In one of our previous blogs we already discussed how to use ImageJ to visualize image statistics and make dark noise visible. However before analyzing the dark noise with a program like ImageJ, it is important to measure the dark noise correctly. Here are some tips on correctly measuring the dark noise.
A good reference measurement of the camera noise profile could potentially be used to do image corrections on the final images of interest. In most cases this is not necessary as the noise correction functionality inside the camera will already be sufficient, but there might be situations where this correction is not sufficient. For example, in high resolution cameras often only column based corrections are performed. The column based corrections might not be sufficient to provide a flat enough image. Subtracting a black reference image can improve your final image quality in such a situation.
So what should you take care of when measuring the dark noise profile? Well actually you should make sure that you measure the full noise profile. This can be realized by measuring the dark noise with an artificial black level offset.
The black level offset has to be such that when an image is captured there is sufficient signal swing upwards and downwards to capture the full signal including downward noise spikes. The noise needs to be digitized.
To clarify this let’s discuss what happens when there is not sufficient signal swing. For example, when the average signal level equals zero, there is only space for noise spikes above the average signal. This is illustrated in Figure a). Noise spikes below the average signal cannot be recorded.
It is thus preferential to measure the noise profile with an artificial black level offset such as is illustrated in Figure b). With an artificial offset there is space to also digitize downward noise spikes and thus to record the full dark noise profile of the camera.
Be sure to use an artificial offset, such an offset is often available as a camera feature. Do not create an offset by illuminating your sensor with light. Illuminating your sensor with light would add an additional noise source which is not related to the camera characteristics.
Note: The line plots show vertically averaged horizontal noise profiles.