Dead pixel masking

#1

Hi there,
We’re using numerous Mako 319-B cameras and have noticed that all tested so far (~6 units) have quite a number of faulty/dead pixels (especially on dark images, or high gain settings). I understand that this is the nature of the sensors and common with regards to manufacturing processes, etc, but I have read that there are techniques for handling this, such as dead pixel masking.

Is there a way of accomplishing this via CVB?

I read somewhere that manufacturers can save such a mask on the camera device (obviously a mask specific to the defects on that particular camera), but I assume this is for mass consumer units such as the camera devices used within a typical mobile phone.

This is not critical to our application, but it would be great to understand if CVB was capable of doing this, ore if there was a way to configure the camera with such a mask so it’s handled up-stream.

Best regards,
Lee

#2

CVB does not implement such a feature.
In general it is better that dead pixels are handled in the camera firmware so that sensor specific characteristics can be taken into account.

I took a look into our database and found out that your camera model does not support the Pixel Defect Masking feature from Allied Vision.
Some other Mako models like the G-419 supports this feature.

1 Like
#3

One thing you could try is the CreateFlatfieldCorrectedImage function exported by the CVCImg.dll. It applies - as the name implies - a flat field correction derived from a dark and an average gray image. This is not the same as defect masking (which in many cases is implemented by interpolating the dead pixels’ content based on the surrounding pixels) and it will definitely not cure all your problematic pixels. But it’s the closest thing we have, might help with at least some of your problematic pixels and is therefore probably worth a try.

1 Like
#4

This is good to know for me too, as I did not know that such a function exists in CVB.

Thanks @illusive for the hint.

#5

I see- thanks for confirming Sebastian. It wasn’t too clear in the camera specs.

#6

Awesome- thanks @illusive; I’ll let you know how it goes if we look to implement.