night shooting tip?

[cosmoepic]

lookin around on the internet i came across something that said if you take a night shot and it comes out kinda pixely/noisy you can take the same shot with the lens cap on, paste that image over the night shot in photoshop, then change the layer to difference and it will help clean up the image


has anyone ever heard of this? i have yet to really try but thought it was kinda odd

 

[Johnboy2978]

That makes no sense. It would be the same as taking an image and then adding a layer that was just a black background. When you go to difference, all you're going to see is the outline of what the original image was. To reduce noise use noise ninja or neat image or equivalent.

 

[cosmoepic]

yeah i didnt think it made much sense to me either never used noise ninja ill look into it

 

[astrostu]

Actually that's how you do it. Sorry, Johnboy, but it does make sense ... at least for the non-random noise due to the detector not being complete even (as in every pixel the exact same sensitivity.

This is called "dark subtraction." Every astronomer does it (at least if they're processing the image properly). The dark (the image with the lens cap on) has to be taken for the same amount of time as the original image. This is also how in-camera noise reduction is done (and why it takes 2x as long as the original exposure).

 

[JerryPH]

This is just to reduce the noise caused by the camera sensor overheating, not high ISO noise.

High ISO noise should not be an issue in most (but of course not all) cases.

 

[cosmoepic]

thanks guys cant wait to try this i have a night shot in the beginner section and it doesn't look as sharp as i would like so ill try this

 

[astrostu]

This is just to reduce the noise caused by the camera sensor overheating, not high ISO noise.

High ISO noise should not be an issue in most (but of course not all) cases.

That's what I partially said, but you are partially wrong.

There are two types of noise, random and non-random. Photon-counting noise (amplified by amplifying the ISO) is random. It cannot be reduced in this way. The only way to legitimately reduce it other than using nearest-neighbor or similar averaging techniques is to use a longer exposure so you have more photons and less relative noise (this noise goes as the SQRT of the counts). Thermal noise - random noise created by electrons being recorded as photons because they have a finite temperature - is also random.

Non-random noise is introduced because not every pixel is created equally. Some are more sensitive than others. Some are always "on" (hot pixels), and some are always "off" (cold pixels or dark pixels). Then there's also non-uniform heating in consumer/prosumer/professional digital cameras where part of the detector will become warmer than another part, creating non-random thermal noise on top of the random thermal noise. If you take a picture with the lens cap on, you will see the non-random noise. On my camera, there are obvious rows and columns of hotter pixels, various hot pixels scattered throughout, as well as a corner of the CCD that's hotter than the rest. This is why I said you're partially wrong, because you neglected non-uniform pixel sensitivity.

For these non-random noise signals, dark-subtraction takes care of it (assuming the CCD is the same temperature as it was when the image was taken). You're just subtracting out the non-random noise.