Lacking in contrast!

[wimvb]

22 minutes ago, Rustang said:

Il have some more goes at my recent data which hasn't been helped by the lighter nights. 

Brighter nights are light pollution; they mean that you may have to decrease your exposure time from normal (but probably not by much). They also add more noise. The added light is subtracted by bringing in the black point. The associated noise is best reduced by increased integration time, or careful noise reduction.

[wimvb]

1 hour ago, ollypenrice said:

In Pixinsight try Local Contrast Equalization (which you can also replicate in Photoshop.)

While local histogram equalization (lhe) increases local contrast, it unfortunately increases noise as well. I've replaced it with multiscale median transformation (mmt) in my workflow. I find that mmt gives me much more control over where I want to add contrast; either in very small scales to enhance details, or in larger scales to enhance local contrast. Either way, it doesn't increase noise nearly as much as lhe. I should probably do a write up of it to show what I mean.

[tooth_dr]

30 minutes ago, Rustang said:

I fully understand that it's not for the faint hearted and needs a certain degree of dedication.

I don’t think my dedication can be called into question!  Running three cameras, three sets of calibration data, and trying get it all carefully coordinated through a 600mm dome slot without obscuring the field of view both pre and post-flip has proven tedious.  The adjustable dovetail saddle can’t handle the weight of two cameras without moving position after the flip, so I’ve decided just to stick with the two cameras and scopes!

 

 

[tooth_dr]

double post!

Edited June 26, 2021 by tooth_dr

[Rustang]

18 minutes ago, tooth_dr said:

I don’t think my dedication can be called into question!  Running three cameras, three sets of calibration data, and trying get it all carefully coordinated through a 600mm dome slot without obscuring the field of view both pre and post-flip has proven tedious.  The adjustable dovetail saddle can’t handle the weight of two cameras without moving position after the flip, so I’ve decided just to stick with the two cameras and scopes!

 

 

I wasn't questioning your dedication, I meant as a hobby for everyone in general because of the level of frustration it can bring amongst other things Its a shame you couldn't get the 3 rigs up and running, it would have made shooting narrowband a quicker process.

[Rustang]

A 32bit tiff of some Ha data for your processing pleasure. Its from this month so not the darkest of nights at the moment.

HOOHAStack.TIF

[ollypenrice]

4 hours ago, Rustang said:

A 32bit tiff of some Ha data for your processing pleasure. Its from this month so not the darkest of nights at the moment.

HOOHAStack.TIF 32.62 MB · 3 downloads

Thanks. This is processed for contrast, since that's the topic. It's not necessarily what I'd do 'for real.' I haven't edge cropped it and this is full size so JPEG losses.

Olly

[vlaiv]

[ollypenrice]

13 minutes ago, vlaiv said:

 

13 minutes ago, vlaiv said:

 

Nice. Just some black stellar cores (from masking?)

Olly

[vlaiv]

31 minutes ago, ollypenrice said:

Nice. Just some black stellar cores (from masking?)

Olly

Actually that is either stacking artifact or some error when importing TIF into ImageJ

They were that way in linear stage and I did not bother to fix it - I just did basic processing on the image (in fact - I don't usually do more than that):

1. Background wipe

2. Stretch

3. Masked denoise

First step is done in ImageJ and last two in Gimp. Mask is simply denoised layer used as transparency mask (inverted and stretched in opposite direction).

[vlaiv]

Here is linear data in ImageJ

Dark pixels in star core all have negative values while the rest of the image has regular values.

I just realized what happened - signed / unsigned conversion. It is ImageJ import thing - it imported data as 32bit signed integer - but in reality it is 32bit unsigned integer - those negative values are in fact values higher than 2,147,483,647 "looped" to negative values".

[wimvb]

I agree with vlaiv. I just downloaded the image and opened it in windows image viewer (which doesn't stretch). Some star cores are just "rings". @Rustang, how did you stack the images?

Edited June 26, 2021 by wimvb

[Rustang]

22 minutes ago, wimvb said:

I agree with vlaiv. I just downloaded the image and opened it in windows image viewer (which doesn't stretch). Some star cores are just "rings". @Rustang, how did you stack the images?

It was stacked in deep sky stacker, in dumb dumb terms, what is the issue with the stars, I haven't noticed anything odd my end unless I don't kniw what the 'odd' thing is I'm looking for!? 

[vlaiv]

8 minutes ago, Rustang said:

It was stacked in deep sky stacker, in dumb dumb terms, what is the issue with the stars, I haven't noticed anything odd my end unless I don't kniw what the 'odd' thing is I'm looking for!? 

Stars are fine - it is image format used to store data that is causing trouble (or rather how different software interprets pixel values).

It is best to save as 32bit floating point fits. You saved in 32bit integer TIF and 32bit integer can be treated as signed or unsigned - ImageJ and windows image viewer as shown by @wimvb treat TIF 32bit integer as signed integer and that means that very high values are treated as negative (up to half are positive - larger than half are negative - that is some sort of binary format stuff how to represent negative numbers).

Just export data from DSS as 32bit floating point FITS (or even TIF - 32bit floating point has strict definition) and you won't have these issues.

[Budgie1]

The stars are fine when I open the file in PixInsight, no dark centres at all.

A quick crop, denoise, stretch and curves in PI.

[Rustang]

20 minutes ago, vlaiv said:

Stars are fine - it is image format used to store data that is causing trouble (or rather how different software interprets pixel values).

It is best to save as 32bit floating point fits. You saved in 32bit integer TIF and 32bit integer can be treated as signed or unsigned - ImageJ and windows image viewer as shown by @wimvb treat TIF 32bit integer as signed integer and that means that very high values are treated as negative (up to half are positive - larger than half are negative - that is some sort of binary format stuff how to represent negative numbers).

Just export data from DSS as 32bit floating point FITS (or even TIF - 32bit floating point has strict definition) and you won't have these issues.

Apart from the ones around the edges! 🤣 Ok cool, I shall bear that in mind when I neck use DSS, thanks for your help with this.

Edited June 26, 2021 by Rustang

[Rustang]

Thanks again everyone, just need some improvements processing wise by the looks of it then, I think this data isant as sharp as normal even though the focus didnt seem any different so will have to keep an eye on that too.

[wimvb]

In PixInsight, the image opens fine with normal stars.

[Rustang]

39 minutes ago, vlaiv said:

Stars are fine - it is image format used to store data that is causing trouble (or rather how different software interprets pixel values).

It is best to save as 32bit floating point fits. You saved in 32bit integer TIF and 32bit integer can be treated as signed or unsigned - ImageJ and windows image viewer as shown by @wimvb treat TIF 32bit integer as signed integer and that means that very high values are treated as negative (up to half are positive - larger than half are negative - that is some sort of binary format stuff how to represent negative numbers).

Just export data from DSS as 32bit floating point FITS (or even TIF - 32bit floating point has strict definition) and you won't have these issues.

I just had a look at DSS and these are the options I have to save the data, none say '32bit floating point'!? 

[Rustang]

10 hours ago, ollypenrice said:

This is, with absolutely no doubt whatever, a processing issue.  It has nothing to do with capture.  Others have demonstrated the potential for higher contrast contained in the data.  However, we can point you towards the processing techniques which will enhance both global and local contrasts but the extent to which you apply them will remain your call. Many images strike me as over-processed in the way Craig T82 describes above. So I'll give you a couple of ways to enhance contrast but don't take that as my saying you should use them. Make your own picture. Wim has covered the black point so let's look at...

The stretch:

When we stretch an image we do so non-linearly, so we stretch the dark parts by far more than the light parts, so increasing contrast in the dark parts. That's how we extract the nebulosity from the background. However, different stretches will give different levels of contrast. If you stretch in Levels you'll get a logarithmic stretch which looks something like this when seen in Curves:

This is a gentle curve and the difference in stretch between the red and blue lines shows the difference in stretch between those points. This will give a natural looking image with moderate contrasts.

The alternative is an aggressive stretch.:

Here all the contrasts between dark and moderate signal have been massively increased. Note the difference in stretch between red and blue. That difference is contrast.

I've shown this stretch before on here and it has raised eyebrows and disapproval but it remains one of my go-to operations, above all in narrowband where its increase of star size is not such an issue. When I demo these curves on this data with guests, most prefer the hard stretch. It was the stretch used in the Ha and OIII of this final version: https://www.astrobin.com/327970/

If you use Photoshop consider Noel's Actions (now called Pro Digital Astronomy Tools) which has an excellent routine called Local Contrast Enhancement. In Pixinsight try Local Contrast Equalization (which you can also replicate in Photoshop.)

Olly

 

Thanks for the heads up on this, I actually forgot I already have those actions so Ive taken a look at the 'Local contrast enhancement' and it works wonders. The images below show with and with out the preset and the areas it enhances are perfect!

[vlaiv]

41 minutes ago, Rustang said:

I just had a look at DSS and these are the options I have to save the data, none say '32bit floating point'!? 

In DSS it is apparently called - rational for some reason (although rational can be fixed point as well - but that is technical stuff - just use 32bit rational).

 

[powerlord]

jfyi - I've found DSS's 32bit floating point incompatible with affinity.

Saved 32 bit floating point from APP import fine into affinity (all affinity processes, filters, etc work in 32 bit btw unlike photoshop), but I found nothing that worked for DSS - one of the reasons I moved to APP.

stu

[ollypenrice]

25 minutes ago, Rustang said:

Thanks for the heads up on this, I actually forgot I already have those actions so Ive taken a look at the 'Local contrast enhancement' and it works wonders. The images below show with and with out the preset and the areas it enhances are perfect!

A tip regarding Local Contrast Enhance: it pushes the darks down and the lights up, not unreasonably given its name, but that means you need a high black point to run it or it will black clip and bring the darks down too far. Bear that in mind. Even when you have a high black point, as I did when I ran it, it pushed the darks down too far. Not to worry: apply it 'As Layer on top' which is an action in the list, and then, in the Layers menu, change the blend mode to Lighten. When you do that, the darkening aspect of Local Contrast Enhance will be ignored and only the lightening aspects applied. You can then run it again and try the normal blend mode or give it a second dose of Lighten, as you wish.

And a tip regarding DSS: ditch it and use something else! I do my stacking and calibrating in AstroArt.

I had no black stellar cores when I opened your data in Pixinsight. Because I like working in Ps I exported the file as a 16 bit TIFF.

Olly

[wimvb]

Regarding local contrast, here's the same target captured and processed by Sara Wager

https://www.swagastro.com/ngc7822.html

Btw, @Rustang to get the most out of your data, you should apply flats.

Edited June 26, 2021 by wimvb

[Rustang]

21 minutes ago, wimvb said:

Regarding local contrast, here's the same target captured and processed by Sara Wager

https://www.swagastro.com/ngc7822.html

Btw, @Rustang to get the most out of your data, you should apply flats.

Yeah her work is amazing. You have me worried now as I do use flats, I stack with Darks, flats and dark flats. Do you think the flats aren't working then, if so how can you tell!?