Knowledge Vault 5 /27 - CVPR 2017
Computational Imaging on the Electric Grid
Mark Sheinin, Yoav Y. Schechner, & Kiriakos N. Kutulakos
< Resume Image >

Concept Graph & Resume using Claude 3 Opus | Chat GPT4o | Llama 3:

graph LR classDef flicker fill:#f9d4d4, font-weight:bold, font-size:14px classDef brf fill:#d4f9d4, font-weight:bold, font-size:14px classDef acam fill:#d4d4f9, font-weight:bold, font-size:14px classDef analysis fill:#f9f9d4, font-weight:bold, font-size:14px A[Computational Imaging on
the Electric Grid] --> B[Light flickers from
AC power. 1] B --> C[BRF: bulbs temporal
flicker profile. 2] C --> D[BRF: type, phase,
spectral energy. 3] C --> E[D-Light: BRFs for
bulb recognition. 4] B --> F[AC lighting modeled
with matrices. 5] F --> G[Inverse problem recovers
single-source images. 6] G --> H[Window reflections separated
by illumination. 7] G --> I[Decomposition allows virtual
bulb swapping. 19] B --> J[Cameras struggle with
sub-cycle imaging. 8] J --> K[A-CAM: synced mask
captures sub-cycles. 9] K --> L[A-CAM handles dynamic
range cleverly. 10] K --> M[A-CAM syncs with
voltage fluctuations. 11] K --> N[A-CAM: DMD mirror
as shutter. 20] A --> O[Experiments demonstrate
single-source decomposition. 12] O --> P[Passive sensors analyze
AC scenes. 13] P --> Q[AC analysis enables
new applications. 14] B --> R[Electric grid is
visual cue. 15] B --> S[Daylight treated as
DC bulb. 16] A --> T[Works with different
source BRFs. 17] G --> U[Inverse problem needs
multi-frame sampling. 18] class B,R,S flicker class C,D,E,T brf class F,G,H,I,U analysis class J,K,L,M,N,O,P,Q acam


1.- Artificial light flickers due to AC power, creating a unique temporal signature for each bulb type.

2.- Bulb Response Function (BRF) defines the temporal flicker profile of a bulb within a single AC cycle.

3.- BRF depends on the bulb type, phase, and spectral energy.

4.- D-Light database compiles BRFs for common bulbs, enabling bulb type and phase recognition in scenes.

5.- Image formation under AC lighting is modeled using sub-cycle images and light transport matrices.

6.- Inverse problem solving recovers single-source images, allowing scene rendering under different illumination.

7.- Window reflections can be separated using the constant nature of sunlight and flickering indoor bulbs.

8.- Conventional cameras struggle with sub-cycle imaging due to short exposures and read noise.

9.- A-CAM uses a programmable optical mask synced with the flicker frequency to capture sub-cycle images.

10.- A-CAM handles dynamic range by unblocking dim pixels in all cycles and bright pixels in fewer cycles.

11.- AC voltage frequency fluctuations are addressed by real-time sensing and syncing of the A-CAM.

12.- Lab and street experiments demonstrate successful decomposition of scenes into single-source images.

13.- Passive sensors can now analyze AC-illuminated scenes without active illumination.

14.- AC illumination analysis opens up new possibilities in computer vision tasks and broader applications.

15.- The electric grid itself is a newly induced visual cue, with AC sequences revealing grid structure and behavior.

16.- Daylight acts as a constant illumination source, treated as another bulb with a DC response.

17.- The method works as long as the BRFs are different for each source.

18.- Inverse problem solving requires sampling the flicker cycle in multiple frames (e.g., 26 frames).

19.- Decomposed scenes allow virtual swapping of bulbs without physical changes.

20.- A-CAM uses a DMD mirror as a shutter, transmitting light with minimal loss when unblocked.

Knowledge Vault built byDavid Vivancos 2024