06-22-2011, 06:13 PM
(This post was last modified: 06-22-2011, 07:28 PM by Brightcolours.)
[quote name='wim' timestamp='1308764617' post='9412']
It doesn't actually record different directions, to be very honest, at least not to a very strong degree. It does capture at different distances in the focal plane, likely by having a slightly 3D type design of sensor (not a lot of displacement needed at the sensor side). Which is also why it can do 3D, and also why it probably doesn't do extreme high quality yet.
It's an interesting concept, based on a development in the 1990s, if I am not mistaken. Canon also showed something similar, which also can record video, a bit over a year ago, I think.
BTW, here is more and an interview with the founder of Lytro:
[url="http://techcrunch.com/2011/06/21/lytro-launches-to-transform-photography-with-50m-in-venture-funds-tctv/"]TechCrunch - Lytro[/url]
Kind regards, Wim
[/quote]
The idea behind lightfield capture is to capture the direction of the light. Not focal planes.
This is the reason why one can look behind subjects, somewhat, and can vary DOF and focal plane during the visualization of the data.
How exactly they capture the direction vector (together with intensity and colour) is not disclosed yet. Nor how high (or low) the data density will be.
*edit:
It is explained how the directional vectors are gathered, in this thesis:
http://www.lytro.com/renng-thesis.pdf
Basically, if I understand correctly, a "conventional" sensor is used to capture many tiny images of the same scene, each image "fed" by its own micro lens.
This gives many images with views from slightly different viewpoints, with with the lightfield is being calculated.
The resulting lightfield is then used to compute the different focal plane projections when viewing the image.
It doesn't actually record different directions, to be very honest, at least not to a very strong degree. It does capture at different distances in the focal plane, likely by having a slightly 3D type design of sensor (not a lot of displacement needed at the sensor side). Which is also why it can do 3D, and also why it probably doesn't do extreme high quality yet.
It's an interesting concept, based on a development in the 1990s, if I am not mistaken. Canon also showed something similar, which also can record video, a bit over a year ago, I think.
BTW, here is more and an interview with the founder of Lytro:
[url="http://techcrunch.com/2011/06/21/lytro-launches-to-transform-photography-with-50m-in-venture-funds-tctv/"]TechCrunch - Lytro[/url]
Kind regards, Wim
[/quote]
The idea behind lightfield capture is to capture the direction of the light. Not focal planes.
This is the reason why one can look behind subjects, somewhat, and can vary DOF and focal plane during the visualization of the data.
How exactly they capture the direction vector (together with intensity and colour) is not disclosed yet. Nor how high (or low) the data density will be.
*edit:
It is explained how the directional vectors are gathered, in this thesis:
http://www.lytro.com/renng-thesis.pdf
Basically, if I understand correctly, a "conventional" sensor is used to capture many tiny images of the same scene, each image "fed" by its own micro lens.
This gives many images with views from slightly different viewpoints, with with the lightfield is being calculated.
The resulting lightfield is then used to compute the different focal plane projections when viewing the image.