0

Assume there is a black box there is no light in it and there is a camera in this box. Camera starts capturing and what it captures is nothing - everywhere is pure black in box. But there will be differences in the sizes of captured frames due to various types of noises (thermal noise, quantization noise etc). I want to decrease/eliminate the effects of these noises in software side so that in a completely isolated black box, all captured frames will be exactly the same. Resolution, depth, color etc none of the properties matters after processing, accuracy/quality of the captured frames in the end doesn't matter. Any kind of filtering, downsampling etc every solution is acceptable. Reference is the black box, frames should be as identical as possible.

Any suggestions ?

y33t
  • 649
  • 4
  • 14
  • 23

2 Answers2

0

There are many approaches to how to remove noise, the noise you are talking about is probably Gaussian noise. The simplest thing you can do to remove it is to run a Gaussian blur on the image, and then to use threshold to remove (make the value zero) everything that it's value is under "a", when "a" is a parameter you should play with a little bit to find the most suitable value.

OopsUser
  • 4,642
  • 7
  • 46
  • 71
0

ok this is a bit funny answer since you wrote accuracy/quality of the captured frames in the end doesn't matter, i'd say simple return black a software generated black image. You could do this without a camera ;)

hm more seriously i think you want to calculate the average nearing to black. each extra sample will have less effect, and it builds up over time to create a noise pattern for your camera.

Then substract that image from new captures.

Or return a statistical result so you return values and the probability of that value.

Peter
  • 2,043
  • 1
  • 21
  • 45