Using different datasets to create an image

[kirkster501]

Hi, how does one combine completely separate datasets please (each set containing lights, flats, BIAS etc) that may be of differing exposures etc?  Does one register and integrate them as complete images in their own right?  For instance you could have a finished image for night1 another finished image for night 2 (could be years later) and then register and integrate these two completed images?  Does that make sense?

Thanks, Steve

[davew]

When doing narrowband I have no option. I never finish in one night !

Let Registar become your friend,

Dave.

[kirkster501]

Sure thing.  But how do you combine the datasets?

[davew]

Download the trial.

Then, import all your images and register them. Read the notes and decide how to crop if necessary.

If a mono image, save each file and load into PS. Stack and save the PSD file. Work on it. ( Recommended by me )

Or, after registering you can stack with user chosen weights ( Not recommended by me )

If a colour image is the final output ( IE. Ha/OIII/OIII ) you can stack as an RGB image.

If an LRGB image is required then you could stack the RGB and save the L separately. It goes on as much as you want it to. Have a play and decide if it's for you.

Dave.

[steppenwolf]

1. Align each master data set using MaxIm DL or RegiStar (other software will also align).

2. Open all (three aligned) datasets in PhotoShop

3. Select the 'Red' Data and type Ctrl A followed by Ctrl C

4. Select File - New and set the format to RGB

5. Select 'Channels' and choose Red. Type Ctrl V to paste the 'Red' data into the red channel

6. Select the 'Green' Data and type Ctrl A followed by Ctrl C

7. Re-select the RGB file, choose the Green channel and Type Ctrl V to paste the 'Green' data into the green channel

8. Repeat for the blue channel

Voila, one RGB image (although the actual data my be Ha, OIII, Hb or SII data rather than actual R G B data depending on which filters you used.

[kirkster501]

Ok guys thanks.  Gee whiz though.....  If the images are at the same scale can you not just stack two or three completed images to make one "super" image?

[Gina]

Ok guys thanks.  Gee whiz though.....  If the images are at the same scale can you not just stack two or three completed images to make one "super" image?

Sometimes but I find there is often a small misalignment between the datasets and then RegiStar puts it right

[davew]

Ok guys thanks.  Gee whiz though.....  If the images are at the same scale can you not just stack two or three completed images to make one "super" image?

You could. The original question did mention completed images over a few years though.

Just average the two images if they are roughly the same size etc.

There are at least two ways to do anything. Simple and complex. Who enjoys simple ! 

Dave.

[kirkster501]

There are at least two ways to do anything. Simple and complex. Who enjoys simple ! 

Dave.

..I'll settle for simple thankyou  I am on a steep enough learning curve in multiple areas as it is!

[ollypenrice]

I often combine 'finished' images from different years, either at the same scale or totally different scales. In order to judge how to weight them I 'Registar' them together but don't combine them. I take both co-registered images into Ps, paste one onto the other, zoom in to see the noise level and choose an opacity - effectively a weighting - which gives the lowest noise. Once flattened this 'super image' can then be stretched and sharpened further than was possible in its component images.

Usually the images will need adjusting to make them work well together. You can measure the RGB values at the same place in each image and use levels and curves to bring them close. To be honest I do this a lot and have developed a reasonable eye for it so I can usually guess what needs doing, but I measure when I can't guess it.

Olly

[Gina]

Yes, I've done this too

[kirkster501]

Yes, great tips there. I think I need to get this Registar software.  Gee whiz, not exactly cheap is it?  £100 just for aligning images.  Boy....

[ollypenrice]

Yes, great tips there. I think I need to get this Registar software.  Gee whiz, not exactly cheap is it?  £100 just for aligning images.  Boy....

No, any old software can align images. Registar aligns, resizes and re-curves images of different field curvature. Now that's clever! There's no such thing as a flat field and this is where Registar struts its stuff.

You can take images from any focal length, any pixel size, any field curvature and combine them. To the best of my knowledge this is unique. It also does it with the utmost simplicity. Given the number of times I nip in and out of Registar to make this fit with that, etc etc, I'd consider it the bargain of the century. I literally couldn't manage without it with my workflow. Sometimes when I combine RG and B in Astro Art the red channel is nearly but not quite correctly aligned. Registar! I've seen the same sort of errors in guests' images from other software. Split the channels, Registar them, recombine them and the picture can be so much better. Then you can can get your stars looking great. This much I do know; I wish I were on commission!

Olly

[Gina]

I too consider RegiStar to be good value for money.  When you consider how much the hardware costs, the software is not expensive.  Just because you can't touch it doesn't mean it isn't part of your kit - a very important part.  Without software you couldn't do anything.

[kirkster501]

Guys, I take it these principles would apply for different datasets on exactly the same equipment but on different nights?  For example, I got a ton of subs on M33 Monday night and I have the rig running again tonight.  I cannot mix them up and calibrate all these frames together in one go because the flats will be different right?

So I have to go through the Registar?

[Gina]

What I do is to take my data sets from different scopes or lenses and firstly stack all from the same scope/lens that match the flats with DSS (other stacking software is available )  then I align all the separate stacks with RegiStar.  That matches up the star parrtens and puts all the data on the same size canvas.  So if there is misalignment there will be extra margins on different sides of the originally misaligned images.  The result is that when all the images are combined in Ps or other processing softare they align perfectly.

I use Union mode in RegiStar which increases the canvas size to cover all images.  When the data is combined and given it's appropriate colour there are small different coloured areas around the edge of the canvas where there used to be misalignment.  Once processed I crop the final image to remove these leaving an area where all origonal image frames overlap.

I hope this all makes sense.  I think there are videos of using RegiStar and the subsequent processing but I don't have any links ATM.  RegiStar allows you to take data sets from different cameras with different resolutions and even width to height ratios (aspect ratio) and combine them.  The thing to watch is that most of each set covers roughly the same area of sky.  Pixel sizes and focal lengths etc. in themselves don't matter but if your sky areas are different then only the smallest will match with the others and you will be wasting CCD real estate.

If anyone would like examples of this, I have Ha images taken with a 135mm lens and Atik 460EX combined with OIII and SII images talken with 314L+ cameras and 105mm lenses.  Pixel sizes differ and sensor sizes by 30% also.  Even the aspect ratios are different but something like 80-90% of the area is common because the smaller sensor of the 314L+ is matched roughly by the shorter focal length of lens and the FOV only differes by some 10%.

[kirkster501]

So following that procedure Gina it would not matter if the exposures were different lengths on differing imaging runs?  Ultimately it would all be additive anyway?

[Gina]

So following that procedure Gina it would not matter if the exposures were different lengths on differing imaging runs?  Ultimately it would all be additive anyway?

That's right

[kirkster501]

Well, thanks Gina.  Worth checking into Registar.

However for subs from the same scope and camera combination and roughly lined up at the time of capture, in Pixinsight I think I could calibrate each set of subs with their respective flats and BIAS.  I will then register them all with respect to one master.  I could then integrate them all to create a "superframe". I then repeat for all channels.  They then wouldn't need to have the same exposures, although I tend to work at five minute subs at the moment.  I may start to increase this to ten mins.

Does that sound sensible?

Thanks, Steve

[ollypenrice]

In my experience flats deal with everything from the filterwheel backwards. That means that if you rotate the filterwheel and all that is behind it as one, your flats will still apply. However, if your focuser is not perfectly aligned, this won't be the case. Assymetrical vignetting will get you. Try it and see. I only use one set of flats for all colours quite simply because it makes no difference. If you have a lump of chewing gum on your red filter this won't work!!!   I can rotate the camera in the scope's rotator and still use the same flats.

Olly

[kirkster501]

I have sorted it with Pixinsight.

I let the Batch Pre-processing script chew through each individual set of subs and spit out the calibrated lights.  If I do this for separate imaging runs I will get the equivalent number of sets of calibrated lights.

Using the registration tool I then register all of these calibrated lights with respect to one luminance from one set of lights.  That will then spit out all these calibrated and registered lights into a new directory I created.  I then integrate them all together to create a "master green", "master luminance" etc.

Works really well!  Of course, they are all from the same scope/camera combo.  I'd imagine if using different cameras then thats where Registar struts it's stuff.

My post processing skills are very much a work in progress so I am embarrassed to post my [now] six hours of data on M33     At least, not just yet....