**Note** from Bead: Through The Lens Campanile · [canonical source](https://redfish.acequia.io/guerin/.agents/7610bb17-52c1-4e5b-991b-16d61e06ed46/2026-06-16/notes/01-document-index.md) · session 2026-06-16 · discussion: Talk: Through The Lens Campanile
Catalog of the prior-art folder for future reference. Source URL: **https://redfish.acequia.io/dev/team/owen-duncan/image-match/** (large mp4s stay there; cite by URL). Captured originally in chat session `7b79c9e2` (yt-dlp downloads into Owen Duncan's team folder). | File | Size | What it is | Use it for | |---|---|---|---| | `Gleicher-through-the-lense-camera-control.pdf` | 165 KB | **Gleicher & Witkin, "Through-the-Lens Camera Control," SIGGRAPH 1992.** The foundational paper. §3 = the constraint solver (eqs. 8, 11); §4 = the 8-vector quaternion camera model + analytic derivatives (page 7); §5 = the control gallery. | The math we implement. Equation/section numbers in `plan.md` refer here. | | `Through_the_Lens_Camera_Control-aPraFT6JilE.mp4` | 18.9 MB | The accompanying **SIGGRAPH '92 video** for the paper. | The "table-decoration" segment (~5:13) is the closest worked TTL example. | | `Through_the_Lens_Camera_Control-aPraFT6JilE.en.vtt` | 22 KB (552 lines) | Transcript of the above. | Quote/search the TTL video without re-watching. | | `Viewfinder_-_How_to_Seamlessly_Flickrize_Google_Earth…-LRAj21y23r8.mp4` | 128 MB | **Michael Naimark's "Viewfinder" Google Tech Talk** ("How to Seamlessly Flickrize Google Earth"). Direct ancestor of `#image-pose-uxui`. | The whole posed-photo-onto-Earth idea. Key timestamps below. | | `Viewfinder_-…-LRAj21y23r8.en.vtt` | 397 KB (9 288 lines) | Transcript of the Viewfinder talk. | Search the talk by keyword/timestamp. | | `campanile-camera-calibration.md` | 7.9 KB | **Shot-by-shot analysis of the 27:05 campanile demo** (Owen Duncan / CE). The 4-UV-point DOF table, the LM ~500-iters/match note, vertices-vs-non-semantic-features (Debevec vs Duncan). | The canonical description of the LM-resectioning paradigm our `solver.js` implements. | | `plan.md` | 13.8 KB | **Plan for a three.js Gleicher–Witkin TTL demo** of the tower (procedural Sather Tower, quaternion camera, differential solver, drag-the-projection UI, truth toggle). | The runnable spec for the *differential* TTL paradigm (the LM one is already built in `#image-pose-uxui`). |
## Key timestamps — Viewfinder talk (`LRAj21y23r8`) - **~22:10** — Debevec discussion: polyhedral vertices vs. non-semantic features for correspondences. - **~25:30–26:30** — Viewfinder's 2D↔2D hack against a fake Google-Earth frame-grab server (COM API). - **~26:42** — "two points should nail all four coordinates" (the 2D↔2D, 4-DOF case). - **~27:05–27:21** — the standalone **2D↔3D campanile-vs-wireframe** demo; 4 UV pins, "~500 iterations per match." Analysed in `campanile-camera-calibration.md`.
## Cross-refs - Implemented by `#image-pose-uxui` (`74c30681`) — `repo/solver.js` (LM) + the campanile pin loop. - This bead's [notes/00](00-goal-and-approach.md) maps the two paradigms; [notes/02](02-modern-ttl-survey.md) surveys modern approaches.