Local galaxies. 10s frame luminance to add sharpness to colour

[timh]

Following on from viewtopic.php?t=7711 have extended the 'semi-lucky imaging' approach of stacking multiple selected 10s frames followed by RC BlurExterminator to get high quality luminance with which to better inform the resolution of colour OSC images of M51, M101 and M106 galaxies (data collected previously at a lower resolution).

The luminance swap approach seems to work quite well although the colour images don't have quite the same 'punch' somehow as in grayscale -- the black and white images look better somehow and it seems that not quite all of the resolution has translated across and particularly in the case of M106.

The starting luminance images integrated from hundreds of selected 10s frames continue to have a slight star shape problem with the integrated images showing Eccentricity Av values between 0.45 and 0.7. This was consistent across the field and the long axis of this elongation always aligned close to the RA axis so I am pretty sure that it is a recurrence of a problem with an 2-3 px oscillation in the RA axis - which I hope maybe to suppress with heavier East loading. RC Blur Exterminator is amazing software though. The star correction process seems very effective at dealing with this particular aberration and after non-stellar sharpening as well the final images appear to be resolving features separated by not much above an arc second -- i.e 3 pixels . So I am pretty happy with the resolution nonetheless.

Again the set up was a VX12 12 inch telescope F4 on a CEM70 mount with PDS guiding via a top mounted 80 mm refractor. The camera was an ASI294 MM at 46Mb BIN1 with 2.315 u pixels giving an image scale of 0.406 arcsec/ pixel. Subs were 10s at gain 200 using a ZWO UV/IR filter only.

The M106 luminance was derived --under Bortle 7 skies - from 1041 x 10s frames, the whirpool - M51a from 1326 x 10s frames and M101 from a total of 867 x10s files. The RGB information was obtained using the same set up but using an ASI 294MC camera at a lower resolution of 0.81 arcsec/ pixel and usually 40s frames at gain 124 (near unity) for total integration times of 2.5h, 12h and 3.8h respectively. M51a has the deeper colour information.

[turfpit]

Tim

I like the M51 colour image - well processed, star colours and a couple of small faint objects. The combination of mono and OSC data is very effective and works well in a Bortle 7 zone. Thanks for the details.

Dave

[Menno555]

Great captures Tim!
I really like M106 and especially the color. If you ever get the change/possibility, see if you can capture and incorporate Ha narrow-band data into it. It has some very interesting Ha features.

Menno

[timh]

Many thanks Dave and Menno.

Firstly I must confess that I swapped out and replaced the original M51 images with better ones which included even more luminance frames. The details of this and of the processing are appended below ..

There are two things here that I think play into two discussions meandering along over on Astrobin that I am interested in. The first is "short frames versus long frames" where there seem to be two schools of thought. One can perhaps be fairly summarized as "longer subframes are always better and stacking lots of short frames not a good idea " and the other essentially says just do the calculations based on sky brightness - as per Robin's model in fact - and act accordingly depending upon your particular objective. Here I think that using a higher gain and actually more than 3000 subs - some included within SC autostacks - has delivered not only resolution but also a level of SNR about commensurate with B7 skies.

The other thread - that I began - was to ask what folk's views are on the power of RC Blur Exterminator to correct and derive such astonishing detail from even rather flawed input data - and then, rather provocatively perhaps - to ask whether it is reaching the point where it is no longer worth the time or money to even fix certain types of equipment flaw (in my case probably RA motion blur) - although that feels very wrong. Everyone with any knowledge of RC Blur Exterminator - included Russell Cronan - affirms that better data in yields better data out - and that surely must be true. Nevertheless when I compare the detail at the core of M51a with that visible in the publicly available HST image of the same my BlurXt sharpened B7 image does show nearly all of the correct detail at not much over an arcsec resolution. So then I am left wondering how much headroom is left for further improvement in resolution? Better input data might be produce better final images - but how perceptible would the improvement be?

It would be fascinating to better understand how the neural net aspect of image processing works in BlurXT - speculatively there is a huge library of 'typical' DSO features broken down into a tiny subscale of fragments that are substituted into an image where-ever the program recognizes a blurred version of any one of these sub-pieces according to whatever Blur function is calculated as consistent for a given image via deconvolution?

To Menno's comment - the great thing that HST images do show that is more or less invisible to just luminance -and certainly especially with M106 - is all the interest and detail in HA. It is one thing missing from the current luminance approach that I have adopted and I'd love to experiment now with just what it is possible to get away with in terms of sub length, gain, filter width in capturing HA. Of course once again it is calculable but I suspect it will be a case of high gain, subs at least 30s and probably sampling at 0.81 rather than 0.406 to get HA data that could never be quite so sharp to now add in. An interesting objective anyway!

Tim




--------------------------------------------------
The luminance for the M51a image comprised 1900 x 3s subs at gain 285 ASI294 MM acquired as a number of autostacks in Sharpcap using selection based on FWHM and Brightness, 664 x 10s subs and 662 x10s subs at gain 200 ASI294MM. The image scale was 0.406 arcsec/ pixel with 2.315 u pixel size selected. All these about eight separate integrated images from different sessions were separately cropped, background-gradient corrected using the gradient and DBE tools in PixInsight before correction using RC Blur Exterminator followed by non-stellar sharpening by ca 0.65X. The starting images had FWHM values estimated between 1.8 and 2.4 - Eccentricities were not so good varying from about 0.35 all the way up to about 0.69. In each case the Eccentricity was aligned close to the RA axis and similar across the entire image consistent with some RA motion blur. RC Blur Exterminator correction appeared very effective at restoring star shape from this particular source of aberration. The corrected and sharpened images were then all aligned and then combined into a single image using PixMath being added in proportion to their relative PSFsignal weights as estimated in Subframe Selector. The final luminance image combining contributions from more than 3000 short frames was then stretched using Histogram transformation, exponential and arcsin functions before noise reduction with RC NoiseExterminator (0.3) and black level, contrast adjustment in Affinity.

The RGB image was built from many 40s subs at gain 124 using an ASI 294MC OSC at an image scale of 0.81 arcsec/ pixel . Integrations were 1x CFA drizzle integrations, gradient subtracted and colour corrected using SPCC and aligned and combined together using PixMath before stretching using Histrogram Transformation. Finally the image was aligned with the luminance image obtained using the mono camera and the mono luminance transferred using the LRGB process.

[turfpit]

Tim

Just needs some Ha now.

Dave