Re: [ccd-newastro] Luminance vs color FWHM question

Expand Messages
• 1) All I was saying was that, with three color filters, each filter is going to pass approximately one-third the light of a clear or missing filter. So if you
Message 1 of 39 , Apr 1 12:14 AM
1) All I was saying was that, with three color filters, each filter is going to pass approximately one-third the light of a clear or missing filter. So if you want to achieve the same S/N as the clear/lum data, you need 3X as much exposure time. that's all, just that simple mathematical relationship applies.

This is important because you will have achieved some level of success with deconvolution, and that success is based strongly, if not completely, on the S/N of the lum data you used. If you expect to get good results in deconvolving color data, then you will need that same S/N.

Given (as Max points out) that the human eye sees color data with less detail than luminance data, you can see why it is not often that color data would be deconvolved. But it _might_ be useful, it _might_ show something unexpected, so for someone who wants to try, they just need to know that S/N is very important in successful deconvolution (that is, deconvolution without artifacts).

As you point out, if the color data is 1x1 you could consider adding it to the lum data before deconvolving. But you should make sure that, in fact, it will improve the S/N to a degree that is useful. It might be that the noise contribution from 1x1 color data would _decrease_ S/N if it's noisy enough for whatever reason (shortness of exposure, for example). Clever statistical combining might help, but you might wind up finding out that the effect of a fancy statistical combine would be to blur the data! In which case, you would have been better off with your 2x2 binned color, or in leaving out your 1x1 unbinned color.

(Only averaging completely avoids blurring effects in a combine. All statistical combined run the risk of removing real data as well as outliers.)

There's no free lunch. The universe is quite stingy with signal. <g>

Ron W

On Mar 31, 2010, at 4:08 PM, Neil Fleming wrote:

> A couple of questions and a statement on this...
>
> Questions: Ron, you said, "Since the color data is filtered, you might need 3x the color data relative to lum in order to have good enough S/N to deconvolve." I have always wondered about this.
> 1) Do you mean that if you take 1 hour luminance, that you would suggest taking 3x1 hour (total of 3 hours) filtered data?
> 2) This would be due to the pass-through statistic of the filter in question?
>
> Statement on the decon:
> If everything were binned 1x1, and you had lots of data, then I would try (and probably advocate) that you add the luminance info from the color data (synthetic luminance or "sLum") to your luminance master ("Lum") to get an sLum+Lum result. The weighting for the combine would be derived from CCDStack's normalization process or calculated from the relative durations of each data set crossed with the filter pass-through. I would then denoise and deconvolve that combined result.
>
> ...Neil
>
> www.flemingastrophotography.com
> Direct from Boston - brilliant diamonds in pea soup
> Also check out the astro_narrowbandYahoo group!
>
>
>
>
> ________________________________
> From: Wodaski - Yahoo <yahoo@...>
> To: ccd-newastro@yahoogroups.com
> Sent: Wed, March 31, 2010 12:13:05 PM
> Subject: Re: [ccd-newastro] Luminance vs color FWHM question
>
> This really depends on the quality of the data you have in hand. Even if the color is unbinned, do you have enough data, to get to low enough noise, to even be able to deconvolve? <g>
>
> Since the color data is filtered, you might need 3x the color data relative to lum in order to have good enough S/N to deconvolve.
>
> So if you think it might work out, you can give it a try, and see for yourself. If the color data winds up with artifacts, you'll know it wasn't good enough (or, if you know how to do the noise math, you can do a calculation and see if you are at all close with whatever you have).
>
> But trial and error both ways, or with maybe simple sharpening instead of decon, might give you some ideas.
>
> Generally, it's normal to spend time figuring out what you can do with the quality of data you have collected, to determine the best way to get to the best final result. Professional astronomers (and statisticians) deal with that all the time; it's the bread and butter of their work, really.
>
> Ron W
>
> On Mar 31, 2010, at 8:19 AM, glibbins2003 wrote:
>
>> I'm wondering what folks are doing with this issue: I have a nicely deconvolved luminance file, nice resolution etc; now I have a LOT of color data which is also unbinned; should I try to match the luminance with the same degree of decon and try to process out the color noise or not; if not what effect does that have if my luminance stars are fwhm 2-3 and the color versions are around 4?
>> Thanks,
>> Dave
>>
>>
>>
>> ------------------------------------
>>
>>
>>
>>
>
>
>
> ------------------------------------
>
>
>
>
> [Non-text portions of this message have been removed]
>
>
>
> ------------------------------------
>
>
>
>
• Hello Folks,   Here is a reprocessed image of the edge-on sprial galaxy ngc4546.  This was image data taken last year with a TMB152 f8 APO and ST10xme.  The
Message 39 of 39 , Apr 14 4:04 PM
Hello Folks,

Here is a reprocessed image of the edge-on sprial galaxy ngc4546.  This was image data taken last year with a TMB152 f8 APO and ST10xme.  The image sequence was 6 x 900 seconds each RGB using Gen I filters.

My goal in the reprocess was to bring out more detail in the galaxy, and to reveal the many background galaxies in this composition.  I hope you enjoy the finished result.

http://imagingtheheavens.com/?page_id=101

Clear Skies,
Mike Sherick

[Non-text portions of this message have been removed]
Your message has been successfully submitted and would be delivered to recipients shortly.