4 ms·
I work on photosynth, so allow me to respond (albeit with a bias). Photosynth was designed to work with the most common use case - pictures taken with cell pho
by jjcm 13y ago
I work on photosynth, so allow me to respond (albeit with a bias).
Photosynth was designed to work with the most common use case - pictures taken with cell phones at different, undetermined angles, with 10-20 seconds in between each shot.
Contrast that with the camera system used for THORSKAN - http://i.imgur.com/VE1qXwS.png http://i.imgur.com/VE1qXwS.png
That's 72 professional DSLRs set up to take pictures of the object simultaneously. I'm not discounting THORSKAN's accomplishments, these are some extremely beautiful demos. But they can't be created by your everyday user.
Eventually things like photosynth will get on this level, but the camera technology in phones will have to change. Luckily, this is already happening. Intel's CES announcements focused on stereo camera systems in laptops and portables, and MS has an investment in it with the kinect already. Having even two cameras shooting at the same time dramatically improves results.
- bane 13y agoHi thanks for the response. I can appreciate how having consistent equipment at known locations and angles taking images of a scene can help with scene reconstruction. But I'm not sure I follow on the need for very close temporal alignment. Is it because of the consistent lighting conditions? With all things being equal, would it matter if each imager used with THORSKAN took a picture 20 or 30 minutes apart?
- JoblessWonder 13y agoWhen you say "very close temporal alignment" do you meant the stereo cameras he was talking about? If so, (and I'm sure since he works on Photosynth he is infinitely more qualified to comment but this is the internet so I'm going to anyways) then the advantage of a stereo camera is that you can more accurately judge depth as long as you know where the two lenses are in relation to each other. Think of it like like those old-timey 3D images. [1] It appears as though Intel's version is going to be a single camera with a depth sensor like the Kinect.[2] [1]http://en.wikipedia.org/wiki/Stereoscopy http://en.wikipedia.org/wiki/Stereoscopy [2]http://www.display-central.com/free-news/display-daily/3d-cameras-coming-to-ultrabook-and-laptops-says-intel/ http://www.display-central.com/free-news/display-daily/3d-ca...
- bane 13y agoNot exactly, he made several mentions of the important of multiple shots happening within a very tight time frame (or at least that's what I was reading into the comment).
- jjcm 13y agoTemporal alignment matters because, well, things move. This isn't as much of a problem for vast concrete structures that are indoors like in the preview they have here, but in the outdoors things are shifting all the time. It's hard to reconstruct something in 3d if it's changed shape by the time you've gone to take a picture again.