Easy 2D to 3D Conversion, example AVATAR (James Cameron)
Picture 1:
Using off-the-shelf software, this is a 2D to 3D conversion from one of the scenes of the AVATAR trailer. Note the roundness and depth of the male doctor’s body.
These pictures need to be seen on a screen at least 2 meters wide to get a true sense of the depth. As depth budget has been adjusted for such a screen size. You need a pair of red-cyan Anaglyph glasses to view.
There are some “tricks” that can be used when doing 2D to 3D conversions.
many times there is enough stereoscopic information built into 2D movies, that can help in more realistic conversions.
BUT, We at RealVision still say that 2D to 3D is not as good as creating in stereoscopic 3D in the first place, with stereoscopic cameras ( CG or Physical cameras)
The approach used:
1) Analyze a 2D movie for some stereo information – How? look for scenes with horizontal camera panning Or. if the camera is still and the subjects are moving from left to right or vice versa in the scene
2) Grab a stereo pair (using the principles of Pulfrich or time parallax from step 1)
3) Pass the stereo pair through a “depth from Stereo” algorithm
4) Smooth out the resulting depth map
5) Mask / Trace / Track main objects and apply the depth map to these
6) Start the tracking software, and manually clean up bad tracks, masks
7) Rinse and repeat step 6 to perfection
Granted the above is a very simplified way of doing 2D to 3D conversions, but it is just ‘one’ of the methods at the disposal of Conversion houses.
The below screen grabs I have posted did not yet do steps 3 to 7 and yet you will see how good the stereo is from the upcoming film AVATAR.
Note:The actual film itself when released will be a proper Stereoscopic 3D film not a conversion, so the 3 Dimensional effect will be a whole lot better that
the 2D conversions I’ve posted below.
Pulfrich 3D or Time Multiplexed 3D
In this scene, the Helicopter exhibits Pseudo Stereo (wrong depth). This scene would need separate tracking of the helicopter with a roto-mask.
However note the detailing that is automatically converted in the clouds and the vegetation in the scene, because depth information was extracted using Time Parallax.
As long as there is perfect horizontal panning AND NO VERTICAL movement in a scene, this is a perfect way to convert 2d scenes while retaining excellent depth information
Picture 3:
This is a stereo pair extracted from a 2D scene in the movie. This stereo pair could then be passed through a “Depth from Stereo” algorithm to create a Grayscale depth map or Disparity map.
Depth from stereo algorithms are still the domain of University research and robotics, but there are a few commercial adaptations out there already. One such is (was?) the PHILIPS 3D Solutions’ BLUE BOX content creation hardware/software system. This system also allowed for manual and semi automated creation and cleanup of depth maps.
After a depth map is created from such a stereo pair, it can be successfully applied in a scene, if there is not much subject movement. Conversions from 2D to 3D employ a whole different set of tools and methods depending on the structure and nature of a scene.
View this picture as cross eyed” stereo.
IMPORTANT: NONE OF THESE CONVERSIONS HAVE GONE THROUGH VERTICAL DISPARITY CORRECTION. THIS IS JUST A ROUGH GUIDE TO CONVERSIONS.
Picture 4:
Depth Budget was adjusted for this 2 meter screen. (Photo taken with a cell phone camera). On this screen the depth and out of screen effect on Picture 1 of the Aliens toes is very good
NOTE: All pictures grabbed from the trailer of Avatar (2D version)
Note: Now read why we are actually AGAINST 2D to 3D conversions for a full movie. (Using 2D to 3D as VFX when needed, and to clean up errors or when used on hard to shoot shots such as Stunt scenes or time critical events live… is Ok!)
All copyrights belong to the Film’s Producers and Owners. These Screen grabs are being used for illustrative and educational purposes only
