You’ve waited for an hour or more while your subs were getting calibrated and integrated, and now you have your mono integrated stacks that you want to recombine into an RGB image. You get your first glimpse of the colour image and all hope seeps out of you. Your will to live is diminished. What the holy hell has gone wrong?
There can be any number of things that can really mess with you at this point, so I’m going to try to work through some of them, and how to correct them if you can.
Redo Registration
There are a few clues in this image that identify what has happened here:
- Stars that are close together, look similar, but are in different colours
- Elements of the nebula / structure that appear to be shadowed / out of sync, or like some weird attempt at making a 3D image in the 80s.
- In this one we can also see a slight rotation effect, like there’s a whirlpool forming and things are starting to spin around.
This all points to a failure of registration for the integrated stacks – that is, your stacks individually fine, but they are not lined up against one another, so when you add them together it gets all whack. (Yes that’s a technical term.)
Fixing this in PixInsight is fairly quick and painless, at least most of the time.
1. Open up the Star Alignment tool
2. Drag the label of the best stack (the one with the best stars) into the Reference Image parameter – or you can just type in the image identifier if it is easy (like in this case, the identifier is ‘H’ for the Hydrogen Alpha layer)
3. Drag the triangle icon from the bottom of the star alignment tool onto one of the other master stacks
4. If registration can be made, a new image will appear in your PixInsight workspace, with the suffix ‘_registered’
Your new image may have black lines on some of the edges, and may appear slightly lopsided or not straight within the window. This is correct behaviour – PixInsight has discovered that to align the O layer with the H layer, the O layer needed to be rotated say 3 degrees to the right. This gives you the black border effect, because the O stack does not have data in those areas. You can fix this in your workflow by cropping.
5. Repeat this for the remaining layer or layers, not counting the one you chose as your main image to register against. In this example we have a narrowband image with H, S, and O masters to combine. Like the O layer above, PixInsight has determined that it needs to rotate the S layer a similar amount to get it lined up.
6. At this point, assuming that your masters are already saved, I would close down the original copy of each of the layers you re-registered, and then rename the registered versions to align with your usual naming convention. For me, that means I have an H, an S, and an O image.
7. When you combine your layers however you normally do (LRGBComposition tool, PixelMath tool) you will now find that the stars all look nice and are generally white, at least the centre of the stars will be pretty close to white. Also we have lost the weird 3D glasses effect, the features of the nebula are crisper. We still have weird dark bits on the borders, because of the three channels we captured, only 1 or 2 have provided any signal in those areas. Crop tool will be your friend!
But why did this happen?
There are two factors that combine to cause this: camera rotation, and not doing a combined registration.
During data capture the camera angle has changed, meaning some features are present in some subs and not present in others. This is best seen in the corners, where stars are out of the frame in one set and in frame in the other. Camera rotation is not necessarily a problem – it can happen all the time depending on your setup and how you use it. This is also common if you’re working with data captured over a number of different sessions / nights, or if you’re doing a collaboration with other astrophotographers. It’s not something I particularly stress about, because star registration is very good at overcoming it.
Depending on your workflow, you may find yourself registering and then integrating the different channels separately. Integrating the channels individually isn’t the problem, but not having them all registered against the same master is definitely something to avoid when you can. If you are preprocessing with PixInsight’s weighted batch preprocessing tool, you shouldn’t experience this problem but it is good to know that you can also tell it which image to register against. In the bottom right corner of the WBPP window is the Registration Reference Image configuration which normally just stays on auto; but you can choose manual and then give it a specific image to register against. I like to do this because I hate wasting integration and processing time on bits with no data! To speed things up, even if only slightly, I construct a new image to register against that only contains good data (no black bars).
1. With your master channel (integrated and registered stacks) open in PixInsight, open up the DynamicCrop tool
2. Pick one of your masters that have been rotated, and use dynamic crop to grab the area that has good signal, but excluding the areas of black but DO NOT HIT APPLY
3. Grab the New Instance triangle icon at the bottom of the DynamicCrop tool, and drag it onto your PixInsight workspace. This will create a new copy of this DynamicCrop window.
4. Now that you have a new icon on your Workspace (Process18 in my case), you can hit the red X in the DynamicCrop tool to cancel it on the master you have used to define the box (S in this case).
5. Click on / select one of the other masters that has black borders on the edges, and then double click the process icon you created and it will configure the DynamicCrop tool with the same box definition, which you can now compare against another channel.
You will see that although the box grabbed only good data in the S master, here in the O master we still have a black strip of no data down the right hand edge. So we will have to shrink our dynamic crop box further. Thankfully this is easy to do!
6. Firstly, right click on the new process icon you created with the dynamic crop configuration, and choose to delete it. We don’t want use that one any more.
7. Now modify the crop rectangle to contain only signal areas. Try to only reduce the crop rectangle, not move it. Note the cursor has changed to a white square with a solid little rectangle on one side – this is telling you that click-dragging from this point will allow you to move that side of the crop rectangle in or out. If you hover the cursor over the corner of the box, you can grab both sides on that corner and adjust them at the same time.
8. Once you have the rectangle redefined so that it only contains good data, jump back to step 3: save this new DynamicCrop configuration to the workspace, and then check it against all the other masters to make sure that it is only selecting good data in all of them.
9. Having confirmed that your crop rectangle is good on all the channels you want to use / work with, open it up against your strongest image (the H master in this instance). Apply the crop to this image, and save the resulting image out as your new registration master.
Now that you have a registration master image which covers the area with signal across all your data, you will want to register your images against this image. I tend to rerun WBPP with this image, just in case the areas of no signal mess around with the calculation of weightings of the individual subs and wind up applying more weight to the wrong images. To achieve this, simply put the registration configuration to Manual, and load up the master registration file from wherever you saved it to.
Tips
- Pressing CTRL + SHIFT + S saves the currently selected image
- If you want to use the DynamicCrop tool but you’re struggling to get a handle on the sides and wind up moving the window instead, zooming out one step (click the scroll wheel on your mouse one click forwards / away from you) will give you a much easier view to work with
- To rotate the crop rectangle, move the cursor off the image itself and into the dead space around it, and the cursor changes to a symbol with an arrow around it to signify rotation
