StarTools

Hi Ivo and fellow ST users. What is the best way to process OSC data taken with one of these new multiband filters (Ha, O3,S2 for example). In the documentation, it’s recommended to take L data through the filter and then RGB data without it. What’s unclear is if that applies to DSLR or other OSC data. This would help with color balancing. Please advise.

Also I have a small pixel (3.76 u) OSC camera and will be using a long focal length (2128mm) Dall Kirkham scope. I expect to be significantly oversampled. But I can’t bin the camera without losing color. Will the ST binning be a good idea?

Thanks!

Wayne

Hi Wayne,

The 1.5 and *especially* 1.6 versions have you covered if you use a Duo/Quad band filter.
There are two ways you can import your dataset. The best and most flexible way, is to use the Compose module to import your dataset like this;
Set 'Luminance, Color' to 'L + Synthetic L from RGB, RGB', then load your dataset three times, once for red, once for green, once for blue.
If your dataset was acquired with an OSC or DSLR with a Bayer matrix (and was not white balanced), double the Green exposure time (it is twice as reliable as blue and red, as there are 2 green samples per 2x2 patch versus just one each for blue and red).

Now process your image as normal (it will be mono, until you hit the Color module); it has been separated into synthetic luminance, while keeping color intact as a separate dataset that is processed simultaneously.
Once you hit the Color module, you can use the Duoband preset, and in 1.6, you can use the Matrix options to effortlessly cycle through all possible Duoband renditions, while keeping the luminance data intact. E.g. you don't have to decide in advance how you will map your color channels - it is done on the fly for you, while not impacting the luminance (detail) one bit.

Another way to import your dataset is to just use the second open when opening it through the 'Open' button in the homescreen. It is essentially a shortcut for the manual process outlined above. The one disadvantage here, however, is that the Matrix option in 1.6 will then show camera models for the purpose of color matrix correction (as it assumes visual spectrum data), rather than list channel remappings (which is only available for complex composites and narrowband data).

Software binning these days, almost gives you the same benefits as hardware binning. Of course, hardware binning gives you the advantage of reducing read noise as well, but in modern sensors read noise has become very, very low.
Software binning (in StarTools that is, using its fractional binning algorithms) allows you to bin to exactly the resolving limit of your dataset and no further (e.g. you're not stuck with fixed integer sizes like 2x2, 3x3 or 4x4). Besides the diffraction limit and your sensor's characteristics, atmospheric seeing is also a big factor in determining how far to bin and make use of oversampling.

It is also worth noting that a small degree of oversampling may be desirable if you have good, clean data and wish to use deconvolution.

Hope this helps. Any trouble, questions, let me know!

Thanks Ivo, sounds great. I’ve ordered the “early bird” model of the new QHY sCMOS OSC cameras, the QHY268C, with Sony IMX571 BSI APS-C sensor. Super sensitive, 16-bit ADC, very clean, no amp glow. Can hardly wait to try it!

Now it sounds like the process you describe would be great for narrowband targets where all the subs are taken through a Duoband-type filter. In the ST documentation You suggested another process for normal RGB targets. You’d take subs through the filter for a luminance channel then take a number of subs without the filter for the RGB channels. Would that work with an OSC camera? Maybe you could provide some details if you have time!

Thanks again, Wayne

whixson wrote:Thanks Ivo, sounds great. I’ve ordered the “early bird” model of the new QHY sCMOS OSC cameras, the QHY268C, with Sony IMX571 BSI APS-C sensor. Super sensitive, 16-bit ADC, very clean, no amp glow. Can hardly wait to try it!

Ooooh! Nice - let us know your findings!

In the ST documentation You suggested another process for normal RGB targets. You’d take subs through the filter for a luminance channel then take a number of subs without the filter for the RGB channels. Would that work with an OSC camera? Maybe you could provide some details if you have time!

Correct! First, you create a synthetic luminance frame from the filtered dataset. Using the Compose module load that dataset as red, as green and as blue. If you did not whitebalance that dataset yet, set the exposure time of the green channel to twice that of the blue and red channels. Set 'Luminance, Color' to 'L + Synthetic L from RGB, Mono'. Keep this image and save it. This is now your new 'L' dataset - save it. Next, launch the Compose module again, load the dataset you just saved as Luminance, while loading the non-filtered dataset as red, green and blue. Choose 'L, RGB' for 'Luminance, Color'.

You are now ready to process a dataset that truly has the best of both worlds; the correct coloring of a non-filtered set AND the depth of a light pollution-filtered dataset.

Ivo, for all OSC images should I set the Green exposure to double?

whixson wrote:Ivo, for all OSC images should I set the Green exposure to double?

That would indeed be the general recommendation.

Again, the rationale is that the green channel has twice the amount of samples per 2x2 patch. There are 2 green samples per 2x2 patch versus just one each for blue and red.
This only holds true if;

• Your OSC (or DSLR) employs a Bayer filter.
• The stacker and its debayering algorithm have not, in any way, cross contaminated the channels. Cross contamination may occur when using more "advanced" debayering algorithms that "borrow" information from the other channels to interpolate the channel being debayered. VNG or blinear interpolation do not cross-contaminate. Cross contamination my also occur when the stacker applies a camera color matrix correction.
• The stacker has not white balanced the dataset yet. White balancing would otherwise have multiplied the signal by an unknown factor in the individual channels. This means that the 2x samples green sample vs blue and red assumption (equalling ~1.4x better signal) can no longer be made.

Hope this helps!