Hello,
I try to understand the output value of glReadPixels with parameters GL_DEPTH_COMPONENT.
Here a part content of my obj file:
v 47.004219 -60.619961 -34.676708
v 46.93742 -60.368637 -34.942249
v 46.895119 -60.117729 -35.213577
v 46.877819 -59.86705 -35.4883
v 46.885769 -59.61644 -35.763931
v 46.91898 -59.365719 -36.038002
v 49.249428 -63.449627 -32.687481
v 46.97723 -59.114708 -36.308041
v 49.000347 -63.183308 -32.73838
v 47.060051 -58.863247 -36.571617
v 48.755192 -62.918678 -32.813602
v 47.166721 -58.611179 -36.826332
v 48.516319 -62.655716 -32.91267
v 47.296288 -58.358318 -37.06987
v 48.28606 -62.394371 -33.034897
v 47.44759 -58.104519 -37.300022
v 48.066692 -62.13459 -33.17939
v 47.619251 -57.849648 -37.514709
v 47.860363 -61.876278 -33.344997
v 47.809673 -57.593552 -37.711983
v 47.66909 -61.619362 -33.530392
v 48.017082 -57.336121 -37.890038
v 47.49482 -61.363708 -33.734009
v 48.23954 -57.077221 -38.047291
v 47.339272 -61.109211 -33.954151
v 48.474949 -56.81678 -38.18232
My depth buffer output has min/max values ( 0.00011…, 0.00017…). What is the output value unit ? Is it a millimeter value ? Is it a depth value of only visible element ?
If someone can explain me.
Window-space depth values.
Typically 0…1 (with the default glDepthRange( 0, 1 )).
WIth a perspective projection, most values are typically up close to 1.
Though you may be doing something special, like rendering with inverted Z. In which case typically close to 0.
In fact, my questio behind this is how to convert this depth value to real depth value in meter for example?
Is there a way to do this?
Assuming glDepthRange(0,1) (the initial state), convert it to NDC with:
ZNDC = 2*depth-1
Then transform it by the inverse of the projection matrix to get an eye-space coordinate. If necessary, transform that by the inverse of the model-view matrix’ scale factor to get an object-space coordinate.
Your reported depth buffer values only make sense if you’re using either a) an inverted-depth perspective projection (0 at the far plane, 1 at the near plane) or b) an orthographic projection with an excessively large value for the far plane distance.
For a more typical projection, I’d expect to see either a much greater spread of values (for an orthographic projection with sensible near/far distances) or values clustered close to 1.0 (for a perspective projection).
Thank you for the answer. I will try with all that information