Why is that for every pixel on the sensor plane, it will only evaluate the radiance from the ray that passes through one particular microlens?
Is this a correct way of imagining this? If you had 4 pixels, instead of accumulating light from the entire lens into each pixel, you combine the 4 of them and have each one only capture light from a fourth of the lens.
I think that the entire lens aperture generates the little disks for each microlens (or mini-camera?) and then on that disk image each particular pixel corresponds to a specific ray pinpointed by one location on the aperture and one location on the sensor plane.
Very recently, a paper published in 2016 addresses a variety of benchmarking and evaluation methodologies for light fields. In other words, they've come up with a way to measure the quality of a light field, taking into account observations like robustness to noise, texture sensitivity, and foreground fattening. Check it out here: http://lightfield-analysis.net/benchmark/paper/lightfield_benchmark_accv_2016.pdf
Would a metaphor for this be like looking at how the light refracts on the lenses?