Join 3,572 readers in helping fund MetaFilter (Hide)

Tags:

Panoramic Movie or Video from Super 8?
February 3, 2014 3:04 PM   Subscribe

I was impressed with this panoramic version of the Zapruder film. Does anyone know how this was done and if it could be done by a hobbyist on Super 8 home movies with consumer-level equipment (Mac)?
posted by entropicamericana to Media & Arts (5 answers total) 4 users marked this as a favorite
 
I guess I should note I know the theory about how it was done, I'm looking for specifics...
posted by entropicamericana at 3:05 PM on February 3


In a reddit thread about stabilized bigfoot footage, someone said you could do it in Adobe After Effects with a function called "Warp Stabilizer". I don't have After Effects, but at least it's a start for a google search.
posted by bluecore at 3:38 PM on February 3


According to Gizmodo, the uploader added computer generated frames between the originals to give that effect. The uploader also gives some of the history of the files used on the YouTube page. Given that it's only a few hundred source frames to deal with, I'm sure this was done using fairly standard tools on a home computer.
posted by planetesimal at 3:39 PM on February 3


I posted something like this in the Buster Keaton thread more specifically these A and B
Their way is probably fancier than mine but the basics may be the same.
1) get digital version of film
2) extract images ( lots of ways to do this ) i use avidemux
Hi lite section you want using A-B toggles then
File-> Save->Save selection as JPEG Images
3) that is usually way too many images so i have a script with this line that picks out every NUMth file;
cp $(ls *jpg | awk -v nm=$NUM '{nr++; if (nr % nm == 0) print $0}') ./$TDIRNAME/
( default is 10 ) and puts them in a sub directory TDIRNAME
4) You will have installed Hugin and some helper scripts
You can do this from the Hugin UI but i just run the following in the reduced images directory;
nohup match-n-shift -v 15 -a --output putanamehere.pto *.jpg
let it run and when finished open the putanamehere.pto file in Hugin and optimize and clean up any large discrepancies.
You can add control points between images as required to correct errors. Crop the images etc.
5) then for the different outputs you can create the large panorama image with the Panorama Outputs in Hugin or the individual layers under the Stitcher Tab.
6) i then just use a variation of ;
convert -delay 20 -scale 50% -deconstruct *tif -loop 0 animrdC-50.gif
to make the gif

You could probably do this all by hand in photoshop or similar but using the panorama stitcher saves a lot of time. I use this technique a lot to extract panoramas from old movies and its just a variation to generate the multi layers and make a gif.
The original sift algorithm did this automatically from videos this just feeds the code the stills instead.

( on preview ny answer is a bit more rambling than the others but it works )
posted by stuartmm at 3:44 PM on February 3 [5 favorites]


Small update for anyone stumbling across this. There are some small tweaks that you can do to speed all of this up.
I just saw that there was another method written up on Reddit using Hugin after i had posted this. It uses some different methods. Reddit Link with an imgur series of images for description.

I noticed the use of --linearmatch that removed the need to compare every image to every other one, this parameter compares the image to just the next in sequence. Perfect for video images .
The Reddit author uses the gui but some changes to the Hugin helper scripts let them be used without the gui overhead.

I copied the match-n-sift and ptoanchor scripts to lmatch-n-sift and lptoanchor respectively.
In lmatch-n-sift i changed the call to ptoanchor to call lptoanchor. In lptoanchor i changed the line 204 to include the linearmatch call;

$rule->Command ('$(CPFIND_SHELL)', '--linearmatch', '--celeste', '-o', '$(SIMPLE_A_SHELL)', '$(PROJECT_FILE_SHELL)');

You would then run;
nohup lmatch-n-shift -v 15 -a --output putanamehere.pto *.jpg instead of the line in the previous comment.

This is probably fast enough that you do not need to do the reduction in the number of frames but you will then have more to load and edit in Hugin when you get to that stage. So you could probably miss out step 3 in the comment above.

I wish i had found linearmatch sooner :)
posted by stuartmm at 9:14 AM on March 12


« Older I need a therapist in Toronto ...   |  I'm petite, younger, female an... Newer »

You are not logged in, either login or create an account to post comments