Would you bother with the following data?

[swag72]

In my quest to get a better output, I am thinking very carefully about my data that is being inputted to be stacked. My normal thing is to go through my subs, check for star roundness and rogue planes, delete as necessasy then stack the rest.

Today, I have decided to look at it a little differently and would really appreciate some thoughts on this.

I know that we all have different ways of focusing, Bahtinov mask, FHWM to name just two popular ways. I think though it is fair to say that however you decide to focus an FHWM figure does ultimately give you an idea of focus. The lower the figure the better. WOuld you agree on this so far?

So I looked at my 50x300s luminance subs from last night, I picked a star that was saturated, and was probably about the middle size of all the saturated stars on offer on the Maxim screen. I then looked at the FHWM of each of these as a way of discarding those worse subs.

So, my highest FHWM was 4.84 and my lowest was 3.51 and a whole range in between. To me this seemed excessively high and would give me potentially bloated stars and an indication of poor focus.

So, am I right in trying to eliminate poor data in this way. Did I pick the right sort of star to test the FHWM on? How would I know what star to pick?

I was kind of hoping, as this is the luminance data and so giving all the detail that it would be nice and tight - These figures would suggest not.

I am using a SW 120ED (900mm) at f7.5 and an Atik 314L. Perhaps with this resolution this is the best I can expect?

I am really interested to see what you think about this and whether any of this data, based purely on the FHWM figures is worth saving?

[RobH]

Hello Sara.

I don't do any of that stuff....I just go through my images and if the stars are round and the image is in focus to my eye, I use it. If there's a satellite/meteor trail in an otherwise good image that also goes into the stack as if you're using an outlier reject stacking algorithm, it trail will dissapear but you'll get the benifit of the sub helping your S/N ratio.

Cheers

Rob

[swag72]

Cheers Rob - Here's my stack of ALL the luminance from last night. Just an auto strecth in PI to see the detail.

Would you be happy with this as luminance data? It doesn't seem to have much detail in it to my eyes.

[RobH]

Now't wrong with that data

The stars are big due to the stretching and there's great detail in the galaxies.

Run a 16 bit version through CCDSharp (deconvolution), paste this as a layer over a non-deconvolved layer and remove the deconvolved stars (they'll have rings around them. Dont stretch either of them at this stage.

You'll need to match the backgrounds using levels, but to enable you to see the effect, add a levels adjustment layer on top of the stack and do a hard stretch. Then highlight the darker of the 2 image layers and bring up the output level to match the backgrounds up. It's always better to lighten a background in order to get a match, rather than darken, as you guarantee that you don't lose any faint data this way.

Then discard the adjustment layer and flatten the image. This then becomes your luminance image and you stretch as normal.

To avoid star bloat, erase the stars in your stretched layer to an unstretched one underneath....easy with galaxy shots as the background is even....not so simple with stars in nebulae.

The method of using a temporary adjustment layer in order to see very slight changes is very effective and I use it lots.

Cheers

Rob

[swag72]

Thanks Rob. I've downloaded CCDSharp and will have a look and see how that works. If you are taking LRGB, do you deconvolve each channel seperately or once the channels have been combined?

[RobH]

I just deconvolve the luminance data Sara....you only need to deconvolve the RGB if you aren't using luminance.

Cheers

Rob

[Luke]

There's a tutorial here on deconvolution if it helps, Sara, under "Multi-Strength-Decon-Layer Blending":

Tutorials - Presentations - Ken Crawford

I've only just started playing around with deconvolution and I've found the tutorial useful.

Edit: I thought it used CCDSharp but looks like it's CCDStack, hopefully still useful as it goes on about how to blend in Photoshop?

I've downloaded CCDSharp and will have a look and see how that works.

[Ags]

Run a 16 bit version through CCDSharp (deconvolution), paste this as a layer over a non-deconvolved layer and remove the deconvolved stars (they'll have rings around them. Dont stretch either of them at this stage.

You'll need to match the backgrounds using levels, but to enable you to see the effect, add a levels adjustment layer on top of the stack and do a hard stretch. Then highlight the darker of the 2 image layers and bring up the output level to match the backgrounds up. It's always better to lighten a background in order to get a match, rather than darken, as you guarantee that you don't lose any faint data this way.

Then discard the adjustment layer and flatten the image. This then becomes your luminance image and you stretch as normal.

To avoid star bloat, erase the stars in your stretched layer to an unstretched one underneath....easy with galaxy shots as the background is even....not so simple with stars in nebulae.

The method of using a temporary adjustment layer in order to see very slight changes is very effective and I use it lots.

Cheers

Rob

Oo, er, I'll never get my head around this :-(

[Tim]

Sara, you are going to get really frustrated imho if you start being too picky about what you do and dont keep in each stack.

Like Rob I just quickly scan through each image for star shape, I check the background values to see if clouds have spoiled some subs, and star count to make sure it is in sync with the majority, and then just whack the whole lot in the pot for stacking.

When its all done just remember to treat your stars as a separate part of the image. As long as they are sharp, round, and colourful you dont have to worry about too much

BTW, that galaxy data looks to be very good. If you took some of the amazing galaxy shots that appear on here in their raw state they wouldn''t be much different.

Chin up

Tim

[swag72]

So the data looks good - I've heard that before with my M82 data - I managed to kill it!!

So I know the processing is THE biggest issue here. I just can't seem to get anywhere with it.