Lytro Support/Feedback Forums & Feature Requests/Living Pictures
Making Lytro panoramas?
Dan Ingold
posted this on February 20, 2012 07:25 am
Has anyone yet considered the possibilities of panoramic photography using the Lytro? In principle, it seems possible to stitch multiple "living pictures" together to form large-format, refocusable pictures. This technique could be used to overcome the pixel limitations of the present sensor, and to produce large-scale prints whose focus and depth-of-field is determined in post processing. After reading Ng's dissertation, it also seems possible that light field technology might be able to exploit, rather than be troubled by, the parallax effect that occurs when panoramic pictures are composed without considering the lens nodal point. This could allow handheld panoramas that otherwise are difficult to stitch due to parallax distortion among foreground elements. This parallax effect provides further 3D cues, as well, which might also be exploitable. Although panoramas should be possible using manual panning techniques, even more awesome is the potential of using the Lytro with a GigaPan to produce giga-pixel, virtually-zoomable, refocusable images..
Comments
Hi Dan,
Although this is indeed possible with the technology, it's not something we've made any public plans about yet.
Thanks for the feedback.
As always, we're very much looking forward to the wide-open possibilities of light field photography!
While we may not be able to make panoramas that are themselves refocusable, the camera can produce individual shots that can be screen grabbed at the desired focus points and then composited. Hopefully this will not jinx the copyright on the software. Any thoughts
Hi Richard,
That is very resourceful thinking. However, since exporting to JPG is not our main purpose for this first Lytro Light Field Camera, I personally would not expect the final results to exceed those obtainable with another camera for this purpose.
However, if you are on a mountaintop with only your Lytro camera, then yes, this might be a good solution in a pinch.
The answer really boils down to whether the LFP files contain sufficient information to allow the lightfields in adjacent, overlapping images to be combined, such that the composited image-as-a-whole can be refocused. The realization of that possibility sooner rather than later may depend on whether Lytro will reveal (or license) the LFP format, so that others might create these types of capabilities, and more still unimagined, while Lytro concentrates on the underlying technology.
Hi Dan,
Thanks much for following up.
Consider this: when Louis-Jacques-Mandé Daguerre was messing around developing his first commercially successful method of capturing light onto a permanent record, he very likely wasn't considering things like motion pictures, high speed photography. lithography and all the fantastic things that capturing photons onto plates and films that have evolved from this process(think micro-circuitry). This is but the first iteration of what will assuredly be a game-changer for every application that relies on photography at its core. It is the infancy people, dare to dream big. Taking into account the visions of people like Wells and Verne, Da Vinci and many more, it seems that whatever we can dream we will eventually achieve. I look forward to the many new uses of this innovation.
I'm no scientist, but this camera is quite interesting. My BIG QUESTION, is, Could this tech be used for motion pictures, ie movies, film, etc.
One assumes that since the "light field" is 3D-friendly and perspective shift friendly, it should be possibly to stitch together light fields based upon the 3D structure of the images If there is some similar pattern of 3D structure in the images that can be stitched together once aligned in 3D... That is, it could match up 'points' and 'edges' and 'shapes' of some sort, and then stitch based on a 3D "best fit," rotating in 3D as necessary, filling on whatever info as necessary, etc.
Perhaps one can even do a reverse panorama by taking images of all sides of an object to create a rotatable, refocusable 'image' of the entire thing? Effectively a complete 3D model, that's also refocusable to whatever's in the background? So, one cold take a Cthulhu statue replica, position the Lytro at 3-5 places around it, and snap images at app the same distance, then in software have it match up the surfaces, points, curves in 3D and then construct the entire scene in 3D not unlike QuicktimeVR, which never really caught on, but was still pretty cool for a number of applications. Even cooler here, if it can be refocused....
Panaramic lense ?
http://i116.photobucket.com/albums/o26/squiresmurf/Product%20Ideas/...