Specular holography

By tooling an spot-illuminated surface to control the flow of specular glints under motion, one can produce holographic view-dependent imagery. This paper presents the differential equation that governs the shape of the specular surfaces, and illustrates how solutions can be constructed for different kinds of motion, lighting, host surface geometries, and fabrication constraints, leading to some novel forms of holography. 1 Holography via specular glints Holography, broadly understood, is any means of making a light field that contains a range of views of a virtual 3D scene. Using only geometric optics, it is possible to construct a sparse holographic light field by tooling a 2D surface to control the bundle of rays it directionally reflect or refracts. The author has been exhibiting such “specular holograms” in public art venues for a number of years. This paper introduces the differential geometry that determines the geometry of the optical surfaces, and develops solutions for several viewing geometries. The idea of specular holography predates wavefront holography by many years. The centuries-old arts of metal intaglio and engine turning exploit the fact that sharp glints on smooth convex specular surfaces appear—via stereopsis—to float below the surface. Beginning in the 1930’s, several authors have analyzed this effect [1, 2, 3, 4] and proposed to use it for view-dependent imagery [5] and holography [6, 7, 8, 9, 10, 11] using surface scratches or parabolic reflectors. All give reasonable approximations for small ranges of viewpoints; this paper gives the exact geometry for all viewpoints. Fig. 1(a) diagrams the optical principle: Given a point light source and a virtual 3D point in the holographic scene, one shapes a smooth optical surface that reflects or refracts a specular glint to the eye along every unoccluded sightline through the point. The surface appears dark to other sightines. Lacking any other depth cues, the brain parsimoniously but incorrectly concludes that both eyes see a single specularity located at the virtual point. Changes in viewpoint reinforce the 3D percept via motion parallax. A specular hologram combines a large number of such optical surfaces, all designed to conform to a 2D host surface so that they can be easily formed by conventional fabrication techniques such as milling, grinding, stamping, ablation, etc. The host surface has limited surface area, so the 3D virtual scene is represented by a sparse 1 ar X iv :1 10 1. 03 01 v1 [ cs .G R ] 3 1 D ec 2 01 0 sampling of its points, usually a stippling of its surfaces. As an artistic matter, there are many ways to algorithmically produce visually informative and pleasing stipplings. Similarly, the optical surfaces have a simple differential geometry, developed below, that admits a variety of distinct solutions. 2 Geometry of a single-point optical surface Consider making a specular hologram of a single point p ∈R3. Let i ∈R3 be a point illumination source, and let e∈R3 be the location of the eye. Typically the viewpoint e is parameterized by azimuth θ and elevation φ ; p and i can vary parametrically as well. Throughout this paper, a boldfaced variable will refer interchangeably to a three-vector and the function giving its locus. The holographic effect is achieved by producing a specular glint on each sightline through virtual point p at a real point s ∈ R3 where the axis of reflection or refraction is parallel to the optical surface normal n ∈R3. The ideal optical surface is a continuous locus of points swept by s as p, i,e vary. Generally, there is an infinite foliation of such surfaces, each a different distance from p on the sightline. To fabricate the hologram, a host surface H with local normal N ∈ R3 is tooled to conform piecewise to this foliation, changing its local normal to n. As in computational mirror design [12], one needs to reconstruct a surface from a field of arbitrarily scaled normals. Here it is more convenient work with tangent spaces than with normals. Let Ts = (t1, t2) ∈ R3×2 be any nondeficient basis of the optical surface’s local tangent plane at s. For example, Ts could be the Jacobian Js(θ ,φ) of s with respect to θ ,φ . Then the constraints that determine the optical surface are 〈 Ts, i− s ‖i− s‖2 η1 +η2 e− s ‖e− s‖2 〉 = (0,0) (normality) (1) 〈 (e−p)⊥,(s−p) 〉 = (0,0) (colinearity) (2) ‖s−H‖ ≤ ∆ (conformity) (3) where η1,η2 are the refractive indices of materials a light ray crosses before and after encountering the optical surface; 〈·, ·〉 and ‖ · ‖ are the Euclidean inner product and vector norm; and x⊥ is any orthogonal basis of the nullspace of x, i.e., 〈 x⊥,x 〉 = 0. For reflection holography, η1 = η2 6= 0. Differential Eq. (1) (normality) states that the optical surface is perpendicular to the axis of reflection or refraction. Eq. (2) (colinearity) ensures that the specularity is on the sightline to p. Eq. (3) (conformance) specifies that the optical surface lie within a thin shell of thickness 2∆ that conforms to the host surface H . The colinearity constraint can be algebraically eliminated by sliding the eye forward along the sightline ep until it coincides with p or s. It is easily shown (by substitution) that Eq. (1) is satisfied by a foliation of revolute bicircular quartics with foci at i and p. The quartics are revolved around the major axis connecting i and p. The most common case of interest is a reflection hologram with static point light source i and virtual point p. In this case the quartics simplify to conics. Specifically, if p is in front of the reflecting surface, the solution is the interior surface of any prolate ellipsoidal reflector with eccentricity ε < 1. If point p is behind the reflecting surface,