2

Background

I have a Panasonic HDC-HS700 camera that I am using to record indoor office scenes. The scenes are illuminated predominantly by fluoro lighting, with little natural light, however there are substantial lighting variations between corridors and offices/cubicles.

So far, my recordings contain a fair amount of image noise. The video will be processed using image processing and computer vision techniques, so I need to keep image noise to an absolute minimum for the algorithms to work best. However, upstream noise removal (i.e. within the recording process itself) is much more desirable than downstream noise reduction (i.e. image processing algorithms).

Question

I am a novice when it comes to video production, so I am unfamiliar with the "tricks of the trade". What is the best approach (and/or camera settings) to minimise image noise given my recording scenario?

Improve this question

2 Answers 2

Reset to default
1

The first thing is to make sure any gain is off. Gain on a video camera is like high ISO on a digital still camera. It increases the noise floor significantly and causes much more noise in even a good image. This should make the camera make other adjustments to try to preserve the exposure with a lower noise floor.

If there is still too much noise, it may require that the dynamic range of the scene be reduced with properly placed lighting to eliminate dark shadows. If you have particularly bright and particularly dark areas, the camera may attempt to get detail in the dark areas and result in noise. Reducing this dynamic range will allow the detail to be captured clearly in all parts of the image.

It is also possible that there simply isn't enough light. It can be very dark inside buildings. Our eyes are exceptionally good at adjusting to a wide range of lighting intensities so we don't really notice it much, but outside is hundreds of time brighter than inside. An average office hall lighting is about 80 lux. A typical sunny day is over 10,000 lux with over 30,000 lux in directly sunlight. That's a HUGE variation. The camera you mentioned seems to have good low light performance for it's price range from the CNet review, but if you still have too much noise after adjusting the gain to 0 or have too dark of an image, then your primary option is to throw more light at it. Good strong studio lighting should allow the sensor to clearly pick things up in such a way that the images will be suitable for computer vision with low noise.

Improve this answer
1

The answer is always "More light". The higher the illumination of the sensor (within obvious limits), the less noise or grain. If you must deal only with ambient light -- you can't supplement or fill in -- then open the iris. However, in doing so you trade off depth of field, so it's a balancing act.

Generally, avoid increasing preamp 'gain' to compensate for low light. It will increase the noise as much as it increases signal.

Improve this answer
3
  • It isn't always just as simple as "more light". It's a common mistake to turn on the gain to get a brighter image when it's dark and then the setting persists to when the camera is in brighter conditions resulting in the camera dropping the shutter speed to compensate. Often, simply adjusting the gain down is sufficient to correct without losing depth of field. Similarly, if the gain is on, more light isn't going to help with the noise since it will still be gained up and will just make the shutter even faster.
    – AJ Henderson ♦
    Jun 2, 2013 at 4:13
  • Aren't we saying the same thing? For any given amount of illumination, opening the iris to attain higher output will generally increase S/N ratio (and decrease depth of field). Increasing the gain to achieve the same output, while leaving the iris closed, will decrease S/N ratio.
    – Jim Mack
    Jun 2, 2013 at 10:29
  • The full extent of our post says the same thing. I'm just disagreeing with the first sentence. The answer isn't always "More Light". I was clarifying that you can have plenty of light and still have too much noise if settings are wrong.
    – AJ Henderson ♦
    Jun 2, 2013 at 17:14

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.