Carina with 100ED in HDR

[avandonk]

Since it has been raining and cloudy for quite some time tracked down some earlier data to play with. I found I had 19 frames of Carina taken with the 100ED. So I fudged the data to extend the dynamic range.

Large image 2.4 MB
http://members.optusnet.com.au/~cheekyfish/CAR_hdrL_100ed.jpg

Details
Camera Canon 5DH, modded 100ED, Hutech LPS filter.
Exposures 19x2 min at an ISO of 800. I made two versions of this data.
1. Median of all 19 with ImagesPlus and Registar.
2. Median of odd and even frames and then summed them.
Used EasyHDR to produce a LDR image and the levels etc in PS.

Cant wait to get data with a bigger dynamic range when the weather allows.

Bert

[tornado33]

Excellent work preserving the inner detail and the outer. Youve got the HDR process working well.
Scott

[iceman]

Lovely image Bert, I can't wait to capture Eta!

[wysiwyg]

Awesome!
I need to learn more about HDR.

[Ric]

A stunning image Bert, very sharp and detailed.

Cheers

[jase]

The HDR method is really working well for you Bert. Really enjoy the nebulosity tones throughout the entire frame. My only comment is around the star colours. The process seems to alter the colour balance of them. Looking at the image in PS and using the colour range tool to select only the stars, then selectively remove stars mixed in nebulosity, you'll note the colour shift results in pink tones (first image). The histograms per channel indicate this as there is as a lack of green. By using curves on the green channel only, you can restore some of the balance (second image). Alternatively you could decrease the blue and red channel, but this maybe at the expense of nebulosity (if not masked). Overall, the level of detail you've obtained is impressive. Look forward to seeing more. Have you documented this process as yet? I feel sure others are keen to hear of your experience as you perfect the process. Job well done.

[avandonk]

I took your advice Jase and increased the green. Let me know if it is at the correct levels according to your discerning eye.

Here 2.3 MB
http://members.optusnet.com.au/~cheekyfish/CAR_hdrL_100ed.jpg

Because I do not have shorter exposures the star colour information is not there as it should be.

Here is a quick precis of the method subject to change as it is a work in progress.

Take sets of exposures doubling the time for each successive set.

EG 15, 30 sec. !, 2, 4, 8, 16 ...minutes. It is better to have more exposures for the longer exposures to minimise noise by median stacking. You could also double the ISO setting if the time gets to long. That is an 8 minute exposure at 500 ISO is equivalent to a four minute exposure at 1000 ISO.

What this means that for a Canon 12 bit sensor and 8 sets of exposures you will end up with an image with a dynamic range of 19 or 20 bits! Or 19 or 20 stops.

Convert all RAW files to TIFF and correct for flats etc with ImagesPlus.

You should now have LINEAR TIFFS.

I then use RegiStar to median stack each exposure set. RegiStar will stack the dim images even when you can only see a few stars.

You now should have a set of images differing only in exposure.

This is the tricky bit.

Run registar on all these images and make an averaged image but tick the intersection box in Combine. This is most important as you need to have an image where all the images have a common area.
Lets call this com_av.
I then trim by cropping com_av in PS to get a nice regular shaped image without any black bits due to drift or rotation. Lets call this com_av_crop

Move all the images into a new directory with com_av_crop.

Now run RegiStar again using com_av_crop as your starting image.

Now it is just a matter of combining com_av_crop with each registered image in turn.

Again tick intersection and set the weight of com_av_crop to 0.0001 and average. What this does is excise an exact area out of the exposure corresponding in size and content to the reference image com_av_crop.

Always check that only two images are ticked for combination. That is com_av_crop and the relevant exposure.
I then save these as 15srg, 30srg, 1mrg, 2mrg etc

Repeat for each exposure and you will end up with perfectly aligned and identical in size set of images differing only in exposure.

These can the be opened in EasyHDR as it will only take identically sized images.

Generate the HDR image and when this is done check that the EV differences are 1 EV for doubled images.
Tick the raw box as your images are linear.
Now generate the LDR image

There are a lot of controls on the right but only bring down compression to 0.4 or 0.6 say and either preview or process all. See pic below. You could also use the black and white clip controls but be careful you dont clip any real data. The image may look a bit flat but save it and then use PS to adjust levels. I am sure that I dont have all the methods right yet because if you fiddle with levels and curves too much you start to lose dynamic range so defeating the purpose of the exercise..

If anyone has any questions feel free to ask. Any suggestions also welcome as I dont possess all knowledge yet!
Thanks for all your advice Jase. We are all still groping in the dark. You can only learn by doing!

Bert

[h0ughy]

awesome work Bert!

[avandonk]

Thanks h0ughy. None of this would have been possible without interacting with a lot of other people mainly on IIS. The extraneous knowledge seems to slowly diffuse in from all around and one day an original idea pops up. I am only too glad to share it with everyone. I was just making sure I was not chasing what would turn out to be a dead end.

Obsession can do that to anyone. It is for others to try it out and really evaluate it. I would be interested to see what a 16 bit camera can produce with this method. Time will tell.

If you think I have a labyrinthine mind do not be surprised. A very senior scientist at CSIRO said about me "not only Irish AND Dutch but a bit of Indian as well, a very scary combination!" He was a bigot!

He was just scared of ?


Bert

[jase]

Much improved Bert. Well done. The colours may not be as vibrant as the original, but it certainly provides a more aesthetic view of this grand object. Thanks for sharing your HDR workflow. I was going to ask if you used intersect or union for Registar (crop/pad). I guess that if there is any pure black border in the sub frames, this would negatively image the final result. Perhaps once refined, you may feel this is a worthy article to publish. Would be interesting to hear your opinion of the CS2 HDR (detailed here - http://www.iceinspace.com.au/index.php?id=63,294,0,0,1,0) and the EasyHDR process.

Looking at the image again, if you desired to keep the original nebulosity colour tones, but remap the stars back to a more aesthetic color. You could simply use the colour range tool, then make your adjustments on the selection only. For selecting stars in an image, here's an alternative way to the colour range tool;

Star Selection

Make a grayscale copy of the image. I'll call this image #2.
High-pass filter image #2 with a radius of one pixel.
Apply a Gaussian blur to image #2 with a radius of one pixel.
Invoke Image->Adjust->Threshold.
Adjust the Threshold Level one click at a time until just the stars are white and everything else is black.
In the original image, in the Channels Palette, create a new channel. Name it "Stars." Choose "color indicates masked areas."
Paste image #2 into this channel.
Make just the RGB channels visible (i.e. make the Stars channel invisible).
Discard image #2.
In the original image, invoke Select->Load Selection. Choose the Stars channel you just created.
Invoke Select->Expand and expand the selection by a few pixels (e.g., three).
Done!!I think this process came from Russ Croman, but not entirely sure of the source.