Skip to content
Flattum Center for Kinetic Research

Parallax geometry · Kinetic form

Algorithms

There is really only one algorithm here. Cut every image into strips, lay the strips down in rotation, and put something in front of them that shows one set at a time. Everything else — the fold, the lens, the mask — is a choice of selector.

This page is the procedure. The relations behind it, with their derivations and live figures, are on Mathematics.

Interlacing

Given K images cut into N strips each, the sheet is K·N strips in rotation:

for strip in 0 .. N-1:
    for plate in 0 .. K-1:
        emit sliceOf(image[plate], strip)

That is the whole of it. The 3D renderer skips the step entirely — it slices in texture coordinates instead, mapping facet i of rib r to slice r of plate i — but a physical sheet has to be interlaced for real.

Choosing a selector

Each buys something and charges for it. The differences are not stylistic, and this is the decision the rest of the work follows from.

Fold
Selects
by where you stand
Holds
two images, cleanly
Costs
a crease per strip, and 41% more paper than the finished piece
Multi-faceted fold
Selects
by where you stand
Holds
three to five, none of them cleanly
Costs
a crease per facet, and images that bleed into each other
Lens array
Selects
by refraction
Holds
a dozen or more
Costs
a pitch calibration that must be right before you print
Sliding mask
Selects
by where the mask sits
Holds
a whole loop, as motion
Costs
all but 1/K of the light
Mirrored cylinder
Selects
by reflection
Holds
one image, from one place
Costs
no interlacing at all — and most of the picture's detail

The sequence

  1. 01

    Choose the images

    They have to differ in more than subject. Two plates that share a palette or a composition read as one muddled picture at every angle rather than two clean ones at their stations. Opposed on colour and on form is the reliable rule.

  2. 02

    Choose the selector

    The table above is the decision. Most of what follows is fixed by it.

  3. 03

    Interlace

    Cut each image into N strips and lay them down in rotation. On a fold the strips are unequal beyond two plates, because each is printed at its own slant length.

  4. 04

    Print

    At 100%, on stock heavy enough to hold a crease. Each plate needs roughly half the sheet's horizontal pixels for two plates, more for the outer facets beyond that — the Interlacer states the figure and refuses to pretend an undersized file will do.

  5. 05

    Trim and crease

    Cut to the trim marks, then score every tick before folding any of them. Solid marks fold toward you, dashed away. Creasing in order, rather than folding as you go, is what keeps the accumulated error from walking the last strips off the edge.

  6. 06

    Mount

    A folded relief cannot be glazed flat, so it goes onto a rigid board with the valleys adhered down, in a box frame deep enough to clear the ridges.

The instruments

Each step above has something here that does it.

  • FoldTwo to five images on one folded sheet
  • InterlacerImages in, printable sheet out
  • LenticularThe pitch you must actually print at
  • Barrier gridA frame loop under a sliding mask
  • AnamorphosisNothing flat, a picture in the mirror