I've shifted the data by 4 and it looks better now (a lot more recognizable). IR is still black. The attached image shows my hand in the depth view.

Deleted: Screenshot from 2016-11-10 15-50-52.png 183 KB

	Screenshot from 2016-11-10 15-50-52.png 183 KB View Download

I forgot to mention - the image I get when using librealsense and not Chromium to get the image looks fine - similar to the second image.

The bit shift is just a hack and not the right way the data should be displayed. Info about how the format actually looks like should be somewhere here https://github.com/IntelRealSense/librealsense/blob/master/doc/projection.md#depth-image-formats.

mediacapture-depth specification doesn't handle specific rendering of disparity mode for R200 camera [1].
Don't think there's much we can do about this in Chromium - this is how upper byte of 16-bit data looks.
We could check though:
1) Find if it is possible and how to switch disparity (only available in R200) mode ON/OFF. 
2) Propose rendering for disparity mode in mediacapture-depth spec.


[1]
https://github.com/w3c/mediacapture-depth/issues/80#issuecomment-260641134

This is not a problem of disparity mode, it's simply that the R200 uses a different scale. The other depth cameras use a 1/32 mm scale while R200 uses a 1 mm scale, making the normal output look like random noise. We could either just scale it down by 32 with glPixelTransferf(*_SCALE) or something similar, or use some equivalent of librealsense's rs_set_device_option(...) with RS_OPTION_R200_DEPTH_UNITS, set to 31.

Technically, the data currently shown aren't exactly wrong - they will just confuse the users a lot, We could  put this into a JS library on top of Chromium, but we might get bug reports when uses see expected output in librealsense and black with a little noise with Chromium.

The work on this needs to be moved to:
https://www.w3.org/TR/mediacapture-depth/#the-video-element
there it is/should be defined how to render long range, near range, disparity this in video element.

Chromium exposes row 16-bit format, deprojection to real world coordinates (scale) from 16-bit value happens using mediacapture-depth API in JS.