2017 eclipse - HDR

[pfile]

entirely processed in pixinsight! lots of pixelmath to get the moon in there and the proms around it. the moon images were stacked from FFTRegistration of the longer exposures, while the corona images were stacked from DynamicAlignment on Regulus. there are 3 ramps put together with HDRComposition and then integrated together.



rob

[Niall Saunders]

That's impressive Rob. There can't be many images out there showing a star, the Moon and the 'mid-day' Sun - all at the same time.

[pfile]

thanks Niall!

[astroedo]

Wonderful image!

I'm trying to process a few images from a friend (unfortunately I wasn't in the US  )
I can align images with FFTRegistration and easily create the HDR image, but, when I try to compress the dynamic range with HDRMultiscaleTransform I get a Bright ring aroun the Sun.

Probably the wavelet based HDRMT gives a sort of ringing effect cause by the fast drop in luminosity near the moon edge.
If I reduce the number of layers, the bright ring, shrinks but never disappears.

As a reference I'm uploading a sample image (it's a single shot not the HDR one, but the effect is the same).

Do you have any hint on how to manage this problem?

Thank you

Edoardo

[pfile]

can you align on a star? because FFTRegistration, as i understand it, is going to end up aligning on the moon, which of course is moving relative to the sun's face during the eclipse.

what i did was to use pixelmath to replace the moon with the color of the corona that was just outside of the moon's edge. i believe that removing the brightness discontinuity represented by the moon helps the HDRMultiscale transform work better without ringing.

so then you have to paste the moon back somehow. i actually used a stack of FFTRegistration-registered images to construct the moon image.

rob

[RickS]

Very nice, Rob!

Did you try the LarsonSekanina filter?  When I did my TSE image in 2012 that was the secret to bringing out detail in the corona.

Cheers,
Rick.

[pfile]

i did not... i think i'm going to have another go-around with this one since it did get kind of crunchy with the HDRMultiscaleComposition. i'll check out that filter. after 3-4 days of registering images and trying all kinds of different stuff i'm a little burned out.

rob

[RickS]

Give it a go when you've had a break.  I think you'll like it.

[pfile]

i meant to ask, is that filter for linear or nonlinear images?

rob

[Alejandro Tombolini]

Hi Rob, excellent work!

I'm waiting to came back to process mine and every hint in this discussion will be appreciated.

I see a big circular halo of approximately two solar diameters, did you have clouds Glendo? I am worried about this because in our point the sky was not completely clean.

Saludos, Alejandro

[RickS]

i meant to ask, is that filter for linear or nonlinear images?

It was 5 years ago, but I think I used it on stretched data.

[pfile]

Hi Rob, excellent work!

I'm waiting to came back to process mine and every hint in this discussion will be appreciated.

I see a big circular halo of approximately two solar diameters, did you have clouds Glendo? I am worried about this because in our point the sky was not completely clean.

Saludos, Alejandro

i've talked to a couple of other people who imaged the eclipse and the conclusion is that the halo is a common thing. the halo could be caused by the earth's atmosphere, not sure, but i don't think it was caused by the local circumstances - it was pretty much completely clear at glendo.

rick's suggestion to try L-S is pretty key. i think this is somewhat equivalent to the radial filters you read about online. so i ended up doing HDRMultiscaleTransformation, LHE and LarsonSekanina, the last two thru L masks. the L-S filter does have to be performed on the stretched data.



on this version i also tweaked the color a little bit.

rob

[cdavid]

Absolutely stunning Rob!  I'm still tediously registering all my sets....hope to follow your outline and see if I can come up with anything good.  Thanks for your tips the other day!

Re- halo.....yes I am seeing it on most everyone's images.  My location was high cirrus clouds and the halo is quite prominent on my photoshop processed version.

http://astrob.in/309506/0/

Best
Carlos

[pfile]

one thing that's plagued me throughout this is that no matter what settings i use in HDRComposition, there is always banding in the merge. the masks can clean this up, but i'm starting to think something must be wrong.

one idea i had is that since every image is overexposed, there's necessarily some band around the sun in each image that is outside the linear range of the sensor, but is not saturated. HDRComposition could be merging these areas in even though the data in these areas is bad. interestingly i see this problem most strongly in the green and blue channels. i am going to experiment with forcing all data in each image > some threshold to 1 and see what happens.

rob

[pfile]

i don't think this theory is correct - i think that there's some strange behavior in HDRComposition for RGB images. the short story is this - in order to avoid banding in the RGB image, i had to lower the binarizing threshold all the way down to 0.04, but that of course ends up drawing too much information from underexposed images. even with the binarizing threshold that low there was still some banding in the G and B channels - the R channel was fine.

if i split the image into RGB channels and do HDRComposition on the mono files, everything works pretty much as expected with default settings, but there is exactly one band in the image which seems to be caused by the merge of the 1st and 2nd images. this can be cleaned up easily with HT thru the Mask01, but something is still wrong. i'm going to try to crop/narrow down the images and post them for juan to look at.

rob