The depth error for a stereo camera is given by: formula

As the depth increases the error increases quadratically. I'm wondering if there's a way to model that error and then correct the estimate.

On my camera, I notice that the z value of these points are always over estimated as I move the object further. Is this always the case? Or is the error some sort of gaussian such that it cannot be modeled.

I was wondering if I could just subtract some number from the depth value, based on where the object is to account for the quadratic nature of error and correct it?

Thank you

  • $\begingroup$ No. The stereo error mostly comes from quantization errors and small mismatches. The noise is generally gaussian and can't be subtracted out. $\endgroup$ – edwinem Jun 16 '20 at 3:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.