Search code examples
imageimage-processingcamerapixelsensors

What is the reason of periodicity of image pixels like dark-ligth when acquaring with industrial cameras-lens?

I use industrial camera (Make: IDS, Model: UI-5490SE-M-GL) and telecentric lens for getting pure image in order to measure diameter of specific product. When i acquared an image using camera and telecentric lens, adjacent pixels is different from each other with periodicity. The image that i capruted is: Pure image

When i zoom this image, i can see dark-ligth-dark-ligth pixels in boundry regions with periodicty like that:

image-dark-light-dark-ligth

So what is the reason of this dark-light-dark-ligth effect?

Solution

  • The image sensors used in monochome cameras are the very same as used in bayer pattern color cameras, except that they don't have RGB filters in front of each sensor element.

    Due to the different sensitivity of the detector material for different wavelengths, the size of sensor elements is varied to compensate for the different relative sensitivity in case a RGB filter is put on top of the sensor. Without the RGB filter these different sensitivities become apparent.

    That being said: Since the sensor elements of an image sensor do have a very strong bias/gain variation regardless of the later sensor configuration, it is mandatory to perform a per sub-/pixel bias/gain characterization for a given sensor (=camera) anyway. This is to be done by either the camera manufacturer and the compensation being computed in the camera itself, or by making that calibration data available to the end user application.

    If you don't have that data available, you must characterize the camera yourself using a homogenized illumination source (usually by putting the naked camera sensor at the side port of an Ulbricht integrating sphere), taking calibration exposures at varying illumination levels, exposure times and sensor gain settings.