Stacking help

[Purplepanda]

Hello! I recently took 33 images of the Milky Way, just light frames. I stacked them in Photoshop with a median filter and got purple vertical strips, like banding or walking noise. Thus I've come to pixinsight hoping it can help me. What staying parameters should I use in this case? Or is there anything to remove the streaks? Thanks!

P.S. I hope I'm not going against forum rules on posting as this image was processed using Lightroom, but I need it to explain what I mean...

http://docs.google.com/viewer?url=https://www.freepdfconvert.com/result/downloadfile/eb33d706-b6d0-4620-9e44-48170c79a30a

[Niall Saunders]

Hi, and welcome aboard!

First - what kind of overlap do you have between the 33 separate Lights? It may be that one of the PixInsight mosaic-generation tools might be better able to stack your images with less artefacts than other software.

[Purplepanda]

Hi, and welcome aboard!

First - what kind of overlap do you have between the 33 separate Lights? It may be that one of the PixInsight mosaic-generation tools might be better able to stack your images with less artefacts than other software.

Hello Niall!! Thanks for responding I'm not sure what you mean, but I did take the images from a static camera (no dithering) and then just warp the images (free transform) the others to match a base image. I just downloaded a trial version of Pixinsight, how will I go about with the mosaic generation?

[Purplepanda]

Ok, so I just tried using Pixinsight, and I'm not really sure what I'm doing. I used star aligning and aligned all the light frames, which gave me many .xisf files, which I loaded into image integration. Is this correct? While doing image integration, I used the pixel rejection 1 (linear fit clipping). I ended up with 3 images, one of which is the actual file and after using auto stretch on it, I realise it still exhibits the 'banding' pattern. The banding pattern is seen in the high pixel rejection file. Question is, what do I do after this? Is there anything I can do to get rid of the banding?

I've attached screenshots, all of which have been auto 'stretched'

Any kind of help will be greatly appreciated 

[Niall Saunders]

Hi again,

Pne of the PixInsight tools that you might want to consider is GradientMergeMosaic

This tool requires a set of previously aligned (mosaic) images, and aims to eliminate 'image-edge' artefacts as each image is added to the mosaic.

When I looked (albeit briefly) at the images you attached in your previous post, I felt that you had not achieved full star alignment, or at least that there was room for improvement. Have another look at the StarAlignment process, and concentrate on using "Register/Union - Mosaic" as the working mode.

Both of the processes I have mentioned do have a reasoably detailed suite of documentation (unfortunately, not always the case for every tool in the PI arsenal), so spend some time getting your head round the many options available.

To begin with. it is also worthwhile considering simply joining two images together. If this seems successful, you can use the result of the first union as a base image onto which you can try to add a third image - and so on.

There is also the MosaicPlanner script (under Script | Utilities), but this requires a base image which, in turn, must have been plate-solved using the ImageSolver script (under Script | Image Analysis).

However - please understand that, although I am vaguely aware of the various steps required, I have never taken the time to actually perform a full mosaic operation. I did mess around with an M81 / M82 data set, and it was as successful as the dreadful vignetting (produced by my then OTA/CCD combo) would allow.

Stick with it, you are actually taking a leap into one of the deeper pools of PI - which is perhaps not one of the easiest introductions to PixInsight (although it is certainly one way to learn a lot in a short period of time) 

[Purplepanda]

Really appreciate you taking your time to help out Niall  even though I don't understand everything you said (I'm spending heaps of time reading up online), I feel like we may be talking about different things. My image was not really a mosaic, but 33 images at 28mm focal length, just stacked on top each other and aligned. Is this what you are referring to?

I found out after reading tons of stuff online that this is known as correlated noise?

P.s. I basically gave up trying just now, but after I saw your post I think I'll better keep trying!

[Niall Saunders]

OK - for some reason my poor addles braincell had decided that you were working with a group of images requiring a mosaic to create your desired effect. Now that I realise I am an idiot  , life should actually be a little bit easier for you.

Your challenge is then based on the following initial steps:

• <StarAlignment 1> : Find enough stars in each image to allow subsequent star alignment
• <StarAlignment 2> : Align all 33 images in your dataset against one of the 33, as a reference image
• <ImageIntegration> : Stack all of the aligned images to create an initial master image
• <DynamicCrop> : Process the master image to remove the edges where alignment and integration did not have suffucient image data
• <DynamicBackgroundExtraction (or ABE)> : Process to eliminate background colour casts
• <BackgroundNeutralization> : Process to neutralise the background colour
• <ColorCalibration> : Process to neutralise the foreground colour
• <Histogram> : Process to clip to the edge of the shadows and highlights, if possible

If you can perform these steps effectively, in the order I am suggesting, you will have a good starting point for further processing. However, I am not convinced that you have the first step under control yet and so, without that, all subsequent steps become unviable (in my opinion).

Don't give up - the worst-case scenario is that you may want to re-shoot the raw data (and also acquire basic calibration frames at the same time). I has taken me 13 years (!!!) to finally create an image that I consider 'adequate'.

Astroimaging is not for the faint-hearted 

[bulrichl]

Hi "Purplepanda",

You wrote that you took "just light frames" and "did take the images from a static camera". That means: DSLR camera, no tracking/guiding, no calibration of the light frames - right? Then:

1) The hot pixels are not removed from the subframes because you didn't take dark frames and didn't calibrate the light frames. Hot pixels are a severe problem when the camera temperature is high. So you will get better results in winter at lower ambient temperature.

2) What you see in the integration is not "banding". Due to the apparent movement of the stars  the hot pixels are at a different location in each subframe after registration. So when integrating, the unsightly traces are produced. You can see that in the integration result: the traces are circle arcs whose center (the pole) is situated out of the frame.

You didn't indicate in which format you took the light frames: in Raw format or in Jpeg format. If the light frames are in Raw format you can improve the result by taking some dark frames (also about 30):

Take (of course with the same camera) the dark frames also in Raw format, with the same ISO setting, the same exposure time and at the same temperature as with the light frames. Before aligning the subframes they must be calibrated. Then your workflow should be roughly:

1) Make the master dark from your dark frames in Raw format (by plain integration),
2) ImageCalibration of the light frames in Raw format with the master dark,
3) Debayer the calibrated frames,
3) ImageRegistration (= Aligning) of the calibrated, debayered frames,
4) ImageIntegration of the aligned frames.

If your light frames are in Jpeg format you cannot do much with them in order to improve the result.

On principal, better results are obtainable with the camera mounted on a polar aligned mount which tracks the apparent star movement, even better with guiding, and optimal with guiding and dithering.

Good luck!
Bernd

[Purplepanda]

OK - for some reason my poor addles braincell had decided that you were working with a group of images requiring a mosaic to create your desired effect. Now that I realise I am an idiot  , life should actually be a little bit easier for you.

Your challenge is then based on the following initial steps:

• <StarAlignment 1> : Find enough stars in each image to allow subsequent star alignment
• <StarAlignment 2> : Align all 33 images in your dataset against one of the 33, as a reference image
• <ImageIntegration> : Stack all of the aligned images to create an initial master image
• <DynamicCrop> : Process the master image to remove the edges where alignment and integration did not have suffucient image data
• <DynamicBackgroundExtraction (or ABE)> : Process to eliminate background colour casts
• <BackgroundNeutralization> : Process to neutralise the background colour
• <ColorCalibration> : Process to neutralise the foreground colour
• <Histogram> : Process to clip to the edge of the shadows and highlights, if possible

If you can perform these steps effectively, in the order I am suggesting, you will have a good starting point for further processing. However, I am not convinced that you have the first step under control yet and so, without that, all subsequent steps become unviable (in my opinion).

Don't give up - the worst-case scenario is that you may want to re-shoot the raw data (and also acquire basic calibration frames at the same time). I has taken me 13 years (!!!) to finally create an image that I consider 'adequate'.

Astroimaging is not for the faint-hearted 

Hi Niall, when I did my star alignment I only did the second step, what's the first step and how do I go about doing it? Can't seem to find it online

Cheers!!

[Purplepanda]

Hi "Purplepanda",

You wrote that you took "just light frames" and "did take the images from a static camera". That means: DSLR camera, no tracking/guiding, no calibration of the light frames - right? Then:

1) The hot pixels are not removed from the subframes because you didn't take dark frames and didn't calibrate the light frames. Hot pixels are a severe problem when the camera temperature is high. So you will get better results in winter at lower ambient temperature.

2) What you see in the integration is not "banding". Due to the apparent movement of the stars  the hot pixels are at a different location in each subframe after registration. So when integrating, the unsightly traces are produced. You can see that in the integration result: the traces are circle arcs whose center (the pole) is situated out of the frame.

You didn't indicate in which format you took the light frames: in Raw format or in Jpeg format. If the light frames are in Raw format you can improve the result by taking some dark frames (also about 30):

Take (of course with the same camera) the dark frames also in Raw format, with the same ISO setting, the same exposure time and at the same temperature as with the light frames. Before aligning the subframes they must be calibrated. Then your workflow should be roughly:

1) Make the master dark from your dark frames in Raw format (by plain integration),
2) ImageCalibration of the light frames in Raw format with the master dark,
3) Debayer the calibrated frames,
3) ImageRegistration (= Aligning) of the calibrated, debayered frames,
4) ImageIntegration of the aligned frames.

If your light frames are in Jpeg format you cannot do much with them in order to improve the result.

On principal, better results are obtainable with the camera mounted on a polar aligned mount which tracks the apparent star movement, even better with guiding, and optimal with guiding and dithering.

Good luck!
Bernd

Hey there Bernd, thanks for the insight! That's right, in RAW, not tracking, no calibration nor dithering. Do you mean to say take the darks now? At the same temperature? I always thought they had to be done together with the lights.

Also, in future to avoid this issue, would it be sufficient to take lights and the other calibration frames, but without dithering or tracking?
 

[bulrichl]

Hey there Bernd, thanks for the insight! That's right, in RAW, not tracking, no calibration nor dithering. Do you mean to say take the darks now? At the same temperature? I always thought they had to be done together with the lights.

Yes, take the darks. In order to remove the hot pixels, these are the minimum calibration frames. It is OK to take the dark frames at a different date. The master dark is re-usable for the calibration of light frames with the same settings. However, the time distance between light frame and dark frame acquisition should not be too large because the sensor slightly alters by the time. Some people say that darkframes should not be older than about 6 months. Important: adjust all settings as at light frames acquisition (same camera, Raw format, same ISO, same exposure time) and also the ambient temperature must be near that at light frame acquisition.

Also, in future to avoid this issue, would it be sufficient to take lights and the other calibration frames, but without dithering or tracking?

It depends on the camera and the ambient temperature, you must try. If you don't have a mount available I would not go and buy one immediately. First of all you need to experience what is possible with your equipment and make the best of it. So I would go with the dark frames, look at the difference and see whether the improvement is sufficient. Maybe at a focal length of 28 mm tracking is not necessary.

But if you have a mount available, then use it! Tracking has the advantage that you can extend the exposure time of the subframes and set lower ISO which will diminish the noise in the subframes.

What camera, exposure time and ISO did you use, and what was the ambient temperature at light frame acquisition?

Bernd

[Purplepanda]

What camera, exposure time and ISO did you use, and what was the ambient temperature at light frame acquisition?

Bernd
[/quote]

Hi Bernd
, I just took 30 darks, temperature should be roughly the same, but might be slightly warmer, I'll take another set when it gets cooler. Also took some biases. 30 of each, and will stack them tomorrow, fingers crossed it works!!

I'm using an Olympus ep3 with the kit lens 14-42 ii. Exposure time 20s, ISO 2500 (stupid me setting it so high), temp, roughly 16 degrees Celsius.

If I'm in a really dark area, and exposure time is already maxed out before trailing occurs, aperture is at the widest and yet the Milky Way is still too dark on the camera screen, should I just leave the ISO at (for example 1600), or bump it up, but get more noise?

Thanks! I've learned so much from you guys 

[bulrichl]

I'm using an Olympus ep3 with the kit lens 14-42 ii. Exposure time 20s, ISO 2500 (stupid me setting it so high), temp, roughly 16 degrees Celsius.

If I'm in a really dark area, and exposure time is already maxed out before trailing occurs, aperture is at the widest and yet the Milky Way is still too dark on the camera screen, should I just leave the ISO at (for example 1600), or bump it up, but get more noise?

Higher ISO must not mean more noise. Have a look at the following websites which explain the wise ISO setting. The first two links explain it for short exposure time:
http://www.lonelyspeck.com/how-to-find-the-best-iso-for-astrophotography-dynamic-range-and-noise/
http://www.lonelyspeck.com/how-to-find-the-best-iso-for-astrophotography-dynamic-range-and-noise/

In your case (no tracking, very short exposure time, limited by the apparent star movement) use rather high ISO, not below 800. I don't know whether the Olympus EP3 belongs to "ISO-variant" or "ISO-invariant" group of cameras. Maybe your setting of 2500 is OK - you should try what is the proper setting for your camera.

However, the case is somewhat different for guided exposures with exposure time of a few minutes, which is explained in the third link:
http://dslr-astrophotography.com/iso-dslr-astrophotography/

In guided deep sky astrophotography one normally tries to set the longest possible exposure time [exception: cameras with extraordinary low read-out noise, but that are not DSLRs]. The exposure time can be limited e. g. by the quality of the guiding or by the full well capacity of the sensor (the bright stars in the frame should not be saturated yet). I think it is well explained in the links above.

Bernd

[Niall Saunders]

Hi again,

In reality, steps (1) and (2) in my descrition are both part of the same StarAlignment process.

However, as you start working with StarAlignment, you may need to repeatedly adjust parameters until the process is able to identify a reasonable number of stars from each and every one of your source images. That is iffectively stage (1).

Having done that, StarAlignment should then be able to complete with a successful alignment of all of your source data against your reference image. This is stage (2).

I hope that is clearer!

[Purplepanda]

Hello Niall, yep that clears it up. Thanks also to Bernd, il try that later during the day. I took my darks and biases yesterday, but when trying to use batch preprocessing, everything works fine until it tries to integrate light frame, which it says there's insufficient signal. Anyone know why? I've tried it a few times, but get the same message every time.   

http://docs.google.com/viewer?url=https://www.freepdfconvert.com/result/downloadfile/37b1519c-9d3c-43f1-8559-e9300ebdf656