the 'master' luminosity reference mosaic. 
    I think I'm in business now with what look like a set of really good mosaics to go on to usual PI processing from here.  This was extremely helpful to me and is probably a good tutorial for others.
Thanks again.
Clear skies,
-Jeff

[StasV]

Hello!
He is my problem

Using GMM to build 12 frame mosaic I got sharp border at the B1,A1,B2,A2 intersection. screen shot 1
I tried to adjust GMM parameters. No result
 
After that I took only B1,A1,B2,A2 for GMM (same parameters). The result was perfect (screen shot 2)
 

[StasV]

screen shot 2 and mosaic scheme

[georg.viehoever]

...screen shot 1
I tried to adjust GMM parameters. No result

Zoom in to the borders of your images. You'll probably see that they take 1 or two pixels to blend into the background. For GMM, you need images that dont have this kind of smooth transition.

Option 1: Use DynamicCrop or simular tools to cut them away, and try GMM again.
Option 2: Increase the shrink parameter (maybe to 4) , and try again. This often (but not always) helps

If this does not help: Upload 2 of the problematic images somewhere, and I will have a look.
Georg

[StasV]

Georg, problem was really stupid. Frame B3 was swapped with preliminary sum B3+B2+A3   I am sorry... Thanks for answer

[georg.viehoever]

 

[StasV]

GMM generates one composite mask. Is it possible to get separated masks for each frames?

[georg.viehoever]

You wont get additional masks from with GMM. But you should be able to generate them from the original images.

[StasV]

But you should be able to generate them from the original images.

Yes, but it will a sharp mask. I need masks with smoothed borders according GMM "feather" process.
I intend to use them for a final adaptation by hands

[georg.viehoever]

Generating the smooth masks from sharp masks is not difficult:
- Create the sharp mask
- If you use "shrink>=1", then apply a MorphologicalTransform with operator Erosions, size shrink+1. This moves the boundary of the mask inwards by shrink pixels
- You can smooth the resulting mask using either Convolution (for instance with Library, Box (5)) or AtrousWaveletTransform (disable some of the lower layers, for instance 1+2).
Georg

[Warhen]

I was reviewing this thread in preparation for writing the mosaic chapter for my Inside PixInsight book for Springer Press. I just wanted to say that although I knew this, I wanted to state it publicly. What a kind and giving guy Georg is! Not only has he given us great tools for the program, but he's so helpful and to the best of my knowledge doesn't have any financial stake in the game. Thanks for being you George!

[pfile]

roger that, Georg has contributed so much to the PI ecosystem!

rob

[georg.viehoever]

 

[KuriousGeorge]

Hi folks, new to this forum with a 30-frame mosaic of M31 in the works (see attached). I'm hoping GradientMergeMosaic will make this project possible. Any ideas on how I could reduce or eliminate the distortions you see here? If feather radius is less than 250 they get worse with black holes.

Individual frames are 2758 x 2208 with 20% overlap and cropped to eliminate border artifacts from stacking. Used...

- DNA Linear Fit v0.3 on adjacent frames (not sure this was needed nor if it made a difference).
- StarAlignment > Register/Union - Mosaic with Frame adaptation, no distortion correction and Projective Transformation. Needed to use previews often to get a match. Saved as .tiff, 32-bit IEEE 774 Floating Point, Associated Alpha Channel.
- StarAlignment > Register/Match Images with frame adaptation.
- GradientMergeMosaic as shown in the figure (i.e., Overlay, 1, 250, 0).

Any help would be greatly appreciated.

[pfile]

you're talking about the artifacts around the bright stars, right? unfortunately i think you have to go in to where the overlaps end up on each image and paint out the bright stars...

rob