Projected Structure: When a Video Stops Being a Surface
A video is orthographically projected onto a hidden 3D structure. As the camera sweeps around the volume, the continuous image fractures into displaced fragments — revealing the geometry that the projection was “hiding.” At a 90° sweep, the structure collapses back into invisibility and the video reconstructs as a coherent surface. The work tests how viewpoint governs legibility, trust, and spatial inference in computational imagery. In effect, it tests the threshold at which the video stops functioning as evidence of a place and starts reading as computation.
Abstract
This case study stages a controlled conflict between image coherence and spatial truth. From one privileged viewpoint, the footage reads as a stable 2D scene. As the viewpoint shifts, the same footage becomes evidence of a 3D volume — splitting across faces and exposing the mechanism that previously remained invisible. The system treats the camera not as a neutral observer, but as a switch that determines what counts as “the image.”
Research Question
When does a moving image behave like a surface, and when does it reveal itself as a projection distributed across space?
More specifically: how much viewpoint change does it take for an image to stop being believed as a continuous record and become read as a constructed object?
System / Method
A source video is mapped orthographically onto a 3D proxy structure (a hidden volume).
- The camera performs a controlled rotational sweep around the proxy.
- As the view departs from the privileged axis, projection continuity breaks — the footage fragments across multiple planes, spatial discontinuities become legible, and the proxy reads as “the cause” of the image.
- At approximately 90°, the proxy becomes visually minimized and the video reconstructs into a coherent image plane.
Empirical Sequences
Three variant studies — different source footage and proxy topologies — each carrying the full arc from coherent surface, through fracture, to reconstruction within a single continuous camera sweep.
Observations
- Coherence is viewpoint-dependent: continuity is not an inherent property of the video, but an alignment event.
- The reveal is procedural, not decorative: the fracture pattern communicates underlying geometry and mapping logic.
- Reconstruction behaves like a perceptual snap: at the correct alignment, the viewer’s attention shifts from “object” back to “image.”
Key Contribution(s)
- Demonstrates a compact, readable system where camera motion functions as an epistemic test: it determines whether the viewer interprets content as documentation or construction.
- Bridges cinematic language (the sweep/reveal) with computational imaging logic (projection, proxy geometry, alignment).
- Produces a repeatable template for future studies using different source footage, proxy topologies, and sweep ranges.
Why Computational Imaging
Although the source material is aerial/urban footage, the project’s subject is image formation under constraint: how projection and viewpoint produce (or destroy) continuity, and how spatial proxies can “carry” an image until the view reveals the cheat.
Tools / Pipeline
3D software for proxy modeling and UV/projection setup, orthographic projection mapping, and rendered camera sweep. Post used for edit, pacing, and presentation.
Next Iterations
- Test different proxy geometries to control how the fracture reads (grid, radial, architectural proxies).
- Quantify the “break point” (angle threshold) where viewers stop reading the image as continuous.
- Add a split-screen diagnostic render (clean view / proxy-only / UV projection debug) for tighter research documentation.