A normal photo only shows the scene from one angle, which
of course gives a flat image. When you're really there, you see
with two eyes, in "stereo". Surprisingly, it's pretty easy to take two images, and see them in 3D.
You can take 3D stereo pictures with any camera:
Hold the camera level, and take a picture. This is the view from the left eye.
Move to the right a few inches and take another picture, facing the same direction as the first one.
The
further you move the camera, the more of a 3D effect you'll
see, but too far apart and your brain can't fuse the images--just like
looking crosseyed at the tip of your nose. For extreme closeup
images, like flowers, you should move the camera only a centimeter or
so between shots. For really far away stuff, like mountains, you
can actually move the camera by dozens of meters! Check out some of our stereo pairs to see what they look like.
For
objects that move, like people, the separate pictures from the left and
right eye positions aren't quite at the same time, and so the two
images won't
line up. But it's fairly easy to build a little mirror rig to
simultaniously take the left and right eye images--here's a little
mirror camera rig I built, and some
photos from it.
You can then build red-green "anaglyphs" using an image editor
program like Photoshop (the GIMP should work too, but at least in
version 2.2, it doesn't seem to!). Here's how:
Open the left eye picture. Copy the whole picture (Select -> All. Edit -> Copy.)
Open the right eye picture. Switch the "Channels"
box so you're
only editing the Red channel--you want Red highlighted, but nothing
else
highlighted (click "Channels", next to the "Layers" box, then click the
label "Red"). Paste the left eye picture into the red channel
(Edit -> Paste).
Now the red channel has the left eye picture, and the green and blue channels have the right eye picture.
Put on a pair of red-green "3D glasses", and look into the
picture! You may have to shift around the red channel to make it line
up with the green and blue (you can use the four-way-arrow move tool for this).
You can also buy a set of LCD shutterglasses (like this $40 pair from eDimensional),
that latch onto your VGA CRT monitor's vertical sync line. The
glasses can black out each eye independently, and with the right
drivers (which are tricky to set up!), your monitor flashes the proper
image to each eye, and you see in stereo! This sorta-kinda works
with some DLP projectors (sometimes it works perfectly, but the eyes
are flipped and the glasses need to be flipped upside down!
Sadly, folks with DVI or LCD monitors
can't use stereo glasses.
Like we did, you can go to the ARSCDiscovery Lab, deep inside UAF's library, to see an awesome headtracking active stereo setup.
Other Fun Photo Stuff
Mount a camera on top of a long pole. This gets you above
most of the annoying shrubs, trees, and signs that clutter up most
photos. PVC pipe is long, cheap, and works well, but it is pretty
floppy! The trickiest part is pushing the shutter on top of the
long pole--I've tried the following ways to push the camera shutter button remotely:
Set the camera to "continuous shutter" mode, and duct-tape the
shutter button down (simple, light, but takes several pictures per
second, which uses up lots of battery power).
Set up a little solenoid button pusher (complicated and heavy, but works)
Rig a long string to a button-pushing lever (not very reliable, and the string gets tangled)
Open the camera, solder on wires to the shutter button, and run
the wires outside the camera (works great, and can be made extremely
lightweight, but remember the shutter button usually presses halfway
down for focus, and opening the camera voids the warranty!). For
the experiment below, I ran the camera-shutter wires out to a little 8-pin PIC 12F675 microcontroller, running this little button-pusher program. This made the camera automatically take a picture every 2.5 seconds.
First, buy some big balloons--36 inch balloons were what we used, and they only cost a few bucks each.
Second, fill the balloons with helium, which you can buy at a
party store along with the big balloons (should be $10 or
so).
Now tie the camera on to the balloon, and let 'er go! Hmmm... Wait a second...
To pull the balloon and camera back
in, you'll need some sort of tether. Fishing line works pretty
well to hold the balloon down, and a fishing pole works really well to
let the line out and pull it back in. We used 200lb test Kevlar
fishing line.
Here are some photos we took using Neal Brown's balloons and fishing rod setup:
Start with a cheapo webcam that has
terrible-looking color--like where black objects look blue or
purple. These cameras are usually missing an infrared filter, so
the infrared light is screwing up the colors of stuff. If you
block out the visible light, the camera will then see only in infrared.
So block out all visible light using a piece of exposed,
developed color film (the black stuff at the end of your rolls of
photos). Color film has pigments which absorb each color, leaving
only infrared. You may need to overlap two or three layers to get
a good IR response.
Take normal pictures using the camera--now you're seeing in IR!
Or, if you have several thousand spare dollars, buy a much cooler thermal infrared camera,
which operates in the far-infrared regime. At that wavelength,
even objects at room temperature are hot enough to glow--and a human
being is white-hot!
Here are some near-infrared homemade pictures of trees
(which reflect strongly in infrared), asphalt shingles (which are dark
and absorb infrared, getting hot), colored marker pens (which don't
reflect much IR at all), and automobile glass (which strongly absorbs
infrared--one reason why cars parked in the sun get so hot!).
Using a flatbed scanner, scan your hand or face while sliding, wiggling, or rotating it as
the scan head passes by. Scanners have a weird non-pinhole
perspective, where things get darker and narrower, but do not change
vertical size, as they get further away.
Google Sketchup is a nice little 3D
modeler with a very friendly interface. You start by drawing a 2D
rectangle (or any closed flat shape), and then use the "Push/Pull" tool
to sweep the rectangle up into 3D. To see the beauty of Sketchup
in action, watch this little tutorial video. Cool, eh? Now download it and follow along with the tutorials or the manual.
To texture a building face, first choose the menu item "Tools"
-> "Materials". Add a new material by clicking the weird "+"
icon on the top right of the Materials toolbar. Drag your texture
image onto the "Texture filename" box. Click "OK" to make a new
material with that texture. Hit "Edit" to set the texture size to
a reasonable value--by default, textures are only one foot across, but
a building photo is usually at least a hundred feet across. Use the
paint bucket to apply your new texture to a building face.
Right-click on the new texture, and choose the menu item "Texture"
-> "Position". This brings up four pushpins. Drag the
red pushpin first (to set the texture origin), then the green pushpin
(to set the texture scale), and finally the blue pushpin (to set the
texture height). I don't recommend messing with the yellow
pushpin. Choose the arrow tool again. OK! You've got
a properly-aligned texture face!
One thing Sketchup doesn't do well at all is photo texturing on
building surfaces, when your photo needs some perspective
correction. Here's how I do perspective correction:
Find a good photo of the building face you want. Look for
one without interfering objects (like the cursed trees and bushes), and
with a reasonable angle.
Open the photo in the GIMP,
my favorite free image editor and Photoshop clone. I usually do
this by dragging the photo onto the GIMP tools, since I don't like
GIMP's "Open..." box.
Crop the photo with the X-acto blade tool. This speeds up the next step.
Choose
the perspective tool (looks like a trapezoid).
Make sure the perspective tool is in "Backward (corrective)"
mode. Drag around the corners of the image to match the
corners of the building wall you want to extract. You can zoom in
and out with the "+" and "-" keys for precise alignment. Press
"Transform", and the perspective tool will distort the selected region
to fill the screen.
You
might hit "Image" -> "Scale Image..." to set the image to be only a
few hundred pixels across. Huge images aren't usually needed for
texturing.
Here's an example of how to do this. Note the final
perspective-corrected image is now entirely suitable for use as a
SketchUp texture.