Comments on: Anaglyphs With Pixel Bender & Depth Map

By: Booney

Booney — Thu, 25 Aug 2011 13:10:58 +0000

ALso there is a cool software which converts your 2d video files and plays them is snsglyph on the fly!
convert-to-3d.com

By: Jozef Chúťka

Jozef Chúťka — Tue, 30 Nov 2010 10:59:04 +0000

filippo check out also DAISY method “An Efficient Dense Descriptor Applied for Wide Baseline Stereo”
http://cvlab.epfl.ch/research/surface/daisy/

By: Jozef Chúťka

Jozef Chúťka — Tue, 23 Nov 2010 14:35:59 +0000

filippo, great source, now if you could find C sources and compile it into alchemy, you should get really close to some good results

By: Jozef Chúťka

Jozef Chúťka — Tue, 23 Nov 2010 14:33:48 +0000

Brmm, I never did, however you can. This effect works best with smooth depth changes, if there is something too far next to too close it may result choppy… use drawTriangles method from http://3ddm.yoz.sk/

By: filippo

filippo — Tue, 23 Nov 2010 13:49:42 +0000

Hi Joseph,

probably I have found a better documentation that can help to understand how to implement it. In this document it’s also great the idea of the deph map for augmented reality.

http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceeding/poster-session1.html (GPU Accelerated Realtime Stereo for Augmented Reality –> .pdf)

There are also on internet other examples using openCV. https://sites.google.com/site/elsamuko/c-cpp/opencv-depthmap

By: Brmm

Brmm — Tue, 23 Nov 2010 08:59:34 +0000

This is really nice,

Have you ever tried making multiple instances of the shader in one displayObject?

Like if you just need simple 3D, as an alternative to Papervision where you don’t need that much panning, like for top view flash games.

Or is that just impossible to due flash performance?

Great work.

By: Jozef Chúťka

Jozef Chúťka — Mon, 22 Nov 2010 22:01:18 +0000

filippo,
now I can see what you are up to… I am affraid you will not be able to use pixel bender while there are no loops in current version. I can see some algorithm public available
http://www.google.com/search?sourceid=chrome&ie=UTF-8&q=algorithm+two+image+depth+maps … I am briefly reading http://cvlab.epfl.ch/publications/publications/1993/Fua93.pdf , I guess that is the way to go, anyway I believe it would not be fast enough to run realtime (as you could see with photosculpt) … however I think this is an interesting idea. Let me know how else could I help you

By: filippo

filippo — Mon, 22 Nov 2010 16:38:24 +0000

Hi Jozef,

thanks for your prompt response.

A kind of Kinect system is what I would like to create. If I/we built a fast system in Pixel Bander to get a depth map it’s easy after to choose the objects in the environment we want to consider for the interaction. We can get an image with only the closest making transparent the others.

If you can help me obtaning a Pixel Bander solution using two images, taken from two cameras separated by a distance of a few centimeters, as input and getting a depth map as output I can make it in realtime.

By: Jozef Chúťka

Jozef Chúťka — Mon, 22 Nov 2010 15:12:32 +0000

Filippo, I do not know about any realtime AS3 attempts. Well maybe will be interested in http://kinecthacks.net/

By: Filippo

Filippo — Mon, 22 Nov 2010 09:23:45 +0000

Many thanks Jozef for sharing but I am looking for a realtime solution. My goal is not only for a stereoscopic 3D effect but I would like to find out the most distant objects and those closest.

http://www.pages.drexel.edu/~nk752/depthMapTut.html