Kinect Produced Autostereograms (Magic Eye Pictures)

[Kyle McDonald], working collaboratively with [Golan Levin] at the Studio for Creative Inquiry, has come up with an application that can produce autostereograms. These are pictures that appear to be three-dimensional thanks to a visual illusion created by forcing your eyes to adjust focus and vergence differently than they normally would. The program is called ofxAutostereogram and it comes with a couple of examples. Both are show in [Kyle’s] video (embedded after the break), starting with a depth image of a shark. This combines with a texture tile, then is processed through the openFrameworks software in the package to produce the final image.

If that’s all it did you might find it rather unimpressive… these images have been around for some time although they were never so easy to produce on your home computer. But the second example is a pretty fantastic one. You can use a depth image from a Kinect as the starting point. As seen above, there is a preview window where you can adjust the clipping planes in order to include the correct depth. This also allows you a preview of your pose. Once it’s just right, snap a pick and process it through the software.

[vimeo http://vimeo.com/21666082 w=470]

[Thanks AnarchyAngel via Kinect Hacks]

8 thoughts on “Kinect Produced Autostereograms (Magic Eye Pictures)

  1. @pRtkL xLr8r watch the video, it was – but as he mentions they way magic eyes works is not ideal for this, as I watched it yes I could see him moving in 3D, but the tiling method used for these caused a very noticeable repetition throughout the image overlapping sections that shouldn’t be
    When he turned off the “live update” the repetitions melted away because of no more motion (they were still there, the eye just cant discern them as they don’t move anymore)

  2. The problem with live update is just that the temporal correlation makes it seem like the tiles are sliding sideways. He could fix it by using a randomly generated, temporally uncorrelated, pattern.

    Also, there are some depth artifacts which could be simply fixed. If you want the current pixel to be a repeat of the pixel n pixels to the left, you should not do row[index] = row[index – n], you should do row[index] = pattern[(index – n) % pattern_length]

Leave a Reply to lwr20Cancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.