Snaping the Milky Way - noise reduction

Discussion in 'Open Discussion' started by rossi46, Feb 24, 2013.

  1. rossi46

    rossi46 Mu-43 Regular

    141
    Mar 1, 2012
    Hi all,

    This is one lovely type of photography I had yet to try.

    Had read quite alot on the net for tips of capturing the stars...

    One thing that I am still confused is the Noise Reduction function.
    It is recommended to turn the Noise Reduction on when capturing the milky way due to long exposures of about 15 seconds bringing some noise.

    But if we are shooting Raw, do we still need to turn on the Noise Reduction function? Correct me if I am wrong, it would not make any difference to the Raw file (by turning on Noise Reduction)?
     
  2. David A

    David A Mu-43 All-Pro

    Sep 30, 2011
    Brisbane, Australia
    Noise reduction applies a process called dark field subtraction to eliminate noise due to the long exposure. It does make a difference to the RAW file because what happens is that after the actual exposure, a second exposure of equal length is made with the shutter closed. The second exposure will reveal only the noise associated with the exposure, and that noise signal is then subtracted from the original RAW data to produce the data for an image without the noise. It's not "noise reduction" of the sort you apply in image processing. The camera setting for that is the Noise Filter setting.

    So, best to leave noise reduction on even if you shoot in RAW.
     
  3. dafloyd

    dafloyd Mu-43 Rookie

    23
    Feb 24, 2013
    Munich, Germany
    Dirk Essl
    This all depends, which camera you use.
    On my EM5 I have turned noise reduction off. My EP2 has it turned on.
    I would suggest you experiment which setting suits you.
    There is also software that can do darkframe substraction. This means you shoot 1 picture with the Lens cap on with the same settings as you will use for your picture, let's say you will use 30 seconds for your shots as exposure time. You can then take many shots with those settings (30 seconds, same ISO) and use that darkframe to reduce noise. This way you save a lot of time while shooting as you don't have to wait for the camera to do the darkframe substraction for every shot. Software that can remove darkframes are Registax and Deep Sky Stacker.

    Cheers,
    Dirk
     
    • Like Like x 1
  4. F1L1P

    F1L1P Mu-43 Veteran

    388
    Jan 2, 2010
    Europe
    If you want better results, turn it OFF.

    Look at it like this: do you want your camera to do automatic noise reduction or do you want to reduce it manually, put some effort and achieve better results?

    It's much like JPEG vs RAW debate.
     
  5. rossi46

    rossi46 Mu-43 Regular

    141
    Mar 1, 2012
    Can you please explain more about the technique of using long exposure (eg. 30 seconds) and taking multiple shots...

    - you mean you take multiple shots during that 30 seconds exposure period?
    - why do you need to take a picture with the lens cap on?
     
  6. rossi46

    rossi46 Mu-43 Regular

    141
    Mar 1, 2012
    Noted.

    How do you normally perform noise reduction for such long exposure images of milky way? Using the Adobe Camera Raw does a good enough job? Better than even the in-camera noise reduction?
     
  7. arad85

    arad85 Mu-43 Veteran

    477
    Aug 16, 2012
    He means taking multiple exposures of 30 seconds each. The idea is to increase the total exposure time so that you can get more signal. The problem you have is that the earth is rotating, so unless you have a tracking mount, you are limited in the exposure time you can have without the stars elongating and turning into star trails. You then use stacking software to automatically combine the images into a single one.

    The reason you take a photo (or multiple photos) with the lens cap on is to give you some frames which show just the sensor noise. The software can then apply processing and remove some of the noise in the image by subtraction as the dark frames (as they are known) contain just the noise, whilsdt the light frames contain signal + noise.
     
    • Like Like x 1
  8. arad85

    arad85 Mu-43 Veteran

    477
    Aug 16, 2012
  9. RobWatson

    RobWatson Mu-43 Hall of Famer

    Best to use the noise reduction feature (poorly named - actually does not do noise reduction but dark field subtraction) as it is more accurate. The levels of error signals are temperature dependent so taking the correction frames in close temporal proximity to the actual light exposure gives a more accurate correction.

    In astrophotography typically the camera is temperature controlled so it is easy at some other time to obtain corrective frames at the same temperature as the light images. In the very extreme ends of the 'spectrum' cameras (sensors) are cooled to liquid nitrogen temperatures effectively turning off the error signals due to thermal effects.

    Doing the correction in post is tricky but not impossible (certainly not a accurately). Once some curves or white balance is applied then the non-linearities induced make accurate correction exceptionally difficult.


    Basic brute force correction is a 3x3 median pixel blur ... you can imagine the impact this can have as opposed to simply using the NR feature.
     
  10. rossi46

    rossi46 Mu-43 Regular

    141
    Mar 1, 2012
    Are you referring to the in-camera noise reduction software's working method?
     
  11. rossi46

    rossi46 Mu-43 Regular

    141
    Mar 1, 2012
    Wow,...Thanks for your efforts to explain why in-camera noise reduction is more accurate...

    Though, this still seems alittle too complicated for my proper understanding, but most important thing for me is that, now I understand its easier and better to use in-camera noise reduction.

    Another quick question for you guys, is it still necessary to perform post processing noise reduction after applying the in-camera noise reduction?
     
  12. arad85

    arad85 Mu-43 Veteran

    477
    Aug 16, 2012
    In camera noise reduction can refer to one of two things:

    • Noise reduction: applying filtering to a JPEG image to reduce the noise. This typically loses fine detail and gives blurrier images the higher the ISO
    • Dark frame subtraction: after a light frame is taken, a dark (one with the shutter shut/lens cap on) is taken. This gives the sensor noise without any exposure. The camera can then subtract the two images to give less noise in the exposed image.

    "proper" astrophotography (where you take multiple long exposure images of the same thing using a tracking mount) will typically have many 10's of light frames, and perhaps a dozen dark frames. Specialist software then processes these set sof images to provide a single image with much less noise than you would have than if you took a single image alone.
     
  13. F1L1P

    F1L1P Mu-43 Veteran

    388
    Jan 2, 2010
    Europe
    easier - yes, better - no.

    A dedicated software running on your quad core PC averaging several light frames and subtracting all the noise, hot pixels and sensor noise from multiple images can only be compared to in-camera NR as better option.

    Again, it is much as JPEG vs RAW debate. RAW needs to be "developed" and that process contains more steps, but yields higher quality.

    Images taken by dannat,
    single capture:
    [​IMG]

    VERSUS

    multiple captures: stacked with DSS
    [​IMG]

    do you see the difference?
    Ignore some errors due to wrong alignment in PP, notice wider dynamic range, details...

    More info, and original post from dannat can be found here:
    https://www.mu-43.com/f88/astrophotography-m4-3-what-have-you-got-29340/
     
  14. RobWatson

    RobWatson Mu-43 Hall of Famer

    The simplest thing for you to do is try with and without the in camera processing and see which results you like better.