A question about holograms

MfA

Legend
I've been reading a bit about Computer Generated Holograms and I came upon the paper A Framework for Holographic
Scene Representation and Image Synthesis
.

What amazes me is how detailed their reconstructions are considering they are only computing 1024^2 samples per hologram, whereas photographically created holograms use films which can resolve sub-micron features. Can someone explain to me why this can work inside the computer but not inside physical reality? (Physical reconstructions from such low resolution holograms look like crap.)

PS. hey Simon, I guess pre-filtering is possible after all :p
 
Last edited by a moderator:
What amazes me is how detailed their reconstructions are considering they are only computing 1024^2 samples per hologram, whereas photographically created holograms use films which can resolve sub-micron features. Can someone explain to me why this can work inside the computer but not inside physical reality? (Physical reconstructions from such low resolution holograms look like crap.)

For many of the same reasons an amateur 35mm photograph of a computer screen usually looks worse than a screen capture converted to a JPEG... namely imperfect optics, imperfect media, environmental factors (we don't live in an isothermic 0K perfect vacuum), human error, etc. The theoretical resolution of the media is simply not an issue given the budget and technical expertise you are talking about. We also don't use idealized pinhole cameras in the creation and viewing of holograms, nor do we usually view holograms from exactly the reverse direction of their illumination direction to 64 bits of floating point precision. Really without starting or ending with something physical their technique is little more than a sophisticated encoding of color and Z data of a single 3D frame.
 
For many of the same reasons an amateur 35mm photograph of a computer screen usually looks worse than a screen capture converted to a JPEG... namely imperfect optics, imperfect media, environmental factors (we don't live in an isothermic 0K perfect vacuum), human error, etc.
Not quite enough to explain multiple orders of magnitude (the best holograms would have around 4 orders of magnitude higher dpi than a 1024x1024 pixel image on a decent monitor).
The theoretical resolution of the media is simply not an issue given the budget and technical expertise you are talking about.
Either the information is in there or it isn't.
We also don't use idealized pinhole cameras in the creation and viewing of holograms
Neither did they, they used virtual apertures in the reconstruction (there are no cameras at all, virtual or physical, in hologram creation though).
nor do we usually view holograms from exactly the reverse direction of their illumination
Now this is important, I didn't immediately realise they never did any off axis reconstructions.
Really without starting or ending with something physical their technique is little more than a sophisticated encoding of color and Z data of a single 3D frame.
No, it's a limited resolution encoding of the object wave (a pure 2D representation I might add). There are some nice animations showing off axis viewing angles and varying apertures on his site.
 
Last edited by a moderator:
and Z data of a single 3D frame.

Let me amend that to "depth" rather than "Z", as it isn't simply depth from camera to object, but rather an approximation of depth from camera to object to simulated emitter being encoded through simulation of wave propagation in the technique described in the paper.


---
beyond3d lies technical marketing
 
Actually they don't simulate wave propagation, that's just to explain how holograms work. They use a FFT based technique to compute the interference patterns for individual textured polygons.

PS. well they simulate it on reconstruction, obviously ...
 
Last edited by a moderator:
Oh hey I'm an idiot ... this is why it works for them :
Note, that by simulating synthetic objects only three orders of magnitude bigger than the wavelength we are able to keep the hologram size reasonable and still get nice images, since the resolution of the final image and not the physical size of it is important when rendering to a screen (cf. Tab. I).

So you can get really nice detail in physical holograms too with only 1024x1024 samples ... they just have to be tiny.
 
Back
Top