ECEN 4616/5616 2/4/2013 Digitally-Detected Holograms: By interfering two coherent waves at a detector, we can record the areas where they are in phase and out of phase. This constitutes a hologram. An example of an optical system that does this is: Here is a picture of the bench-test optics: This system was designed (via the detector, magnification, and maximum Numerical Aperture) to detect objects as small as 2 microns. Here is what the detected hologram of a 5 micron wire looked like: pg. 1 ECEN 4616/5616 2/4/2013 The top images are and in-focus image of the wire, and a hologram taken 1 cm out of focus. (The DOF of the system was about 100 microns for a 5 micron object.) The bottom images are reconstructions of the wire’s image by propagating a modified version of the detected hologram. Note how the background noise has been suppressed. A typical hologram of cloud drops (simulated by polystyrene beads in water for the bench) is: The difference between just propagating the image (as an amplitude-modified plane wave) and propagating a modified image is shown in the partial resconstruction of two beads in the upper right of the above image: pg. 2 ECEN 4616/5616 2/4/2013 Finally, here is the image, hologram, and reconstructed hologram of a barb from a guinea feather: pg. 3 ECEN 4616/5616 2/4/2013 The algorithm used to decode the above holograms is in Appendix A. pg. 4 ECEN 4616/5616 2/4/2013 Appendix A Improved Algorithm for Reconstructing Digital Holograms (From 1995, unpublished) Holograms A standard representation for a (monochromatic) optical wave field generally traveling near the z direction is, (in complex notation): O Ao ( x , y )eik o ( x , y ) where Ao ( x , y ) is the (non-negative) amplitude, and o ( x , y ) is the (arbitrary) phase of the field. (In the standard notation, z dependence is left implicit, since it can be deduced i t from the x and y dependence, and the time dependence, e , is left off, since it would simply cancel from both sides of the equations.) With this notation in mind, let's define O to be the light scattered from an object we wish to make a hologram of, and R is a (nearly) plane-wave reference beam. The hologram is made by allowing O and R to interfere at the site of a detector (a CCD camera, say). At the detector, then, the total wave field is ( R O) and the detected intensity is proportional to: H 2 R O 2 R O R O (1) R O RO R O, 2 2 where indicates the operation of complex conjugation. This expression describes the detected hologram. Optical Reconstruction of Holographic Images If the reference beam, R , is a good approximation of a plane wave, then the amplitude of R: AR ( x , y ) constant and the phase of R: R ( x , y ) constant , and: R AR ( x , y )e i R ( x , y ) RC where RC is a constant which can be made real by choosing the origin such that R 0 . Then, the 4th term in (1) (ie.: R O ) is proportional to the original object wave, O . Given this approximation, we can recover the image of the object by passing a plane wave through a mask (film) with transparency equal to H. This wave will then be pg. 5 ECEN 4616/5616 2/4/2013 modulated so that it is proportional to H , and the object-wave term R O RC O will produce an image at the original z-distance away from the hologram. Equivalently, we can simulate the propagation of a wave field equal to H digitally. Of course, R is never exactly a uniform plane wave, so distortion and noise are thereby introduced. The other terms in (1) that are not proportional to the object wave also introduce noise into the image. In addition, if the wave used in reconstructing the image optically is not a uniform plane wave, then even more distortion and noise are present in the image. The meaning of all of the terms in (1) are: 2 R : This is the background reference illumination; it is what the detector would see if no object wave were present. 2 O : This is the diffraction pattern of the object; it is what the detector would see if no reference wave were present. Generally, O R . 2 2 RO : This is the conjugate object wave, distorted by any non-uniformity's in the reference wave. The conjugate object wave is the same as the object wave, except it focuses in the opposite direction; thus the conjugate wave will be out-of-focus at the point where the object wave forms an image and vice-versa. RO: This is the object wave, distorted by non-uniformity's in the reference wave. A distorted reference wave will distort the final image. Special Digital Reconstruction of Holographic Images One of the advantages of having access to a digitized hologram is that we are no longer bound to the limitations of optical reconstruction. Since the detected reference wave, 2 R , can be recorded when no objects are in the beam, the first step in improving the hologram, H, is to subtract out the background: 2 H H R O RO R O. 2 Our first improved hologram, H , has one less noise term than the standard `optical' hologram. For further improvement, we would like to remove the distorting effect of multiplying the object wave, O , by R . Unfortunately, we do not know the amplitude function R directly, as we only have access to the detected intensity of the reference 2 wave, R . pg. 6 ECEN 4616/5616 2/4/2013 At this point, we make some reasonable approximations. Remember, by definition, R AR ( x, y )e i R ( x , y ) . In general, AR ( x , y ) constant , since significant variations in the intensity of the background is commonly seen in holograms. Variations in the phase, R ( x , y ) of R , however, would be directly related to components of R propagating in different directions. Since R is a well-collimated laser beam that propagates in (nearly) one direction only, evidently R ( x , y ) constant i. This, plus the fact that AR ( x , y ) 0 (by definition), means that: R R 2 R . The next improvement in our digital hologram is therefore: H H R 2 O RO RO 2 R 2 2 O O O. R This modified hologram, H , is much improved over the original detected hologram, H . The background variations are removed, the object wave is distortion-free, and the 2 (already small) diffraction pattern overlay, O , is reduced by division with the (generally larger) background. Virtually the only noise term of note left is the out-of focus conjugate object wave, O . Images produced by digitally simulating the propagation of H have 10 20 times the Signal/Noise of optically reconstructed hologram images (and are even better than many in-focus images, made with coherent light). They are a much better target for automatic image analysis techniques than images from optically reconstructed holograms. There is a further, iterative, technique whereby even the last two noise terms of H might be removed, leaving only the object wave. In H itself, we have no way to distinguish between O, O , and. 2 O R . However, if the object that O comes from is compact (that is, it forms a small image), then when we propagate H over distance to form an image of the object, O and O (as well as O R z 2 ) become separated geometrically. Then OZ z is the image, which is compact, real, and localized, while the other two terms are spread over a much larger region. We physically separate OZ z from the other terms by selecting a small area around the image. We can then apply any apriori knowledge we have about the image to improve it (for example, should it be a pg. 7 ECEN 4616/5616 2/4/2013 complete shadow image?), back-propagate it to the hologram plane (ie.; calculate OZ 0 ) from which we can calculate the complex representation of O and the diffraction pattern ( O O O ). These two non-object wave terms would then be subtracted from 2 H to yield a new estimate for O: H , which could then be propagated forward to find the next approximation to the image. We have not tried this yet, since the current algorithm works very well, but it might prove useful for applications where highprecision images are required. R x, y is an actual constant only for the case of in-line or Gabor holograms – for off-axis reference holograms, perhaps a linear phase function can be used as an approximation, depending on the angle of the reference beam. This has not been experimentally attempted, due to the difficulty of doing off-axis reference holography with limited resolution CCD cameras. i pg. 8