Copyright 2003 SPIE and IS&T
This paper was published in Proc. of Electronic Imaging 2003 and is made available as an electronic preprint with permission of SPIE and IS&T. One print or electronic copy may be made for personal use only. Systematic or multiple reproduction, distribution to multiple locations via electronic or other means, duplication of any material in this paper for a fee or for commercial purposes, or modification of the content of the paper are prohibited.

ABSTRACT

Accidental abrasion of a specular surface sometimes produces real and virtual images of the abrading object. Investigation of this phenomenon in 1994 led to a simple technique which almost sounds like a joke: a method for creating white light holograms by scribing interference fringes one at a time onto a plastic plate by hand. The optics of these abrasion holograms is similar to Rainbow holography, and the similarities reveal techniques for controlling distortions, for producing images of opaque objects, as well as for producing animation effects and images that extend out through the film plane. As with any rainbow hologram, spatially coherent or point-source illumination is required, both a transmission mode and a reflection mode exist, and conjugate illumination produces pseudoscopic images. This abrasion holography highlights the fact that the zoneplates comprising a Rainbow hologram function independently not only of illumination frequency but also of fringe spacing. Size-independent fringes suggests that truly enormous holograms can be engraved on a wide variety of very crude everyday surfaces.

1. INTRODUCTION

In inventing holography Dennis Gabor created a method where an object generates a complex zoneplate lens. This lens can distort a point light source image and reconstruct a three-dimensional image of the original object. Benton improved upon the technique with his Rainbow method. Under this method the monochromatic illumination requirement was removed and the zoneplate created by each point on the object no longer covered the entire hologram. Several other non-holographic techniques for producing freely observed 3D images also exist. These employ arrays of tiny cylinder lenses or slotted aperture plates, but these "Lenticular" devices do not involve optical wave interference and they lack most of the unique behaviors of holograms.

There is another method for 3D image reconstruction that lies halfway between the diffraction optics of holography and the geometrical optics of lenticulars. This third method sometimes appears in nature, and so it has been repeatedly discovered by observant researchers. I call the plates produced by this third method "scratch holograms", while Liebermann [1] calls them "holographic engravings," and Plummer and Gardner [2] name them "mechanical holograms." Abramson [3] calls them scratchograms or "Poor man's holograms" and also refers to the technique as Incoherent Holography. The physics itself suggests the terms "single-fringe holography" or "giant-fringe holography." Since the required materials were available to humans ever since we started using tools, with tongue in cheek I also call them "prehistoric" holograms. All these terms refer to the same phenomenon and the same techniques.

1.1 A Chance Observation

In a sunny parking lot its easy to observe strange patterns of highlights apparently moving across the car hoods. Rows of tiny parallel scratches in the paint behave as specular line-scatterers. If the angle between the observer and the surface is changing, these scratches produce patterns of reflected sunlight that appear to move across the hood. On one occasion I noticed a particularly striking example. The highlights on a black car hood were not just highlights, they were large images, and they appeared deep within the surface. They also had a distinct structure: close-spaced radial filaments surrounding an array of whitish patches. After some time in observing these, the nature of the shapes suddenly became clear. They were images of a lambs wool polishing mitt, complete with matted fibers in the center, straighter fibers forming a radial halo, and with the glowing shape of a human hand visible in the matted fibers. Even more intriguing, most of these patterns appeared to float many centimeters deep within the car hood. Apparently they were created as the hood was polished by a gritty polishing-mitt, and the mitt produced a large number of microscopic, curved, and somewhat parallel scratches in the paint.

	Figure 1 Car hood after careful "polishing" by a dry paper towel

The phenomenon was easily reproduced by rubbing a very dirty car hood with a dry paper towel. Figure 1 shows the results. The hand that holds the towel must not rotate, and its overall trajectory must be a circular arc of ~90deg with radius < 30cm. The resulting scratch patterns were observed under sunlight. Images of handprints were clearly being reconstructed, with the images appearing deep within the painted hood rather than upon its surface. The depth of each image was somehow proportional to the radius of the curved scratches and to the angle of illumination. At some observing angles the images would even appear to float above the surface rather than below. Some handprint images were quite sharp, with the surface texture of the paper towel clearly visible as in Fig. 2 below.

	Figure 2 Close-up photo of a "virtual image" handprint pattern

1.2 Independent Re-invention

Although the greater optics community may not be familiar with the above phenomenon, it is actually somewhat well known. Plummer and Gardner [2]] analyzed a similar effect in 1992 after noticing 3D images during inspection of metal mirrors subjected to an automated lapping process. Artist Gabriel Liebermann harnessed the effect even earlier in 1981 to produce 3D images in abraded metal, such as his work entitled World Brain described by Garfield [1]. Apparently Liebermann kept his technique proprietary, although Garfield mentions that it was based on an NC milling machine. And as Abramson [3] points out, the inventor Hans Weil created a version in the mid 1930s that encoded stereo pairs of 2D photographic images in the form of angled parallel abrasions with two different angles, a sort of "multiplex stereogram" version of the above technique. The present paper is based on work from 1994.

---

2. CURVED LINE-SCATTERERS PRODUCE IMAGES

Plummer and Gardner [2] show that the scratches producing the above images act as curved line-scatterers. (Their paper delves into the geometry far more rigorously than below, so please refer to ref [2] for a full treatment.) For simplicity the dipole pattern of light reflected from a line-scatterer of thickness << wavelength is ignored. Also assume that these scratches lack a blaze angle and reflect the incoming light equally in all radial directions, where each incoming ray is scattered as a hollow cone of rays. In following diagrams I symbolize the scatterer as a bent rod, although in reality it may have internal structure.

A curved line-scatterer formed from a segment of a circle can act as a convex or concave spherical mirror, producing either an astigmatic real or virtual image of the distant point source. In figure 3, we have a curved line-scatterer in the XY plane of radius r and center of curvature C. Light from a distant point source PS shines downwards and reflects from the scatterer. Each ray from PS scatters into a hollow cone of rays where the angle of the cone depends on the position of the cone's vertex along the length of the curved scatterer. The cones' angles depend on the angle a. The axis of the cone is tangent to the curved scatterer.

	Figure 3 Light from a distant point source scatters off a curved line-scatterer, where each ray from the point source produces conical sets of scattered rays.

In fig. 3, other parallel rays from PS striking the scatterer at different locations produce other cones. The rays from all these cones taken together, if extended through the cone vertices to locate virtual image points, intersect to form two separate astigmatic virtual image loci. One virtual image locus is positioned at the vertices of the cones: it is distributed along the core of the line-scatterer itself. A second virtual image is shown in fig. 4.

	Figure 4 The intersection of lines extended from sets of cones of rays produces an astigmatic virtual image of the light source: an ellipse-shaped locus of points in the YZ plane

When the scatterer in fig. 4 is lit from above, and for small values of angle a, the intersection of the rays cast backwards from the cones forms a virtual image in space. Rays a1, a2, etc. from each cone are projected backwards as a1', a2', etc. They intersect with corresponding rays from other cones, and the intersection takes the form of a U-shaped locus in the YZ plane. The rays form a tilted cone intersecting a plane, so the locus is a conic section, and for finite illuminator distances the locus should be an ellipse. The major axis of this ellipse-shaped virtual image is aligned with the rays from the distant source PS, with the nearer focus coincident with the center of the line-scatterer, while the distant focus coincides with the location of the illuminator.

	Figure 5 Light from a distant source creates a virtual image point P when observed by eyes positioned at O1 and O2

Human stereopsis lets observers perceive either of the two virtual images. In figure 5 an observer's eyes are positioned at O1 and O2, and they observe the ellipse virtual image described in figure 4. However, the observer is able to see only a tiny segment of this ellipse by looking through the curved scatterer, as if the scatterer behaves as a slit-aperture. This slit masks the vertical extent of the ellipse virtual image locus, so the observer's eyes at O1 and O2 perceive only a single bright point P positioned somewhere behind the curved scatterer. A second point lying on the ellipse-image in the positive half of the Z plane is not seen. This second point can be observed only if the eyes are moved to the opposite side of the XY plane. Or in other words, the line-scatterer produces both a reflection-mode image sent to one side of the XY plane, and a transmission-mode image sent to the other. This resembles conventional hologram optics.

The source of illumination in fig. 5 is required to have very limited extension, otherwise the observed image point becomes a horizontal line segment and the image suffers an astigmatic blur effect. This blur is similar to that observed in Rainbow holograms [4]. The blur is reduced when radius r is made shorter (less blur when the virtual image P is very close to the film plane.) Thus we see another similarity to conventional holography.

Note that the curved scratch scatters light in the vertical Y dimension as well, so if human eyes are positioned at points O3 and O4 in figure 5, stereopsis would force them to perceive the light as coming not from point P, but instead from the vertices of the cones of rays located within the curved scatterer itself. For this reason the image produced by the curved scatterer can have horizontal parallax only. If viewed with eyes turned 90 degrees, the image lacks depth: it appears at the location of the scratches on the surface of the plate. Obviously this is a major similarity to Rainbow holography.

What is the depth (focal length f) of point P in figure 5? Knowing that P lies on the ellipse-shaped locus in figure 4, the value for f must vary with the position of the observer and with the angle of the rays from the distant point source. In figure 5 we have an observer positioned broadside to the curved scatterer, with eyes at O1 and O2. Inspecting the diagram from a position above it, we find that certain rays from each cone described in fig 4 can be extended through the cone vertex to converge on the virtual image point P in figure 5. Angle b is equal to angle a since the lines forming them are both part of the same cone of rays. Since the side of the triangle opposite to angle b is also shared by the side of a second triangle opposite to angle a, we have identical triangles, therefore focal length f simply equals radius of curvature r of the scatterer. This is only true for vertically-illuminated "broadside" views of the scatterer under vertical illumination and for small angles a (i.e. for observer distance >> interocular spacing.) . For example, if the light from the scratch was viewed from above (i.e. looking down from the location of point source PS), the curved scatterer would act instead like a spherical mirror, and the focal length measured from the scatterer to the virtual image point would then lie between the scatterer and point C. Focal length f would then have the usual value of 0.5r. On the other hand, if the observer's eyes O1, O2 are moved downward in the negative Y direction, the image point P migrates upwards along the ellipse, and the image depth increases (it eventually becomes nearly the same as the distance to point source PS.) But for viewing angles close to those shown in figure 5, the virtual depth remains close to the value for r.

Virtual depth of a point approximates the scratch radius r. This is a useful result. Suppose we were to employ a double-pointed compass (a dividers) to scribe a curved scratch onto a plastic plate. Make the scratch resemble a circular arc as shown in figure 5. We could hang this plate on a wall, illuminate it with a distant point source placed vertically above the plate, then observe the plate with two eyes oriented horizontally. Wed see a glowing spark of light shining from within the curved scratch. The virtual depth of the glowing spot would be the same as the radius of the scratch. Now suppose we lay down several hundred similar scratches, each with a different XY position, plus a Z position as set by the spacing of the dividers. Could we not draw arbitrary objects or scenes in 3D as sets of glowing points? This actually works very well.