Revision history [back]

In my experience (Kinect V1 & V2, orbbec astra, RealSense)

Kinect V1 had a really weird depth format, where you have to shift bits, etc.

Kinect V2, RealSense, Astra, and probably the rest, use the standard of 1 Unit = 1mm as PeteBlackerThe3rd says...

But there's also special values:

If you get a pixel with value 0, it doesn't mean that at that pixel the measure was zero, it actually means "unknown" value. This usually happens when the sensor is not able to acquire a value for that particular pixel.

To some degree, zero value pixels can be considered as "transparent".

This is important, because it needs to be taken into account for processing the depth image. For example, RealSense has a "Spatial Hole Filling" post processing technique that fills the empty gaps using the surrounding bits, but I think it's not enabled by default.

I've also recall a depth camera that also had a non zero special value to indicate some sort of overload in the pixel measurement.