Blooming suppression with orthogonal frames

[LucMurphy]

Hi,

As anyone figured out the detailed workflow for suppressing bloom when you have 90 deg frames? as suggested by this topic:
https://pixinsight.com/examples/NGC6914-CAHA/index.html

I simply can't figure out the details...

Thanks
Luc

[msmythers]

Luc

I don't have a CCD and I'm confused by the mask part in Vicent's example. I decided to use his example image, crop the image into a separate Horizontal and Vertical bloom image. I then made a copy of both so I ended up with 4 images total. Image Integration won't work with less then 3 images. Here is the result I got. I had to change the clipping factors to extremes. I think that is because this is a stretched image. Anyway not perfect but it would be better with more subs.

Mike

[msmythers]

Luc

Just to add to this I went back and in the pixel rejection section I tried other rejection methods with the same result. I then for clipping I only checked the high pixels. I got the same results. I also changed to no normalization just to check.


Mike

[pfile]

to get the mask of the blooms, you could do something like this in pixelmath:

iif($T>0.98,1,0)

adjust the value of 0.98 to make sure you get the blooms. it's possible that you may want iif ($T==1, 1, 0). apply this to both H and V master images, setting pixelmath to create a new image.

now you have 2 masks. you can merge them with max(mask1,mask2) or even iif((mask1==1)||(mask2==1),1,0). again have pixelmath set up to create a new image.

now you have one mask. you can then blur the mask a little bit by removing 2 or 3 scales with AtrousWavelets or MMT or any other tool that will let you turn off wavelet layers.

then apply the mask to the master image with both H and V blooms.

now you can apply the pixelmath expression:

min(H_master,V_master) to the masked master image. for any given pixel, that should pull the pixel value from the image that is not saturated and replace the target image's pixel with that value. thanks to the mask it will only be applied where there is blooming.

i did this blind so hopefully i did not make a syntax or logical error...

rob

[jkmorse]

Rob,

I have to try your PixelMath solution.  What I have done in my situation with my 6303 that doesn't have an anti-blooming gate is to shoot a second set of subs with a short enough iteration to eliminate the bloom, then combine the two master sets with HDRCombination in the linear phase.  That seems an equally valid solution and has the added benefit of not needing the 90 degree rotation, as well as saving time by having the second, shorter set taking much less time to shoot.

Love to get your thoughts,

Jim

[pfile]

yes i think HDRComposition is also a great way to handle this. maybe better, in fact.

you can probably use the long-exposure, rotated images as input to HDRComposition as well, but i have not tried that.

for cameras that have rectangular sensors it's a better solution.

rob

[LucMurphy]

I have tried Rob's workflow concept (add thew min(H_master,V_master) to a master masked image) but a much more crude way of creating the master mask.

I have also tried two successices HRDC (master+H, then master1+V) but the second pass reintroduce some blooming that is very un-esthetic. A more precise mask may help.

What I did not write in my intro and that is not obvious in the images  i have uploaded because of the non-linear jpeg is that all the blooms have "shadows". This greatly complicate the creation of an effective mask. I have a 6303 and I would like to know from Jim if he experience something similar.

In anycase, thanks for your suggestions. I have a few idea to try now.

Luc

[jkmorse]

Luc,

I have to admit, when I use HDRComposition, I do not bother with masks.  The beauty of that tool is that is simply subs out those portions of the image that are saturated but leaves the rest untouched.  As such, you keep everything in the master image stack but the part that is replaced by the non-saturated bits from the shorter iteration stack.  That said, you need to make sure that the shorter iteration does not saturate all the way to the core.  What I have found in a couple of instances with my 6303 is that the shorter iteration stack looks fine at first glance, but when you get in close, there is still saturation in and around the core that looks bad if you get in really close to the saturated star.  Unfortunately my best test will come up if I ever get clear skies when the moon is down so I can get my 60 second iterations of M13 in the bank.  Once I get those I will play with this and report back.

Jim