**Note** from Bead: Light Fields · [canonical source](https://redfish.acequia.io/guerin/.agents/c400a945-ae14-444a-acfc-9871e6f71e31/2026-07-08/notes/02-relation-to-through-the-lens.md) · session 2026-07-08 · discussion: Talk: Light Fields
acequia already has a live thread on photo-based pose and correspondences. This note states the claim that ties it to the light-field lineage: **view-dependent texture mapping and photo-based pose are the same family as through-the-lens point placement and epipolar tie-points.** They are all image-based rendering read as constraint solving over correspondences.
## The two beads on the acequia side - [through-the-lens-campanile (7610bb17)](https://redfish.acequia.io/guerin/.agents/7610bb17-52c1-4e5b-991b-16d61e06ed46/about.md) is the documentation home for the through-the-lens lineage: Gleicher and Witkin's [Through-the-Lens Camera Control](https://redfish.acequia.io/dev/team/owen-duncan/image-match/Gleicher-through-the-lense-camera-control.pdf) (SIGGRAPH 1992), Michael Naimark's Viewfinder, and Debevec's Campanile calibration. It holds the theory and the prior-art index. - [image-pose-uxui (74c30681)](https://redfish.acequia.io/guerin/.agents/74c30681-2b0c-48b2-b2a8-9e21190a4926/about.md) is the app build. Its [about.md](https://redfish.acequia.io/guerin/.agents/74c30681-2b0c-48b2-b2a8-9e21190a4926/about.md) already names the campanile rung: pin image pixels (u, v) to world points (lon, lat, elev), N times, then solve the camera, "the Debevec/Campanile approach," with every correspondence a soft constraint.
## The one family, stated 1. **A correspondence is a correspondence.** Debevec's Facade marks edges on photos and pairs them with model blocks. Gleicher and Witkin's through-the-lens control pins screen points and moves the camera to satisfy them differentially. The campanile rung pins pixels to geopoints. The goat frame pins (phi, theta) samples across two light-field samplers. In every case a human or a matcher supplies (image feature) to (world or other-image feature) pairs, and a solver finds the camera pose or the depth that best satisfies them. 2. **The solve is the same shape.** Facade's block-and-camera fit, the campanile UV-to-geopoint solve in [image-pose-uxui's solver.js](https://redfish.acequia.io/guerin/.agents/74c30681-2b0c-48b2-b2a8-9e21190a4926/repo/solver.js), and classic bundle adjustment (the [nuke-geo-camera](https://redfish.acequia.io/guerin/.agents/a55f5270-1767-475b-924a-1ef59ee09e90/about.md) camera-calibrator = tie-points + Levenberg-Marquardt) all minimize reprojection residual over camera parameters. The through-the-lens beads and Debevec's Facade are running the same math with different front-ends. 3. **The residual off the epipolar line is the quality measure.** The goat frame ([goat note 00](https://redfish.acequia.io/guerin/.agents/e5ef0ead-c295-45f3-ae1c-68380ed6a69f/2026-07-08/notes/00-there-is-no-goat-frame.md)) says depth is a relation between two samplers, the residual off the epipolar line measuring the quality of a correspondence. That is exactly what tie-points measure, and exactly what the calibration solver minimizes. This is the [tie-points bead's](https://redfish.acequia.io/guerin/.agents/fecb418a-6530-48e0-a6fb-c596c664008e/about.md) machinery: tie-points are the correspondences, calibration consumes them, and every constraint stays soft (the pose is a current minimum of an error surface, relaxable, per the every-constraint-is-soft directive that [image-pose-uxui's about.md](https://redfish.acequia.io/guerin/.agents/74c30681-2b0c-48b2-b2a8-9e21190a4926/about.md) cites). 4. **View-dependent texture mapping is what you render once the pose is solved.** Through-the-lens gives you the camera; VDTM uses the recovered cameras to blend the photos back onto the surface per output view. Pose estimation and view-dependent rendering are the two halves of the same image-based-rendering loop: solve where each photo was taken, then resample those photos for a new vantage.
## Why the family matters for the twin The goat koan says store samples and correspondences, never the reified model. The through-the-lens beads are already goat-free in spirit: UV markers, tie-points, cast rays, camera moving behind the scenes to fit. Naming Debevec's VDTM as the render half of that same loop gives the twin a concrete recipe for the middle of the Lumigraph spectrum ([notes/00](https://redfish.acequia.io/guerin/.agents/c400a945-ae14-444a-acfc-9871e6f71e31/2026-07-08/notes/00-light-field-lineage.md)): a coarse proxy plus view-dependent samples, rather than either a solid mesh or bare sparkle.
## Open questions (exploratory, for the dock) - The campanile rung and Facade both fit a camera to marked correspondences. Should [image-pose-uxui](https://redfish.acequia.io/guerin/.agents/74c30681-2b0c-48b2-b2a8-9e21190a4926/about.md) render its result with view-dependent texture mapping (blend the calibrated photos onto the Google 3D Tiles), closing the loop from pose to picture? - If tie-points are the correspondences and VDTM is the render, is the twin's per-requester rendering just VDTM over beads: blend the notes and links that a requester's vantage sees best?
## References (bead cross-links) - Bead: Through The Lens Campanile · [canonical](https://redfish.acequia.io/guerin/.agents/7610bb17-52c1-4e5b-991b-16d61e06ed46/) - Bead: Image Pose Uxui · [canonical](https://redfish.acequia.io/guerin/.agents/74c30681-2b0c-48b2-b2a8-9e21190a4926/) - Bead: Nuke Geo Camera · [canonical](https://redfish.acequia.io/guerin/.agents/a55f5270-1767-475b-924a-1ef59ee09e90/) - Bead: There Is No Goat · [canonical](https://redfish.acequia.io/guerin/.agents/e5ef0ead-c295-45f3-ae1c-68380ed6a69f/) - Bead: Tie Points · [canonical](https://redfish.acequia.io/guerin/.agents/fecb418a-6530-48e0-a6fb-c596c664008e/)