I'm doing this reprojection in CUDA with cuda textures (aka cuda Arrays). I'm pretty sure that it's not the half pixel offset issue, because if I use the original screen positions (rather than the screen->world->screen computed ones) the picture is stable.

I'm doing this reprojection in CUDA with cuda textures (aka cuda Arrays). I'm pretty sure that it's not the half pixel offset issue, because if I use the original screen positions (rather than the screen->world->screen computed ones) the picture is stable.

EDIT: the following is wrong.
Also note that with bilinear sampling, the image colors quickly propagate. For example, in a 1D pixel grid at time=1 pixel 1 has pixel 0's information, at time=2 pixel 2 has pixel 0's, etc if the pixel coordinates aren't exact.

I don't remember seeing any negative precision deviations, so the end result is some blurring and floating to the bottom right.

Note that when I say floating, I don't mean that stuff in the top-left disappear. It's still there.

I'm doing this reprojection in CUDA with cuda textures (aka cuda Arrays). I'm pretty sure that it's not the half pixel offset issue, because if I use the original screen positions (rather than the screen->world->screen computed ones) the picture is stable.

Also note that with bilinear sampling, the image colors quickly propagate. For example, in a 1D pixel grid at time=1 pixel 1 has pixel 0's information, at time=2 pixel 2 has pixel 0's, etc if the pixel coordinates aren't exact.

I don't remember seeing any negative precision deviations, so the end result is some blurring and floating to the bottom right.

Note that when I say floating, I don't mean that stuff in the top-left disappear. It's still there.

I'm doing this reprojection in CUDA with cuda textures (aka cuda Arrays). I'm pretty sure that it's not the half pixel offset issue, because if I use the original screen positions (rather than the screen->world->screen computed ones) the picture is stable.

Also note that with bilinear sampling, the image colors quickly propagate. For example, in a 1D pixel grid at time=1 pixel 1 has pixel 0's information, at time=2 pixel 2 has pixel 0's, etc if the pixel coordinates aren't exact.

I don't remember seeing any negative precision deviations, so the end result is some blurring and floating to the bottom right.

I'm doing this reprojection in CUDA with cuda textures (aka cuda Arrays). I'm pretty sure that it's not the half pixel offset issue, because if I use the original screen positions (rather than the screen->world->screen computed ones) the picture is stable.

Also note that with bilinear sampling, the image colors quickly propagate. For example, in a 1D pixel grid at time=1 pixel 1 has pixel 0's information, at time=2 pixel 2 has pixel 0's, etc if the pixel coordinates aren't exact.

I'm doing this reprojection in CUDA with cuda textures (aka cuda Arrays). I'm pretty sure that it's not the half pixel offset issue, because if I use the original screen positions (rather than the screen->world->screen computed ones) the picture is stable.

Also note that with bilinear sampling, the image colors quickly propagate. (e.g. in a 1D pixel grid at time=1 pixel 1 has pixel 0's information, at time=2 pixel 2 has pixel 0's, etc if the pixel coordinates aren't exact)