xanadu

[sketchyj]

i was watching the early 80s disco fantasy musical Xanadu today, the one starring olivia newton john...this movie has some really cool effects, a goldmine of clips for those who don't care about ripping.
anyway, i really liked how they were able to create these glowing neon borders around the dancers. i would really like to duplicate this effect on my own. I'm guessing it would involve chroma-keying, but if anyone who has seen this knows what im talking about, could you break it down for me?

[holly]

It's basicly an edge stroke or a glow with no ramp/fade. You can do it pretty easily in AE, but yeah, you'll need an alpha so it's easiest with chromakey. I'll try to dig up some old clips I made that use this...

[seex]

I dont know if its the same thing, but try those old after effects plugins, there is one called neon edges, i use it a lot with my dancers. You have control over the colour and amount of edge, you cant animate the parametrs tough. Drop me a mail.

[sketchyj]

cool thanks guys, that seems pretty easy

[sketchyj]

oh, one other thing, a lot of the shots are obviously not shot against a blue screen, yet they were able to create this effect. is there a way to make a mask around a figure that will recognize the colors it's masking and move with the figure? i should probably buy an AE book, it would probably open up a lot more knowledge to me...

[dongbamage]

in that case it was probably drawn in frame by frame (i haven't seen the clip but it sounds as though it would be beyond 80's tech to automatically do this, in fact its probably beyond todays wthout using chroma) might be talking bollix tho...

[sketchyj]

yeah i didnt think it was something easy...that would be a great tool though to be able to pull something out of a video like that...

[dongbamage]

yeah but that would just be autovj

[sketchyj]

nah, that would be a great time saver...and if it was that easy everyones content would improve

[dongbamage]

but how would you define what you wanted to include? unless the computer has some way of being able to differentiate between the focal object and the background it can not do it.

If it does however (ie Chroma, luma) then it would be ok, but that normally means a clip made for this purpose rather than just anything...

[sketchyj]

well, im not a programmer, but i would guess that it would require a lot more processing power than what we have today, and some kind of program or algorithym that is able to identify objects based on logic and probability combined with elements of chroma key technology. I've read some articles about projects that for years have been trying to create AI for robots that can function in the real world, but it's not to the point that a computer can look at a scene and identify objects. A smart program would be one that knows that a brown lizard on a brown log is a separate object, not because of color, but because of shapes, movement, and a massive database of information that tells it what is what. That is what some of the projects i read about in wired have been trying to do for years, but that requires thinking that is on a par with the human brain, and programming to accompany it. Maybe there is some simpler way it could be done if it were designed just from the point of using it as a design tool, but i'm probably not qualified to guess how that could be done without BSing you.

[dongbamage]

yep unfortunately current technology is unable to recognise anything as abstract (from its little binary pov) as a lizard or a log...

it would be have to be a huge database and it would also require a lot of cpu power that should be spent on fx...

128 bit computers maybe...

[lowRes]

iep..

i'm also still waiting for direct interface between brain and computer...

ehheeh!

* )
r_x

[johnnylocust]

Originally posted by sketchyj
well, im not a programmer, but i would guess that it would require a lot more processing power than what we have today, and some kind of program or algorithym that is able to identify objects based on logic and probability

This is something that Mokey (http://www.imagineersystems.com/products/mokey/) can do.

If you're interested in the mechanics of CV (computer vision), intel has an open source library called OpenCV (http://sourceforge.net/projects/opencvlibrary/) . It's not for the faint of heart though. nVidia has a similar open source library called GPGPU (http://www.gpgpu.org/) that is usfull for CV applications.

Or... you can do like Disney and hire an animation sweat shop in Korea, and have an army of 12 year olds do it by hand.