From:	David Woods <david@*******.FREESERVE.CO.UK>
Subject:	Re: Ultrasound vision
Date:	Fri, 5 Feb 1999 19:33:06 +0000

Adam Getchell wrote:
>
> >You will have to propose a plausible method by which a detector array,
> >without a mirror or a lens, can image a scene using light. You have not
> >done
> >this yet. Good luck.
>
> Well, you'll have to read the paper then. I'm working on research
> that does diffuse realtime imaging of the body using far IR. This is
> essentially the same problem as reconstructing the temperature distribution
> (spatial and temporal) of a candle on one side of a metal bar, given the
> temperature distribution of the other end.

Yes, I've heard of this work.

> Since it is a diffuse source, the "lens" is linear algebra
> translation matrices (which is how lenses are treated in optics texts). I
> suggest "Introduction to Optics and Optical Imaging" by Craig Scott. Of
> particular relevance to the discussion is Chapter 14, "A Sampling of
> Optical Information Processing Systems Based on Fourier Plane Filtering".

Thank you, I was already aware of such filtering.

> The optical Fourier transform mentioned replaces the sequence of delta
> functions in the mathematical transform with a finite grid of spots.
>
> With good results and proper computer programming this stands the
> chance of replacing X-rays mammography as a realtime imaging tool without
> the drawbacks of ionizing radiation.
>
> So yes, realtime diffuse imagery (a step harder than what you were
> talking about) using light without a lens is possible and practical.
> The technique would obviously have implications elsewhere.

Maybe your neglecting the fact the in most imaging done outside a lab is
of
objects effectively at infinity. This is what I meant by 'real world
imaging'. A set of randomly positioned lambertian objects, at infinity,
will evenly illuminate an unmasked 2D photodetector array. The amplitude
profile will be perfectly flat. There would be no distribution to
analysis.

The direction of the incoming rays must be encoded, before detection;
the
image of the scene has to be reconstructed by decoding the observation
afterwards. Without this two-step process I can't see how your fourier
analysis can do what you claim.

As you are doubtless aware encoding the signal before detection is
called
multiplexing, and is divided into two classes. Temporal and Spatial.
(Caroli et al. 1987).

As the Temporal techniques involve simply scanning the sensor across the
scene they aren't very interesting. I'm aware of two types of device
using this technique that image distant objects. One uses two or more
collimator grids, widely separated, placed in front of a detector. The
second uses one or more arrays of opaque and transparent elements.
Instruments of the former type are called 'Fourier transform imagers'
(Makishima et al. 1978 and Palmer & Prince 1987). The latter are called
'Coded aperture cameras'.

This is indeed imaging without a mirror or lens. However, it also
addresses
my criticisms I made against your previous posts. Specifically the need
to
determine the direction of the incoming photons, a point you
consistently
failed to address.

It does involve an optically active element in front of the detector
array; a point I originally covered.

I can see how objects close to the array (in proportion to the array's
extent), could be resolved. A point source at a finite distance would
create
an amplitude profile across the array. I imagine the resolution of such
a device would be very dependent on target range.

Regards

- David Woods

Caroli, E., Stephen, J.B., Di Cocco, G., Natalucci, L., Spizzichino, A.:
1987, SSR, 45, 349

Palmer, D., Prince, T.A.: 1987, IEEE Trans. Nucl. Sci., NS-34, 71

Makishima, K., Miyamoto, S., Murakami, T., Nishimura, J., Oda, M.,
Ogawara,
Y., Tawara, Y.: 1978, in "New Instrumentation for Space
Astronomy", eds. K.A. van der Hucht and G. Vaiana (New York: Pergamon),
277