Interesting News Story on Robotic Panorama Camera Platform

Interesting. In a nut shell - "The Gigapan robot platform shown with a mounted digital camera was developed by Carnegie Mellon University and the NASA Ames Intelligent Robot Group. The robot takes up to 300 photos in 10 to 30 minutes. Software then "quilts" them together, resulting in a panoramic scene that can be zoomed for amazing detail".

He uses a Canon Powershot. Can't make out the model.
 
The result is a grand, scenic shot that also allows exploration deep into the photograph to reveal startling detail.
Click to expand...

Where's this detail coming from? It's already there, isn't it, whether it's a panoramic or not? A P&S has huge Dof anyway, so how is the GigaPan adding in increasing this startling detail? It all sounds like marketing hype to me. That's not to say this is a hugely useful tool to create panoramics in the first place.
 
Think a series of shots at full zoom, but with the FOV of a wide angle lens when all the shots are combined.

Think of how it would be with a 1DSmk2 and a good 400mm, with a few hundred photos. A several shot stitch from one of those already gives more resolution than scanned MF film, and an automated system to capture panoramic photos in less time is great.
 
RML said:
Where's this detail coming from? It's already there, isn't it, whether it's a panoramic or not? A P&S has huge Dof anyway, so how is the GigaPan adding in increasing this startling detail?
Click to expand...

I think it's all in the software; probably some kind of extrapolation algorithm akin to a digital zoom. I'm pretty sure it's the software that makes Gigapan special because the platform itself looks fairly basic and the camera shown in the article looks like a discontinued Powershot SD-series compact.

Gigapan looks like a lot of fun, though! I would love to try it. It makes my CV 15mm Heliar look like a telephoto.
 
No, it's not like digital zoom. If you tack 6 2000x3000 images together, you'll have a 6000x6000 image, and if it's perspective corrected and stuff it'll look like one exposure, so it has a lot more resolution. If you use longer focal lengths to make the individual images, they'll act like a shorter focal length once it's together, and that's where the detail when zooming in comes from.
 
MadMan2k said:
No, it's not like digital zoom. If you tack 6 2000x3000 images together, you'll have a 6000x6000 image, and if it's perspective corrected and stuff it'll look like one exposure, so it has a lot more resolution. If you use longer focal lengths to make the individual images, they'll act like a shorter focal length once it's together, and that's where the detail when zooming in comes from.
Click to expand...

I can't follow your reasoning. 6 times 2000x3000 makes a single 6000x6000? How? IMO it'll still just be 6 2000x3000 stacked on top of eachother and resolution won't increase. If you stitch them together than you'll get a 12000x3000 image but still the resolution won't increase; just the canvas size increased.
 
RML is right - the resolution is the same as the camera gives. Image size is bigger. No more details, but more subject are in the frame. That's what I see. The very same can be done with 1 shot of Panoramic camera
 
I get it. This is very clever. Note well that the device makes hundreds of photos, at varying optical zoom focal lengths. THAT is the key.

If it were merely a panoramic device, it would be no different than the typical nice level tripod and manual stitching software. You take maybe 3 or 4 shots, each one covering 70-80% new areas and overlapping the rest with the previous photo. Then you use the editing software to stitch together the image and you have a nice wide panorama view, but I'd agree with those who have said it would have no more absolute detail than any single photo would have.

But this is different. The camera is taking up to 300 photos - why? Because it is covering every angle at maximum optical zoom. Each photo covers a small percentage of the overall panorama, but it gives maximum detail, zoomed way in as it is. Then the camera moves a careful and precise amount, a fraction of a minute of angle, and repeats the process.

When this result is stitched together, you get not 3 or 4 photos laid side-by-side to make a panorama, but hundreds. You actually would need software to 'zoom it out' a lot to where you could see the entire thing on screen at one time - but it would have the advantage of being able to 'dive in' as Google Earth does, and the level of detail would be revealed as you dove in.

The Gigapixel Project camera works in a similar way, but it uses film and covers more area, then scans the film at a much higher rate than digital is currently capable of producing.

http://www.gigapxl.org/

This could essentially make up for the differences in a scanned large piece of film, making such 'Gigapixel' photos much easier and cheaper to obtain.

The device would have interest, I would imagine, to the manufacturing, civil engineering, and other professions, and it would be quite interesting to museums and such as well - not to mention the fun factor for hobbyists.

I wonder how the inventors control for such things as light level changes during the exposures?

Well anyway, it's neat, thanks for sharing it.
 
You must log in or register to reply here.
Back
Top Bottom