There’s a reason that more cartoons are in 3D than live action content. That’s because animated images typically are much simpler than real-world objects, which means that they can be manipulated more easily and precisely. This is important when you are trying to adjust the differences between the left and right views of a scene.
The problem gets even worse when you are trying to create a second view from a single image to create a 3D image. You can extract lots of depth cues from a single image — and even more cues if you look at a series of moving images — but it’s difficult to create a natural-looking object in the foreground as the background moves behind it. A company called YUVsoft has developed new algorithms that address this problem, with striking results.
This is just a still image that I grabbed from the YUVsoft website, but I recommend that you follow the link above to see the original animated sequence. I added the yellow circles to point out the serious artifacts that can happen when you try to “move” a foreground object against a background when you have something that you can see through, like hair. See how the hair in the top circle has smeared; in the lower circle, a weird “window” with sharp edges has appeared. The new process from YUVsoft does a much better job of preserving the natural look of the original image.
Why is this important? As 3DTV gains traction in U.S. (and worldwide) living rooms, there will be a huge demand for additional content. And with technology like this matting from YUVsoft, content providers will be able to reach back into their back catalog of 2D content and remaster them for 3D distribution faster and at a lower cost than they can today. The toolkit for improving this conversion process is growing every day, and developments like this from YUVsoft show just how much progress we’re making.