Junkyard Jumbotron: join all your screens into one big one, no software needed | Hacker News Reader

Junkyard Jumbotron: join all your screens into one big one, no software needed

boingboing.net

55 pointsdanshapiro16 comments

Slightly OT, how do they do it on TV shops where they put a lot of TVs to form a big screen? How are the images split? Can I do it with any TV?

i think there are boxes that do it. the googleable you are looking for is "video wall"

quote:

RickBorovoy

  To make your own Junkyard Jumbotron, please use  
  http://jumbotron.media.mit.edu
  The jj.brownbag link above is an earlier version of the software. 
  And please post your feedback / results on the Facebook page at    
  http://www.facebook.com/pages/Junkyard-Jumbotron/207846822564634

This is similar to the digital billboard panels on the sidelines at sporting events. There are some augmented reality applications possible with this idea too.

Back in the seventies and eighties before very large TVs having 20 TVs daisy chained together was all the rage to have in your coke dealer mansion.

>This is similar to the digital billboard panels on the sidelines at sporting events.

Yes, those are called Jumbotrons.

https://secure.wikimedia.org/wikipedia/en/wiki/Jumbotron

No sir I am talking about the field level panels for advertising purposes:

http://en.cnledw.com/upload/2010/08/30/201008301153735939.jp...

Couldn't get two screens on the same computer to calibrate properly. One of them always showed up with the big broken sign. I had to use two different browsers for it to even create two patterns so maybe there's a bug in the calibration when the two screens are on the same computer.

This could be really interesting for people working with digital art installations (using old PCs and monitors/TVs).

A shame that its not likely to work with the wooden 'mirror' though: http://www.youtube.com/watch?v=HCSbk9JDwPY

combining cell-phone screens automatically? lame.

combining video projectors automatically? awesome.

i wonder if the code can deal with overlapping displays.

There's tons of research on this, including a yearly workshop called PROCAMs (PROjectors and CAMeras).

http://www.procams.org/

Building arbitrary arrays of projectors is now thoroughly understood, including projecting on surfaces with corners, bends, different textures, etc. Adding one or more cameras to the setup allows systems to self-calibrate and/or dynamically be changed. However, I don't know if there is some standard software to do all this.

Here's a description of some older work done at our lab on this: http://www.cs.columbia.edu/CAVE/projects/pr_any/

Former PROCAM Oral presenter (NYC 2006) here.

Mersive, http://www.mersive.com/, sells a solution with their proprietary software on a box connected to a camera communicating over the network to rendering machines connected to projectors.

I hacked together a prototype X server that worked seamlessly across several projectors (casually aligned/partially overlapping/etc) for them a few years ago.

I also wrote high-dynamic-range camera code for them so that a camera can easily measure the full dynamic range of even overlapping, bright projectors.

I haven't done any computer vision since, but those 2 things (particularly the former) still get me a lot of attention with recruiters and in cover letters.

Cool to see another PROCAMmer on here. I'm quite familiar with Shree's work, as I worked in color correction, too.

That is insanely awesome. I always wanted to do more graphics stuff.

As apu points out, this is really well understood now -- has been for over 10 years, in fact. I think the fact that projectors are still fairly expensive (though much cheaper than years ago) and most researchers aren't big participants in the open source community is why we don't have anything widespread.

At the University of Kentucky lab where I worked in 2002 (The Metaverse Lab), we had a 24-projector geometrically/photometrically-corrected only-casually-aligned system projecting on 2 walls and the floor with all 3 planes meeting at a corner. Additionally, the system was cordlessly head-tracked (Intersense) and the world view would dynamically change so that you could look around objects, look under tables, etc. It was amazingly cool.

I'm pretty sure I could write the code for the geometric calibration of a 4-6 projector (whatever a single, stationary camera could see most of) single-plane system in a day and then use the homographies (plane-to-plane mappings) to warp and display large images geometrically seamlessly on the multi-projector display. There would still be extremely apparent photometric seams where the projectors overlap, though. You can use some simple models to get rid of some of this, but REALLY making it look good in practice is hard, and requires some engineering.

As with most things, making it look 80% good is easy. It's that last 20% that requires years of engineering and a dedicated company like Mersive to make it happen. Their stuff looks good.

Oh, of course. It's just that sometimes making the 80% go from easy to trivial is as interesting as the last 20% :)

For sure -- this is the essence of learning new fields without being a master.

I'm pretty certain if you buy 2 projectors, a camera (get a half-way decent one, not just a cheap webcam), and read/study a few relevant papers, you'd have a calibrated overlapping setup that you could send images to and show reasonably correctly.

The great thing about this is that it's a lot like graphics programming -- once you get the taste of the positive feedback of small successes, you'll want to start playing around with it a lot.

No software needed _except a browser_ much less amazing than title suggests.

Reply on news.ycombinator.com