What is the basis for creating a depth map for an existing image?

Software
1

In other words, how does such a utility decide the distance from the camera of say groups of pixels and therefrom decide the level of each pixel in the map?

I do know what a depth map is and what the pixel levels therein are supposed to represent.

Thanks for looking.

P.S. I’m not talking about cell-phone images with data from a built-in depth-sensor.

2

I’d say to get a reliable depth map you need a special sensor of some kind (actual depth sensor, stereo camera, …). Without that, I don’t see how you would infer the depth information, as that information is simply not included in the pure RGB image.
Maybe nowadays some AI tools could give acceptable results, but I don’t know any.

3

IIIRC there have indeed been AI attempts to infer the depth map from bokeh basically (the blur of objects depends on distance/depth). There is more related stuff on “depth from defocus”, but seems to require more than one image captured at different focus points, or having “split” right/left phase detect pixels…

4

Thanks but I am talking about the depth map options in e.g. GIMP/G’MIC or in the RawTherapee DeHaze function which has a ‘Depth Map’ check box and a ‘Depth’ slider.

Both work on ordinary single images although G’MIC does have some 3D stuff too.

5

some cameras capture the depth map while capturing the image and store them in the file as metadata.

6