How does a plenoptic camera enable post-capture refocus?

WHAT THIS TESTS The interviewer probes whether you understand that a light field captures directional ray information, not just a flat image, and that refocusing is a reintegration of that data rather than image sharpening.

A GOOD ANSWER COVERS A conventional sensor integrates all rays hitting a pixel, discarding their direction. A light field is the 4D function describing radiance along rays, parameterized by two-plane coordinates that encode both where a ray crosses the aperture and where it lands. A plenoptic camera captures this by inserting a microlens array just in front of the sensor: the main lens forms an image on the microlenses, and each microlens spreads the incoming cone across several sensor pixels, so each pixel records a particular direction through the aperture. From this raw mosaic you can extract sub-aperture views, slightly different viewpoints of the scene. To refocus, you choose a virtual focal plane and integrate the rays that converge there; in the 4D representation this is a shear of the light field along the directional axes followed by summation over directions, which synthesizes an image focused at the chosen depth. Different shears yield different focal planes from the same capture.

COMMON WRONG ANSWERS Describing it as a multi-lens array camera only, with no directional encoding. Confusing post-capture refocus with deblurring or with depth-from-defocus alone. Forgetting the spatial-versus-angular resolution trade-off the microlens array imposes.

LIKELY FOLLOW-UPS The resolution trade-off, computing depth from the parallax between sub-aperture views, synthetic aperture imaging, and limitations of refocus range.

ONE CONCRETE EXAMPLE With a Lytro-style camera, one exposure stores the full light field; afterward you shear and sum to focus on a near flower or, from the identical data, reshear to bring the distant background into focus, with no second shot required.