[Pixel_Outlaw] has been working on a method to capture 360 images with his camera. He’s using a shiny Christmas ball ornament to reflect the entire room into the lens of the camera. In the unwrapped image you can make out the three legs of his tripod. In that snapshot he laid the ornament on the floor and pointed the camera straight down from above.
What catches our attention is the post processing he used to unwrap the image. He loaded up The Gimp, an open source image manipulation program, and used just three steps to unwrap the image. First he cropped the picture so that it was square and the spherical ornament was perfectly centered. Then he ran the polar coordinates filter. Finally he scaled the image, setting the width to be Pi times the height. Works pretty darned well for something that doesn’t take much fiddling.
The ornament wasn’t perfectly smooth (or maybe it was a bit dirty) but you can get a much better starting image if you use a bulb with a silver reflector like we saw in this older hack.
From what I understand, this is exactly how the pros do environment mapping onto composited 3d objects. A silvered ball is placed where the object is to appear in the scene, then the resulting image used to create light from surrounding objects.
Can he do this with realtime video?
Would it be possible to reduce distortion by using a mirror pyramid instead?
The great part is that there is not a hemisphere on the photo, but actually more than 99% of the environment is reflected on a hemispherical reflecting surface.
Photoshop has the same polar filter. You loose some detail at the top, but it’s a very nice method of capturing the whole environment.
@Hirudinea: It’s very easy to make a GPU-shader that does polar coordinates. Perhaps something like matlab can use shaders on realtime video?
A friend worked on a similar project for video conferencing about 10 years ago. I don’t think it got as far as the market. I can’t remember if they used a hemisphere or it’s inverse, an upside down funnel shape.
I once saw one of these where the camera was held in place by a single post which raised a center-drilled lens filter to a hole drilled in the top of the mirror sphere. Wa-La, no tripod in the image.
This is quite brilliant…and free.
Thanks for the interest everyone.
I just had an idea to give people a cheap way to take 360 degree pictures. You could take this camping quite easily and “unwrap” the pictures at home on your computer! The idea is to put as much image data as possible into a single top down photograph.
Happy photography! Have fun.
@localroger – the word is spelt ‘voila’, it’s French. Just a friendly tip :)
Nice security camera idea. Though how you explain the contraption hanging off the ceiling is beyond me.
This has been well known in the photography world for a very long time and there are commercial cameras based on this which do the unwrapping in real time. Most 3D software has instructions for creating reflection or hdri lightmaps using this method in the manuals as well. It’s still cool but it wasn’t an original idea at all.
Hmmm… I’ll bet that ‘Killerchrome’ would make for a good reflective surface: http://www.alsacorp.com/products/killercans/kc/killerchrome.htm (Check out the videos at the bottom.)
Apple patented this ten minutes ago.
This guy is a frickin genius! What a unique idea and it works perfectly.
Well to be fair, I’d not known about HDRI photography until it was mentioned above . This was sort of created out of my own need. Still it is a very cheap solution that is just plain fun.
If you want good picture try a better viewing ball and camera. You just might be able to rival those expensive setups if you don’t mind a bulky rig and upping the project a few dollars. Take my initial idea, modify it, personalize it, and make it better.
Far out a use for those garden mirror ball decorations. I’m going to have to remember this when Christmas rolls around I have a tripod that should take photos to stitch up up to male a 360 photo. I never did try because I thought it would be too much a hassle to get level.
@by _gl.tch Actually it’s voilà :)
—-
As for people saying it’s well-known, well yes it is, in fact several tv shows and movies used it over the years it in an exaggerated way by zooming on an doorknob or ornament in a picture or some CCTV footage and then creating a flat image from the reflection, that then invariably is of incredible quality and detail that makes NASA people that work with 43MP cameras fall off their chair in awe.
Hmmm… If you use two, you can use this system for 3D computer vision: http://people.csail.mit.edu/koch/omnivis2010/papers/paper-2-OMNIVIS2010.pdf
hdrshop is free and will allow you to do HDRI as well as spherical unwrap. Yes this is exactly how environment maps are gathered for film effects work. There are several places to source better light probes, garden “gazing balls” will usually do the trick in a pinch.
Good work using gimp, it is always better to have a free, open-source alternative to something under license by an evil corporation like USC :)
In case you are interested, here is a link:
http://gl.ict.usc.edu/HDRShop/
@_gl.tch, whatnot – Actually it’s a large violin.
Greetz from BrmLab – Prague HackerSpace
Our robot is using Tescoma ladle to have some sense of space around :-)
You can see video of first try (or first fail if you need :-) to move beer keg from point A to point B:
http://www.youtube.com/watch?v=s43x2ZOyOBk
And some (also kinda funny) implementation details and debuging info:
http://brmlab.cz/project/brmbot_outdoor
http://lifehacker.com/378490/stitch-photos-into-panoramas-with-free-software
@doronbc this isn’t about panorama but about spherical mapping and one-shot 360° shots.
Furthermore having to make all those control points seems silly when point-and-shoot cameras can stitch without it, and when it’s quicker to load the damn pictures in some image editor and manually move them to line up than it is to set tens of precise controlpoints.
Still thanks for the link though, some people might benefit from it.
It just occurred to me that if you took a picture of the inside of a large cylindrical box with a fine checkerboard pattern printed on it, you could map the sub-distortions (e.g. small dents), and create a correction map.
@DeadlyDad That’s not an altogether bad idea, but you’d need the right software to apply such corrections though.
I think The Gimp has some displacement filter that uses a greymap to do alterations, although I’m not sure it would be up to this job.