3

I'm trying to take some timelapse photos with a webcam. To reduce noise, I'm trying to take a dark frame. However, with the lens covered up totally, there is absolutely no noise in the image. The whole frame is uniformly black. Performing a graphic equalise yields nothing whatsoever. There is no image information. Noise only appears when there is a little light available, but then I'm just taking a picture of the light source. It seems that the camera sensor turns off at very low light levels.

Has anyone else seen this behaviour? My camera is a Foscam ip camera, with a CMOS sensor. My understanding is that noise from a variety of sources is present at all times...

To give a concrete example, see image below...

Partial noise frame

This is an image from my web cam, with a bit of light passing through a slit. You can plainly see the bright bit, containing noise (especially towards the bottom). However, either side there is no noise what so ever. I have selected the black area with a tolerance of 0. You can see from the histogram that there is absolutely no noise. Mean = 0, STDDEV = 0. You will also note that the brightest bit has not blown out, so the whole image is well within the cam's optical range.

I don't think that the sensor is cutting off the noise internally as suggested, unless it does it on a per pixel basis. This is impossible though I believe because how does one pixel sensor know if it's noise or just low light? There seems to be a threshold value below which the sensor pixels cannot register anything. This is not the same behaviour that I see from my D80.

Can anyone posit a theory as to what's going on, and how I can judge the level of light just enough to take low light pictures..?

Improve this question

3 Answers 3

Reset to default
2

Your webcam will be applying a threshold to the signal coming off the sensor and zeroing any values below the threshold. Since the darkest parts of the image are mostly noise anyway this scheme reduces bandwidth.

Seeing as your webcam is basically erasing these values there is no way you can record a dark frame. If you really really wanted to you could record a uniform dark grey frame, but you'd have to work very hard to make sure what you're photographing was uniform grey with no brightness gradient. It's almost certainly not worth it.

If noise in you timelapse is a problem, I would look into noise reduction software designed for video, which will be able to smooth noise over many frames.

Improve this answer
3

What is probably going on has little to do with the camera sensor, per se. When the signal from the sensor is processed there is noise reduction applied. It seems that in the case of your webcam, the processor recognizes a "noise only" black frame and goes ahead and eliminates all of the signal to get rid of the noise.

Improve this answer
2
  • It's probably nothing more complicated than clipping the blacks to a level greater than the noise. I highly doubt the manufacturers of the webcam developed a processor that would recognize a frame that contained only noise and process it differently.
    – Matt Grum
    Oct 31, 2014 at 9:54
  • Setting the blacks to clip at a threshold that includes read noise is pretty much what I described - everything below the S/N ratio of a dark frame is rendered as 0, 0, 0.
    – Michael C
    Nov 1, 2014 at 15:19
0

When using CCD and CMOS sensors, the distribution of the DC component of the sensor noise rarely changes with time (unless you plan on using your camera in extreme temperatures). This allows for effective factory measurement and calibration which allows them to measure and remove the dark frame automatically. So I don't think there is much use in trying to do this yourself.

This is not the case for other sensors such as cameras beyond the visible spectrum that need constant cooling and such.

Improve this answer
3
  • You seem to be implying that it's possible to largely eliminate noise from CMOS sensors but my DSLR disagrees with you.
    – David Richerby
    Oct 24, 2014 at 9:17
  • no, it is not what I implied. You cannot eliminate the entire noise. Thats impossible. What is done with dark images is measure the average value of every pixel and transform the image to be more uniform. This is used to effectively reduce the non uniformity in the detector offset which is naturally occuring as each pixel has a slightly different response. So using dark images, you do not eliminate the noise, you zero mean it.
    – user2324712
    Oct 24, 2014 at 18:39
  • 1
    Since a webcam typically stays on for long periods in roughly the same environmental conditions (i.e. your environmentally controlled living room), read noise will generally remain fairly constant. This is not so much the case with a DSLR, which is designed with a wide variety of environmental conditions and for which the amount of time the sensor is energized can vary widely.
    – Michael C
    Oct 24, 2014 at 18:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.