View Single Post
Posts: 93 | Thanked: 52 times | Joined on Oct 2008 @ Victoria BC Canada
#22
Well, how about this approach:

http://talk.maemo.org/showthread.php?t=30387

Basically, rather than doing "in camera stitching" it would just grab video frames and auto-assemble them into a big picture. Getting that to run on hand-held devices would blow away any megapixel ratings. The resulting picture would be as big as you want, irrespective of any sensor limitation. Of course, it would only work on static scenes.

I was interested in the N810, but the process should be able to support multiple devices. Also, there's not too much difference between generating a flat verses 360 panorama.

There is software available that does stitch video frames to a single image, so it's just a matter of getting that to run under a Maemo device, getting the data from the video camera in the first place, and getting that data run through the app.

I've no idea if it's possible for something like an N810 to run this much data crunching, at least in a reasonable amount of time, but at least there's a known process

Quoting my original post here:

Originally Posted by fixerdave View Post
In doing some research, I've found: Videoorbits from: http://comparametric.sourceforge.net

Which basically does what I want. It takes a video (in a very specific frame by frame format) and runs it through a 4-stage process, ending with a single composite image.

Now, I have no idea how fast it runs, or if it can even run at all on something like an N810, but I'd like to find out. However, it's a little beyond my current abilities.

As I see it, here are the steps to get there:

1) compile videooribits on the N810 (I've not set up a sandbox nor compiled anything yet.)

2) get some app that can pull a video stream from the N810 camera.

3) convert that video stream to the format required by videoorbits.

4) run the video through the conversion process

...
 

The Following User Says Thank You to fixerdave For This Useful Post: