Processing or stacking issue? Help needed please :)

[CosmicOrangutan]

Hi Everyone,

I'm currently working on a picture of M33 and am having some issues either when stacking or during post processing of the image. 

This is the first time i'm getting these artefacts appear on my images after stacking so i'm a bit confused. I have a feeling it might be to do with dithered subs and D.S.S as the amount of RGB dots is exactly the same as the number of subs stacked in the image. This said however i have stacked a few other pictures using dithered subs and the same stacking method without this issue. I'm also unsure about the waves of noise emanating from the centre of the galaxy out to the edge of the image. I have attached an image from pixinsight to show what i'm on about. All i have done on it is a quick DBE and then stretched it right out to accentuate the noise pattern and strange RGB dots to show you guys. I think it must be to do with stacking as the RGB dots are visible as soon as you transfer the image from D.S.S to pixinsight. I have also tried stacking with and without calibration frames(dark,bias and flat) to see if it makes a difference and it doesn't.

Any insight into these issue would be great and a remedy even better.

many thanks,

George  

[Freddie]

They are hot pixels in a slightly different position on each image so when stacked show that grouping. Your DSS stacking software has some cosmetic settings that should be able to remove them. Not used DSS for a while so can't remember exactly where those settings are, sorry.

[wimvb]

Why not do the stacking in PixInsight?

Use bias and flats and possibly dark frames.

To more aggressively remove the hot pixels in your image, use cosmetic correction. You can use the same master dark for cosmetic correction as for ordinary image calibration.

When doing image integration, use sigma clipping and decrease the high end sigma value. This will increase the clipping of hot pixels. That should remove the hot pixels.

With DBE, you have to be careful where to put samples. increase the sample size to a large value, and use only a few samples (usually less than 15). Place samples in each corner, along the edges and towards the middle. Be careful not to place samples over stars or nebula. Examine the extracted background before applying DBE.

When you apply DBE, you subtract the background signal, but not the associated noise. This will always give the impression that the image is noisier. In truth, the S/N is decreased, but only because the S part (signal) is decreased. The noise is the same as in the uncorrected image. The noise can be removed reduced by MultiscaleMedianTransform or TGVDenoise.

 

Good luck

[Uranium235]

When using DBE (for simple gradients), try this:

(first, crop the image to remove stacking artefacts)

1: Create a preview of the image, stretch that preview using levels so you bring out most of the stars.

2: Use the "make image from preview" function

3: On that new image, place sample points (I usually like to set a sample radius of 6, fairly small) in each of the four corners (avoiding stars). Then one centre top, bottom, left and right (remembering to avoid stars). That should give you a nice, smooth gradient model.

4: Drag that selection out as a process, cancel the current DBE process.

5: Go back to your unstretched image, reactivate dbe then drag the process you just created into the DBE window - that will place the markers you did for your stretched image.

6: Run DBE (sample output 1:1), that will create your gradient model. Save it off for future reference.

7: Subtract the created gradient model from your image using pixel math (Can be done in PI, Maxim or even Photoshop!)... job done!

 

Youre looking to create the smoothest gradient model possible, that way you are avoiding the "black holes" you sometimes get when placing sample points in the wrong places. Here is an example of a smooth model (data has been stretched for illustration):

As mentioned above, use Sigma clipping when stacking in DSS - but I would suggest you use no less than 16 subs for it (more is better).  Also, its best to avoid any cosmetic hot pixel reduction as in DSS that more often than not ruins an image.

 

[wimvb]

Good description of dbe. Just a few thoughts:

1 it's possible to duplicate an image by simply dragging the image tag (left hand side frame if an image) onto the workspace. New image gets the added "_clone" to its name.

2 why not just apply the dbe to the image? This must be much easier than saving the bg model and going to ps to subtract. Pi uses floating point numbers when doing the dbe process. If you save the bg image and open in another program, you will also convert it and lose accuracy. I usually leave the discard bg model box unchecked, and the replace image box checked. This will leave the image process history intact.

 

[Uranium235]

There is no reason why you cant do it all in PI, but I prefer (personally) to subtract in Maxim because the offset function is slightly better and it allows me to preview the result before applying the gradient model so i can see if its going to work or not. There is always more than one way to skin a cat   you just need to find what works for (and suits) you.

[AKB]

9 hours ago, wimvb said:

When you apply DBE, you subtract the background signal, but not the associated noise. This will always give the impression that the image is noisier. In truth, the S/N is decreased, but only because the S part (signal) is decreased. The noise is the same as in the uncorrected image.

I'd never thought of it this way before, more thinking that in the case of LP you're actually subtracting (smooth) noise. However, the visual impression is often as you say.  Thanks for the alternative POV!

 

[CosmicOrangutan]

These are some great tips, thanks very much guys! I will give these ideas a try this evening and post the results. I have had the odd hot pixel appear before but never in this pattern and to the same scale.

Once again cheers , George  

[ollypenrice]

You have two PI experts on the case. I defer to their considerable knowledge! All I'd say is that I doubt your background model was right when you ran DBE for the image you posted. There are dark patches around the galaxy which almost certainly come from markers placed too close to it. Something has also created that dark half-ellipse midway up the left hand side, possibly a marker placed too close to a star. When you look at the gradient model (which you should!) look for anything that resembles, in basic shape, your target. If you see such a shape the model is wrong, it has read the target as background sky.

Now, the vertical banding... You say you haven't seen it before but could this be because you've done a harder stretch than usual? We suffer from this here on one camera as well (mine as it happens!) It is cured at source by using longer subs. I can also cure it easily in Ps by running Noel Carboni's 'Reduce Vertical Banding' routine from his actions set, now available from Pro Digital Astronomy Tools. However, PI has a 'remove Canon banding' routine, I think. Might that help?

Olly

[Barry-Wilson]

You've got some great PI advice there George and plenty to read and then re-process your image.

The pre-processing scripts of PixInsight are excellent and I would recommend you learn how to use these in preference to DSS - you will get better results.

Plenty of folk here to help you with PI.

Good luck.