Pre-Processing images in Nebulosity.pdf

Pre‐Processing images in Nebulosity 

Craig Stark 

You've taken your images and are now comfortably inside. Now what? How do you 

get all those raw frames to look like a nice pretty stack? Just what the heck is Bad 

Pixel Mapping? Should I try Drizzle?  

The rest of the manual provides answers to many individual questions and 

documents each of the tools. The goal of this section is to let you see how all of these 

it together and to give you the necessary information to choose a path through the 

initial processing of your data. This alone won't give you a full understanding of how 

each tool works (see the individual section in the online manual for each tool), but it 

should help put all the pieces together.  

The basic steps are as follows:  

1. Prepare   any   sets   of   darks,   lats   or   bias   frames   for   use   by   stacking   them   

2. Take   care   of   hot   pixels   (dark   subtraction   or   Bad   Pixel   Mapping),   bias   signals,  

and/or   vignetting   ( lats)   

3. (optional)   Normalize   the   images   

4. Convert   RAW   images   into   color   via   Demosaic   (if   one‐shot   CCD   used   and  

captured   in   RAW,   which   you   really   should   do ) and square‐up your pixels (if 

needed)  

5. (optional)   Grading   and   Removing   Frames   

6. Stack   the   images   (Align   and   Combine)   

7. Crop   the   image   to   clean   it   up   

8. (color   only)   Run   the   Adjust   Color   Offset   tool   to   remove   skyglow   hue   

9. Stretch   the   image   (Levels,   DDP,   etc)   

The last three steps (crop, color offset, and stretch) are covered in more detail in the 

Post‐Processing   How‐To  document. 

Step 1. Preparing the darks, ﬂats, and biases 

If you've taken darks, lats, and/or bias frames for this imaging session, you'll need 

to put them together to form "master" darks, lats, and/or bias frames. If you've not 

got a new set of these, simply skip to the next step as there's nothing to do here. 

Assuming you do have some, what we need to do is take the set of them (e.g. 20 bias 

frames) and combine them so that you can use them to remove artifacts in your light 

frames. Having more than one dark, lat, and/or bias frame is a good thing as each 

individual frame has both the artifact you want to remove from your lights and 

random noise. Stack a bunch of these together and the random noise goes away 

leaving you with a clean image of the artifact you want to remove. Use just one and 

you remove the artifact and whatever random noise that one frame had.  Since it's 

random noise won't be the same as the random noise in your image, using just one 

dark, lat, or bias will actually inject noise into your light frame and make it noisier.  

This is why people take a good number (20‐100) of each of these.  

When stacking these, we  don't want the frames to move . That is, since there isn't a 

star whose motion we want to track, we don't want to align these images. We just 

want them stacked on top of each other as‐is. To do this:  

1. Pull down Processing, Align and Combine  

2. Select "None" for the Alignment method and keep it set to "Save stack" and 

"Average / Default"  

3. Click OK and then select all of your dark frames (or bias frames, or lat 

frames)  

4. When all are stacked, give the resulting combined dark frame a name like 

"master_dark" or "master_dark_1m" (1m being a code for 1 minute ‐ 

something to let you know what kind of master dark this is)  

5. Repeat for any other types you have (lats and/or biases)  

Ugly Details 

At this point, you've got nice stacks of each and the stacks can be ready to use. If you 

want the absolute cleanest pre‐processing and, it's worth considering the following 

issue.  Nebulosity 's pre‐processing just does the basic math for you. It subtracts the 

dark and bias from the image and divides this by the lat. It does not do anything to 

the bias, dark, and lat you pass in during Pre‐processing. It just uses them.  

So what's the problem? The problem is that that dark frame has the bias error in it 

already. The lat frame has the bias error and some amount of thermal noise in it 

(which will lead to hot pixels). So, if you use all of these as‐is, you're going to do 

things like subtract out the bias error twice, which will actually inject the reverse of 

the bias error (still noise) back into your image. Oops.  

The solution is to pre‐process your pre‐processing frames. You can, for example, 

apply the bias frame as the only pre‐processing step for pre‐processing your 

"master dark" and "master lat" frames. You can also have a dark frame taken at 

about the same exposure durtation as your lats and apply this to the lats. Before 

fully going down this route, consider the following recommendations:  

Recommendaons 

•

If you are using normal dark subtraction and not Bad Pixel Mapping to 

address the hot pixels, your darks already have the bias error in them. Do not 

collect extra bias frames and do not use any bias frames during pre‐

processing. Just use the darks and both the dark current and the bias error 

will be removed.  

If using lats, it is worth knowing that  Nebulosity  passes a mild smoothing 

ilter over your lat in any case (a 2x2 mean ilter). This will help remove hot 

pixels in the lat if your exposure duration was long enough to put them in 

there and will also remove some of the bias error. You may still remove the 

bias from this if you like, or simply pass something like the 3x3 median ilter 

over your lat to smooth it out prior to applying this to your light frames.  

•

If using Bad Pixel Mapping, consider using bias frames as well. There is no 

need to clean up your dark frame (i.e. remove it's bias error) as with BPM, 

only the very hot pixels are touched. The bias error in your dark frame is 

ignored completely. If your camera has a strong bias error, grab a stack of 

bias frames once (shortest exposure possible) and grab and stack a bunch of 

these (you only need to do this once). Call it a "master bias" or "uber‐master‐

bias" or whatever you like and apply this during pre‐processing (below).  

Step 2. Taking care of hot pixels, bias signals, and/or vigneng 

At this point, you should have "master" darks, lats, and/or bias frames. If you don't 

and you're processing without these, skip this step. Keep in mind, you can use as 

many of these as you want (or don't want). You can use darks but nothing else, lats 

and biases but not darks, etc. It's up to you and what type of pre‐processing images 

you actually have. If you've got a stack of darks to use, you have a choice to make.  

Dark subtracon or Bad Pixel Mapping?  

Both of these techniques are designed to deal with the thermal noise inherent in 

your images and the resulting "hot pixels" that show up in the same spot on the 

image in each frame. Dark subtraction is the traditional way of doing this. It works 

by simply subtracting the value for each pixel in your "master dark" from the value 

of that pixel in each light frame. If your light frames and dark frames were taken 

with the same exposure duration and at the same temperature, dark subtraction will 

remove the hot pixels (and "luke‐warm" pixels as well ‐ any thermal noise, not just 

the brightest). This can work very well  if you control the temperature, exposure 

duration, and take a lot of dark frames . If you don't do these, you can end up with 

"holes" in the image (black spots where the hot pixel used to be), incomplete hot 

pixel removal, and you can inject noise into your light frames (see above).  

Bad Pixel Mapping works differently. You irst create a "Bad Pixel Map" (Processing, 

Bad Pixels, Make Bad Pixel Map) using a dark frame or stack of dark frames. A slider 

appears to let you set a threshold (feel free to use the default). Values in the dark 

frame that are above the threshold say "this pixel is bad". Bad pixels, and only bad 

pixels are ixed in your light frames by using surrounding good pixels to help ill in 

what this pixel should have been. For many cameras (in my experience, the cooled 

cameras with Sony sensors work best), this is an exceptionally powerful technique 

as the hot pixels are removed effectively with no noise being injected. It's also very 

lexible as you can use the same "master dark" from night to night and from 

exposure duration to exposure duration just by adjusting the slider and making new 

maps as needed.  

Note: If you use Bad Pixel Mapping you will not use Dark Subtraction and vice 

versa. One or the other but no need for both. If you use Bad Pixel Mapping you 

can still use lats and bias frames and it doesn't matter whether you apply 

BPM before or after your other preprocessing .  

Applying Bad Pixel Mapping 

To apply BPM to your light frames:  

1. Create a Bad Pixel Map if you don't already have one. Processing, Bad Pixels, 

Make Bad Pixel Map. Select a dark frame or stack and start off by just hitting 

OK to use the default threshold.  

2. Pull down Processing, Remove Bad Pixels, selecting the one for the kind of 

image you have. If you have a one‐shot color camera that is still in the RAW 

sensor format and looks like a greyscale image and not color (another reason 

to capture in RAW and not color...), select RAW color. If it's a mono CCD, 

select B&W. If it's already a color image, you can't use Bad Pixel Mapping.  

3. A dialog will appear asking you for your Bad Pixel Map. Select it.  

4. Another dialog will appear asking you for the light frames. Select all of them 

(shift‐click is handy here).  

5. You will end up with a set of light frames that have had the bad pixels 

removed. They will be called "bad_OriginalName.it" where OriginalName is 

whatever it used to be called.  

Applying Darks, Flats and Biases 

Here, you get to apply traditional dark subtraction, lats, and biases in any 

combination you wish. To do this:  

1. Pull down Processing, Pre‐Process Color images or Pre‐Process BW/RAW 

images. Color images are already full‐color. BW/RAW images were either 

taken on a monochrome camera (BW) or taken on a one‐shot color camera 

but have not yet been converted into full‐color via the Demosaic process.  

2. A dialog will appear that will let you select your various pre‐processing 

control frames (darks, lats, and/or biases). Select whichever you have by 

pressing the button and telling Nebulosity which ile to use here.  

3. If you are using dark subtraction and you doubt your exposure and/or 

temperature control was perfect, select the "Autoscale dark" option.  

4. Click OK and you will be asked to select the light frames you wish to pre‐

process.  

5. When all is done, you will have a set of iles called "pproc_OriginalName.it".  

Step 3. Normalize Images (oponal) 

All things being equal, your 50 frames of M101 should all have the same intensity. 

They were taken on the same night one right after the other and all had the same 

exposure duration. So, they should be equally bright, right? Yes, but there's that 

nagging "all things being equal" we supposed and, well, all things aren't always 

equal. For example if you start with M101 high in the sky and image for a few hours 

it starts picking up more skyglow as the session goes on, brightening the image up. 

That thin cloud that passed over did a number on a frame that still looks good and 

sharp, but isn't the same overall intensity as the others, etc. All things are not always 

equal.  

If you're doing the Average/Default method of stacking, you need not worry about 

this issue unless the changes are really quite severe. If you're using standard‐

deviation based stacking, Drizzle, or Colors in Motion, it is a good idea to  normalize  

your images before stacking. What this will do is to get all of the frames to have 

roughly the same brightness by removing differences in the background brightness 

and scaling across frames. To normalize a set of images, simply:  

1. Pull down Processing, Normalize images  

2. Select the light frames you want to normalize  

3. In the end, you'll have a set of images named "norm_OriginalName.it"  

Step 4. Converng RAW images to Color and/or Pixel Squaring 

(aka Reconstrucon) 

The last step before stacking your images is to convert them to color (if they are 

from a one‐shot color camera and you captured in RAW) and square them up as 

needed. Some cameras have pixels that are not square and this will lead to oval 

rather than round stars. The process of demosaic'ing (color reconstruction) and/or 

pixel squaring is called  Reconstruction  in  Nebulosity .  

Note, you can tell if your images need to be squared up by pulling down Image, 

Image Info. Near the bottom you will see the pixel size and either a (0) or (1). If it is 

(1), the pixels are square. Of course, the pixel dimensions will be the same in this 

case too.  

To reconstruct all of your light frames, simply:  

1. Pull down Processing, Batch Demosaic + Square (if images are from a one‐

shot color camera) or Batch Square (if images are from a monochrome 

camera or you just feel like squaring up a color cam's but keeping the image 

as monochrome for some reason).  

2. Select your frames  

In the end, you'll have a set of images named "recon_OriginalImage.it"  

Step 5. Grading and Removing Frames (oponal) 

Sometimes bad things happen. The tracking goes awry, a breeze blows, you trip over 

the mount, etc. This is a good time to ind those "bad" frames and pretend they 

never happened. There are two tools to help you here.  

Grade Image Quality 

This will look at a set of frames and attempt to automatically grade them as to how 

sharp they are relative to each other. The idea here being that you'll not use the least 

sharp frames. Pull down Processing, Grade Image Quality and point it to your light 

frames. It will rename them (or copy them with a new name) denoting how sharp 

each frame is.  

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: