Error model of depth map


In my application, it is not enough to only know depth confidence map (0-100). Therefore, could you please provide some documents on how to calculate the depth confidence in SDK? I want to know the mathematical relationship between the estimated depth error and confidence.

Thank you for your time!

Hi @ZhenshanLiang
Welcome to the Stereolabs community.

I’m sorry, but this information is not available. The confidence map reports an evaluation of the goodness of the stereo correspondence for each pixel, and this is the only available information.

Hello @Myzhar,

Thank you for reply. I will try another way.

Hello @Myzhar,

Hope you are doing well. I am not sure the answers to the following questions are open to discuss, but I am OK if it is not.

You mentioned that confidence map is related to goodness of the stereo correspondence. So, if the correspondence is good enough, it means calculated disparity is very accurate, which means the disparity deviation can be considered as 0 pixel. And, I want to know how many pixels of disparity deviation the worst correspondence can cause?

By the way, is it possible that the confidence is very small (smaller means more trusted from SDK docs) but the depth is completely wrong?

Thank you for your time and patience.

We do not have this information.

Yes, this can happen mostly when there are repetitive patterns.