When we look at a lit object, our eyes detect the light bouncing back from that object. That light travels in all directions -- if the object is transparent, the light continues through it as well. If you were able to isolate a single point in space within this lit scene, you would discover rays of light crossing through, traveling in all directions.
A light field describes this phenomenon. You can think of a ray of light as something that has five dimensions. Three of those dimensions are the spatial dimensions we're all familiar with -- what roughly translates to height, length and depth, or the x, y and z axes. The other two dimensions refer to the flow of light along the ray.
The geometric distribution of light is what we call a plenoptic function. The word plenoptic comes from the Latin word plenus meaning full or complete. Optics refer to the behavior of light. To capture the complete light from a scene, we would need a plenoptic camera.
A literal plenoptic camera is just a thought exercise. Opaque physical objects block light -- they cause occlusion. To capture a light field, the camera -- and photographer -- would have to inhabit every perspective around a subject simultaneously without blocking light. We haven't figured out how to position a camera to capture every point of view around a subject or prevent occlusion. But the Lytro camera does simulate some plenoptic functions.
The device has a few things in common with your average digital camera. It has a lens and aperture through which light passes. It also has a sensor that detects light. But between the two is an array of microlenses. These lenses are much smaller than the Lytro's main lens -- a prototype plenoptic camera that was the predecessor to the Lytro had microlenses that were 280 times smaller than the main lens [source: Ng].
The lenses in the array direct light from the back of the main lens to the sensor. The sensor captures the light field between the Lytro's lens aperture and the sensor. The resolution of the photographs depends both upon the power of the sensor and the number of microlenses in the array.
Once you've captured an image, it's time to crunch some numbers.