Some years ago the following scheme was proposed to encrypt an image. Bundles of glass fibers had then been found useful for moving images short distances. The fibers were essentially the Cartesian product of one fiber and an irregular 2D arrangement of pixels of the transported image.

To prepare for encoding images one would take such a bundle, scramble the middle of the bundle while keeping the ends fixed. After scrambling some substance would be added near the middle to keep the scrambled fibers fixed in their scrambled state and then the bundle would be cut there. The two ends of the bundle would be physically transported to the two places between which secure image transmission was needed. To send an image one half of the bundle would be used to produce a film image with the pixels scrambled as the bundle was scrambled. The film (or its image) is transported to the other place and decoded by alignment of the image with the scrambled end of the other half of the original bundle.

This fails as a crypto scheme because a modest collection of cipher text (cipher image in this case) reveals correlations between distant pixels which in fact represent close pixels in the plain image. With such closeness information a total map can be produced.

This sort of analysis is thought to explain how nearby points of the retina come to be wired to nearby points in the visual cortex. (As if “fire together -> wire together”)