inside panoramas / cylindrical textures from real objects

taking 360 degree panoramas from turning the camera to the middle, not outside.

here i have a small toy car which is standing in the middle of a slow moving turntable.
(1round / 30sec)

Step 1 The Object

i placed the car in the middle of a turntable which is turned slowly by a motor from a "disco mirror ball".
the video camera (mini-dv) captures the car as it turns.

with 25 Pictures/Sec and a turning-time from 30 sec i end with 750 pictures.
(copy the DV-film via firewire from the camera and splice him into single images with your favourite movie-tool)

Comments

[mirlen101]

I did this the easy way I just laid an object on my flatbed scanner and rotated the object as the light passed in sync . I even scanned my whole face , sides of my head included ;-) It was sooooo creepy ! Much worse than sitting on a copier ! ;-) A scanners stepper motor could be used to make the rotation automatic and precise . The scans are high def but limited distance . But I've done depths of about 4 inches . I've beat cameras with an old flatbed ! Here's some of the plane flat results ( sans 180 degree method ) . One image the broken squash got 25,000 hits in one week ! The album got 50,000 ! http://entertainment.webshots.com/album/562675198AQuMNr

[gfixler]

Your results came out just great. I got the idea to simulate a slit-scan camera about 5 years ago with a regular digital camera, by taking a video, and chopping out the middle column of pixels.

At first, my idea was to hold the camera up against my car window while making the long drive up and down Sepulveda - an LA road that goes through the mountains - that took me to work and back home. I had no graphical programming experience, so after dumping the ~3 min. videos out to frames, I wrote a Photoshop batch to open each one, copy the center column, close it, and resize the other image it would refocus back to from its center point outwards width-wise by 2 pixels. This effectively built the image to the left, and a blankness to the right. There was no way to paste to the far right edge of the window, so I had to trick it this way. Then I'd just crop out the right half. It worked well, and slowly.

I also tried making several versions of the same pano from different vertical columns, evenly spaced across the image sets. When played back, it made everything in the image appear to spin, and slide a bit, as each was being viewed from different angles when it was in different parts of the frame. It was fun, and the images were monsterously wide.

Then I felt like trying it for your idea here, but had no nice rotational device, so I taped down 2 bars of something to create a V on my desk, and pushed a dead CD into that corner, and put an action figure on it (Rokkon, from He-Man), and filmed from a tripod as I tried very hard to spin the disc slowly, and evenly. I got something kind of like your spins, but very low-res (320x240 video), and very jaggedy, as I just couldn't spin it smoothly enough. It was just a test though, and it worked enough to prove the point for me, so I didn't revisit it. I tend to give up when I see that an idea works, without making anything cool out of it - I guess I'm just all about process, and not results.

Somewhat relatedly, a few years after that, I got into microcontrollers, and processing (www.processing.org), and actually made a little turntable with a platform on a stepper motor, for perfect 1.8° incremental turning (200 steps in a full 360°). I hooked it up to a processing script through a BASIC Stamp connected via serial to the PC, and had it spin the thing, and capture a picture through a miniDV cam for each position. Then I wrote a Processing script to let me spin the object - Pom Pom, a Homestar Runner figurine. You can see the setup here:

http://www.garyfixler.com/pomspin1.jpg
http://www.garyfixler.com/pomspin2.jpg

And you can play with it here:

http://www.garyfixler.com/pomspin/

Click, and drag left and right to spin it, and if you make sure to let go inside the boundaries of the little movie, you can sort of throw him in either direction (let go while still dragging), and he'll spin, slowing to a stop.

[RichardBronosky]

eric1000: "take the middel row of pixels from each of the 750 pictures" I'm totally confused as to how this is done. However, if you were to have said middle column (as in vertical slice of pixels) I can see how this might be possible. Also, what do you do with the result? I didn't catch anywhere that you share how you use these. (I am a skimmer, not a reader though.) Have you ever printed one of these and attached it to a cylinder?

[eric1000]

i've done the printing... but for me its more a "art" thing. maybe some 3-D worker can use it to make more realistic textures to their models, but i like it as "art"... the next thing i plan with this is to make animated pictures with lenticular lenses. i like playing with dimensions (in/out, time/room, ...) and generating a new "view" to common things... :o)

[RichardBronosky]

I printed and rolled the image of the camera. It was quite funky looking due to the lens distortion. The rubber duck looked very good when printed and rolled however. I suspect that this is because it is of [nearly] consistent "distance from center" all the way around. (or you might say cylindrical) This gave me an idea. I'm going to try to take a series of still pictures of my head with my Canon XTi and crop+join them. I think that since a human head is relatively cylindrical, this should produce good results. To eliminate lens distortion, I will use a zoom lens, shoot from a great distance, and only the center 50% of the image area. This could take me months to pull off, but I will follow up.

[eric1000]

i would like to see your pictures... please send me one if its done... even if it looks not that good.. ;-)

[eric1000]

i use the free imagemagick to do this
http://www.imagemagick.org

sliceing the middle row of an 320x240 picture work like this:
convert -crop 1x240+160+0 in.jpg slice.jpg

bringing the slices together like this:
convert -append -rotate 90 slice-1.jpg slice-2.jpg slice-3.jpg out.jpg

[RichardBronosky]

Okay, I see the "-rotate 90" in there. That made me study your turn table image more closely. I see that your camera is rotated 90. So, it makes sense that you crop a 1 pixel high horizontal row from the video and rotate it 90 degrees to use it as a vertical column for the final image. As a Linux user, I could totally bash script mencoder/ffmpeg to pull stills from the video and imagemagick to crop and join. I just could not figure out the logistics of which pixels go where, based on your original description.

[eric1000]

ok.. i thought it would come out clearly...

1st about the "-rotate 90"...
just try it without and you'll see.. ;-)

the "append" feature appends the images vertically...

so two 1x240 slices would end in an 1x480 picture.
since i need a 2x240 picture i have to rotate the slices before appending them.

2nd about scripting...
yes, that was what i was doing and i'm planing a little .exe with graphical interface for "simpler usage" in near future...

3rd about the stacking mehtod...
imagine this movieclip as source:
image 1 = 1a,1b,1c,1d,1e,1f,1g,1h,1i
image 2 = 2a,2b,2c,2d,2e,2f,2g,2h,2i
image 3 = 3a,3b,3c,3d,3e,3f,3g,3h,3i

this would lead to a picture (just the middle row of each):
1e,2e,3e,...

clearer now?

[5oldier]

I'm really into panoramics at the moment and I was wondering about doing something like this, but how do you select the exact middle row of pixels?

[Shark500]

deffinatly the sickest thing i'vwe seen today. completly awesome. great job

[royalestel]

I DEFINITELY think you should add the nuts and bolts for your image processing method. Mine was to crop the images and then offset each frame by 1 pixel, add all the frames together and voila! But the tools you used to manipulate the images are an integral part of a clear instructable. Cheers!

[royalestel]

Hey great! I think you should rename this as "cylindrical textures from real objects". This technique is used to create textures for 3D object doubles. My method used a shake script, but shake costs $500! I'm glad to see a cheaper alternative.

[eric1000]

the "cheaper" alternative mostly this: for PICTURE in `la -A1 IN-*.jpg do # cut a 1 pixel slice from each input frame convert -crop 1x480+320+0 $PICTURE slice-$PICTURE done for SLICE in `ls -A1 slice-*.jpg` do # put slices together convert -append $SLICE finished.jpg done and that is worth $500.. hmmm ok... /me remember to price my windows port of that script higher... ;o)

[royalestel]

Yeah, that's pretty much what my script was, but I didn't know of another image editing program I could do the same thing with. . . Thanks for posting this! Really!

[eric1000]

http://www.imagemagick.org/

woth a view... and its free...
think of it as all possibilitys of photoshop and more without being forced to use a mouse. ;-)
scripting is fun.

you can do almost anything with it as you can with photoshop.

but its not so "easy" to use the first time, because you only have "files" and not "pictures" you edit..

[Taotaoba]

This seems to be a hard part too. Did you use some kinds of special software tools or skills?

[eric1000]

just "imagemagick" and some batch scripting skills.

[ninepound]

Any chance we can get that script?

[eric1000]

PM me, but i want the pictures you did in return! beside that.. its a very simple script, since i described it already here how i did it... the rest is in the manpage...

[Taotaoba]

Thanks. I'm new to panoramas. Are there some software can view your product? I mean, like looking at a rotating car?

[eric1000]

hmmm... the point is not looking at a rotaiting car... the point is looking at all sides of it at once... the "product" (if you name it so) is the flat picture... not an animation or reproduction of it in a common known view... its more ment as "art" or as a physical unrolling of a "round" thing to a "flat" thing. the fun ist the switch from inside equals outside, or around equals front...

[Taotaoba]

I see. I thought it might like this one Create Quicktime VR Panoramas. Yet it does look artistic. Nice job and thanks for sharing.

[Zak]

I've done this in the past with a video digitizer that scanned left to right. By setting someone on a swiveling chair this would produce a 'rolled-off' version of their face. Nowadays this would take a bit of processing video, taking a different vertical line from each following frame. Nice to see this done in a different way now!

[eric1000]

yes, thats the principle of a finish camera in sports... at rotating or other targets it will bring interesting new views... its worth trying... and since it is so easy to do this with digital cameras... so why not... :-) "tech" meets "art" :-) so this is listed at "tech" and "art" LOL happy you all like it! :o)

[beanblog]

Wow - Is there an easy way to automate the photoshop part? I don't really want to select the middle row from 750 individual pictures. I assume there's some batching or macro's involved - ?

[eric1000]

the first ones (some years ago) i did with corel photo paint scripting... but these where done with a linux "imagemagick" script. but... there is a "easy" windows port in planing too... ;-)

[Dorkus1218]

I'm curious how you got that pulsing effect in the last pic of the tape. Did you take a different slice from the original pictures for each of the "frames" of the animated gif?

[eric1000]

yes... just an animation of different slices from the source pictures.

[Dorkus1218]

Wow, very very cool.

[gmjhowe]

well, you could do that in paint, if you can use paint, if not then photoshop/coral draw/ etc

[CementTruck]

Ha! Step 5, Image 4 - looks like a stapler stolen from the Batcave, and image 6 looks like a utility belt.