integrate 2 images seamlessly….

[Emanuele]

Hi all,

So, suppose you have 2 images taken with a relatively long focal length telescope (980mm) and a short fl scope (450mm).

Now, is there any way to integrate the long focal length image in the short focal length FOV, without cropping?

Meaning: I have a large FOV of NGC7331 and Stephan's quintet. Now I want to image with a long focal length scope and image NGC7331 in detail. Then I want to superimpose this last image on the large FOV one, seamlessly.

Is there any way to do that in PI?

Thanks!
E.

[Niall Saunders]

Hi Emanuele,

To me this would seem to be best worked through using the DynamicAlignment process - but I think you would need to 'up-sample' the widefield image to give the same pixel 'scale' as that obtained from the longer FL scope before you started, and then you might need to 'sharpen' that image a little bit first.

DynamicAlignment will always 'crop' the SourceImage to the dimensions of the TargetImage - and you want your higher resolution NGC7331 to 'sit inside' the lower resolution widefield image. If this is going to be workable, you need the (target) widefield image to have FAR MORE pixels than the higher resolution (source) image of NGC7331.

As a start, I would up-sample the target widefield image by a factor of (980mm / 450mm) - the ratio of FLs of the two imaging setups

Let us know if this works - or if my 'though experiment' was flawed (they often are  )

[Carlos Milovic]

Niall, I think Emanuele was concerned on the seamless part, not the matching of the images

The usual approach is to build masks, and then use PixelMath to replace data (or average). One way to create such mask is to binarize one of them (I suggest the longer one., which I suppose fits enterely inside the shoft f frame), so all the signal is white and leave black the areas without signal. Then, use the minimum filter to "expand" the black area, with a low amount, so you'll create some sort of gradient.
Now, before applying the image, use the new LinearFit process to match both images. Then, using the mask on the short f frame, replace its contents with the image of long f (just put the name of the image in PixelMath). The mask will act a a linear "interpolation". This should work.

Another alternative is to try Georg's script for seamless mosaicing. Basically, it merges the images at the gradient domain, and then recovers the merged result. It should give better results, in terms of the seam, but right now the algorithm has several limitations. Anyway, it is worth a try.

[Emanuele]

Thank you Nial and Carlos!

I will try that as soon as I have a new image from the TEC140. Bad weather here.