Computational Photography, an AI-powered Slopendium — 31 Light fields and plenoptic cameras

▸Part 31 LIGHT FIELDS AND PLENOPTIC CAMERAS⧉

---

▸31.1 Light fields 101⧉

31.1.1 Capture rays, not pixels⧉

31.1.2 The plenoptic function, reduced to four dimensions⧉

31.1.3 Ray and point are dual⧉

31.1.4 Reading the 4-D structure through its 2-D slices⧉

31.1.5 Rendering a new view by looking up rays⧉

31.1.6 Light fields vs. plenoptic and radiance⧉

▸31.2 Light field cameras⧉

31.2.1 Integral photography: Lippmann's fly's-eye plate (1908)⧉

31.2.2 The plenoptic camera: a microlens array on the sensor⧉

31.2.3 Lytro: the consumer plenoptic camera⧉

31.2.4 Commercial light-field cameras beyond Lytro⧉

31.2.5 Camera arrays: a grid of full cameras⧉

31.2.6 The spatial↔angular tradeoff⧉

31.2.7 Two strategies, and why arrays and microlenses are dual⧉

▸31.3 Other light field acquisition setups⧉

31.3.1 One camera on a gantry: sample the aperture in time⧉

31.3.2 Handheld, unstructured capture: let the poses be irregular⧉

31.3.3 Catadioptric capture: one sensor, many viewpoints at once⧉

31.3.4 Coded aperture in time: sweep the pupil itself⧉

31.3.5 The everything-else, mapped⧉

▸31.4 Refocusing and synthetic aperture⧉

31.4.1 Reconstructing a photo: it is all about which rays you sum⧉

31.4.2 Digital refocusing is shift-and-add⧉

31.4.3 A focal stack from one capture, and an all-in-focus image⧉

31.4.4 Fourier-slice photography: the fast version⧉

31.4.5 Reading refocus and depth off an epipolar slice⧉

31.4.6 Synthetic aperture: an aperture the size of a room⧉

▸31.5 Aberration correction in light fields⧉

`fig-light-journey` · FUNDAMENTALS part opener — the journey of light: source → scene (reflection) → lens → sensor / retina → visual system, each step labeled with its chapter ✅

31.5.1 An aberration is misrouted rays⧉

31.5.2 Re-routing each ray to the ideal-lens position⧉

31.5.3 The trade: ray bookkeeping instead of glass, and only what you sampled⧉

▸31.6 Light field aliasing and 4D Fourier analysis⧉

31.6.1 The light field is a sampled signal, and its samples are viewpoints⧉

31.6.2 Depth is slope is spectral orientation⧉

31.6.3 The bowtie: a spectrum shaped by the scene's depth range⧉

31.6.4 How densely must you sample? The plenoptic-sampling bound⧉

31.6.5 Geometry buys back samples: the depth-vs-views tradeoff⧉

31.6.6 Under-sampling looks like a ghost⧉

31.6.7 Where this sits⧉

▸31.7 Lumigraph and shape priors for sharper light field rendering⧉

`fig-histogram-exposure` · one real scene at three exposures (−2 / as-shot / +2 stops), each with its luminance histogram — distribution slides left→right, clipping spikes pin to the edge; crushed-shadow / clipped-highlight callouts (BASIC histograms) ✅

31.7.1 Pure light-field rendering blurs because it has no shape⧉

31.7.2 The Lumigraph: reproject onto a geometry proxy, then blend⧉

31.7.3 Unstructured inputs: free-hand views, no grid required⧉

31.7.4 Surface light fields⧉

▸31.8 Light field microscopy⧉

`fig-demosaick-snapshot` · even a plain snapshot needs computation: one color per pixel (Bayer mosaic) → demosaicked RGB ✅

31.8.1 The optical setup: a microlens array at the intermediate image plane⧉

31.8.2 The spatial-versus-angular trade: a coarse 3-D volume⧉

31.8.3 Recovering the volume: synthetic refocusing, then 3-D deconvolution⧉

31.8.4 Why single-shot 3-D is the whole point⧉

▸31.9 Light field networks⧉

⬜ figure not yet created

`fig-light-field-network` (a small MLP maps a parameterized ray → RGB fig-light-field-network

`fig-lightfield-refocus` · **light field → refocus → merge:** refocus a light field by **shift-and-add** (refocused near / refocused far) → the synthesized focal stack merged to one all-in-focus image. placed in **[[11 Advanced computational photography]] § Refocusing** (forward-ref'd from the DoF & focal-stack chapters); registered here next to its focus-stacking siblings ✅

`fig-medium-expressive` · two-panel: the medium's limitations as expression — a backlit estuary with shadows accepted to black (limited display contrast as a graphic), and three kids frozen mid-jump (a single time slice held to contemplate); intro counterpoint to "alleviate limitations" ✅

31.9.1 NeRF: a radiance field rendered by volume integration⧉

31.9.2 Light field networks: a ray straight to color, in one evaluation⧉

31.9.3 A family of neural light fields⧉

31.9.4 Test-time training: the network as the per-scene prior⧉

31.9.5 Hand-off: from rays to radiance fields and generation⧉

▸31.10 Practical aspects of light field cameras⧉

⬜ figure not yet created

may be a sidebar instead.

---

31.10.1 Do you lose all that resolution?⧉

31.10.2 Can you do video? Is it practical? — the bandwidth problem⧉

31.10.3 The time dimension is high speed⧉

31.10.4 Adjacent frontiers, briefly⧉

31.10.5 So, do I get my camera?⧉