Melotte 15 workflow suggestions

[dayers]

The image at http://orlop.net/hr_rgb10.jpg is the result of the workflow below. What do you think and where might I go from here?

I started with Jim Misti's unprocessed Ha, L, R, G, B data from the mistisoftware web site for the Melotte 15 star cluster and IC 1805 nebula (center of the Heart Nebula). My workflow to this point:

LUMINANCE

MorphologicalTransformation on Red image to reduce star prominence a bit.
DynamicAlignment to register Red and Ha images.
PixelMath using Vicent's magic and Harry's tutorial to combine Red and Ha into a new Luminance image.
HistogramTransformation to give new luminance its final stretch.
ACDNR
MorphologicalTransformation to shrink stars a bit.

RGB

ChannelCombination
BackgroundNeutralization with dinky preview upper left as reference.
MorphologicalTransformation, operating on those stars again.
DynamicCrop to get rid of edge artifacts
HistogramTransformation for a final stretch
SCNR to get rid of green cast
LRGBCombination

So there you are. To my eye colors need a lot of work and I haven't applied any noise reduction yet. Some stars have a blue halo that may need some attention. I am inclined to make CurvesTransformation my first stop.

What do you think? Thanks for your suggestions.

Dave

[sleshin]

Hi Dave,

The colors, a muted orange,  are indeed strange for this Ha rich nebula especially considering that you are incorporating Ha data. Presuming that the stated work flow is accurate, I would redo the RGB data and try the following: Do the Dynanic crop first. It's good to get rid of bad edges generally present after registration so these edges are not include when doing DBE or Background neutralization. Perhaps it wasn't needed but I would generally do DBE next and leave "Normalized" unchecked. The resulting image will have the background neutralized. Then do Color Calibration. If you left this step out then perhaps that explains the strange color. Then do the non linear stretch. At this point before creating the LRGB, the RGB image should have correct color, allbeit perhaps muted and requiring an increase in saturation at some point. If the color of the RGB is correct but then gets screwed up during the LRGB combine step, it might be related to poor matching of the luminosity of the HaR(the new L channel) and the CIE* L component of the RGB image. Using the Linear Fit tool to match these should help and it may also be helpful to change the RGB Working Space of the RGB image before doing the LRGB combine. In my recently posted image of the Medussa Nebula, I used an HaR image as Lum much as you are trying to do. To better match the CIE* L component of the RGB image I changed the RGBWS to a ratio of 0.5:0.25:0.25 and left the gamma value at 2.2. The resulting CIE* L component of the RGB image was a much better match for the HaR image so that the resulting LRGB combine image had rich reds rather than a  muted salmon color.

HTH

Steve

[dayers]

Hi Steve,

Thanks for taking the time to make your good comments and suggestions. I really appreciate it and will make another try at processing the data. Stay tuned.

Dave

[dayers]

http://orlop.net/Project-01.jpg

Wow! Am I ever a believer in DBE now, Steve. Thanks for heading me in the right direction. The left image is after ChannelCombination; the right is after applying DBE. I had been trying to ignore DBE because I didn't understand it, so I bit the bullet and studied Juan's video tutorial, with an able assist from Harry's video. Now I think I about half-way understand what it does and how to properly apply it. Maybe.

I'm a long way from through with processing this data of Jim Misti's, but I just had to stop and share my amazement.


Dave

[sleshin]

Wow indeed. Now do Color Calibration with the image still linear. Use the stars as the white reference. If the resulting image has any green still present, use SCNR and that should help.

Steve

[dayers]

http://orlop.net/rgb22s.jpg

I followed your suggestion and applied SCNR and linear stretch. Is this as far as I should go with RGB before combining with HaR?

Thanks again for your valuable help.

Dave

[sleshin]

Hi Dave,

The image could use a boost in color saturation which could be done with curves now or using the saturation transfer function in the LRGB Combine tool. So yes, you could add in the Luminance, your HaR channel, at this point.

Steve

[dayers]

http://orlop.net/Mel15_20110721.jpg

Before I start Yet Another Pass at processing Jim Misti's Melotte 15 data, I thought I would post this HaRGB image, although I am pretty sure I can make some improvements.

On this image I used Harry's video tutorial, which shows it works well on nebulae as well as galaxies. My workflow started with HaR according to Harry, new ChannelCombination using HaR, HistogramTransformation for permanent stretch, BackgroundNeutralization, DBE, SCNR, CurvesTransformation to improve contrast/brightness, and then ATWT for noise reduction.

All comments/suggestions/criticisms are very welcome.

Dave

[netwolf]

Dave, I saw that Jim's page is backup grabbed as much of the raw data as I could. I note though that Jim mentions a RGB scalling factor he uses when he combines 1.6:1.5:1 for RGB respectively. I am not sure if that might in anyway help your processing. But so far you have really provde the usefullness of DBE. So far I have only been running through one set of data I am working on again and again, and I am finding ABE is giving me better results than DBE. But I think its because I am not doing the DBE setup correctly. Will have to give it another go.

[dayers]

I'm glad that Jim resolved his web-hosting problem. Making his data available is a great service to those of us struggling up the PI learning curve.

I have noticed that the ColorCalibration tool gives results similar to Jim's R-G-B ratios, closer when a galaxy is used as white reference than when stars are selected, but close enough. So I don't sweat color balance too much.

Someone has said that when placing background samples for DBE, especially for images without a lot of pure background, Like Mel15, that not many samples are needed if they are "quality" samples. I find that I usually have to increase the default sample radius from 5 to 15 or 25 pixels to make it easier to place samples, and 15 to 25 samples strategically placed seem to be enough for good results. I haven't tried ABE.

Dave

Dave

[netwolf]

Thanks Dave, I will give that a go wit DBE, the image I have been trying is one of M8 captured at my friends observatory. So it has a lot of stars in the background. I have just used the auto generate with some minor movements of the samples to avoid the nebulae and stars. But ABE with defaults settings presented me with what I felt was closest to the pics posted by others. I will give DBE another shot.

I will give Jims files a go and see what I can mange with them. Its nice to have data like this to play with. It has been raining in Sydney for almost 2 weeks now. It just cleared today but the ground is to wet and mushy and the moon half full. Hope the new moon next week will bring clearer skies again.

Regards
Fahim

[dayers]

http://orlop.net/Mel15-HaRGB.jpg

I've taken Jim Misti's Melotte 15 raw data about as far as my present skills permit. The image above is an HaRGB composite. I created the HaR using Harry's PixelMath magic as described in his video tutorial. After ChannelCombination with G and B, I applied in succession BackgroundNeutralization, DBE, ColorCalibration, and SCNR. Then some noise reduction with ATWT while still linear. Then I gave it a Histogram permanent stretch and tweaked it with Curves.

I'm going to rest for a while.

Dave