I think you're discounting something... If you consider how Google Cardboard works, it exploits the fact that the camera has a field of view - the image is per

I think you're discounting something... If you consider how Google Cardboard works, it exploits the fact that the camera has a field of view - the image is perspective. So when you hold the camera at 10 degrees left from center, it takes the right-most column of pixels... And when you hold the camera at 10 degrees right of center, it takes the left-most column of pixels. And these columns of pixels converge - as though your head were as wide as the distance between those two points where the camera was. (Something like that.) Point being, if you move the camera through enough points, aimed in enough directions (even if it's on the surface of a sphere, always aimed out), you eventually capture rays of light as though they passed through every voxel in that space, from every direction. And, just to highlight, they are doing a TON of compression. This is essentially inside-out tomography, or something like that, to derive what the color is, from every voxel, in every direction... And then doing a 5-dimensional compression on that data (2D surface of a sphere of pixels, at every position inside a 3d sphere.) So, I think it's a much more "decent approximation" than you're giving it credit for. Or at least, it has the potential to be, if the algorithms are good enough.