Raising the Bar on 2D-to-3D Conversion

BitAnimate (Lake Oswego, OR) is a small start-up company developing 2D-to-3D conversion technology. They envision their technology being used in a variety of ways- from an on-line conversion service for users to upload their clips and watch them in 3D, to embedded application like 3DTVs. They also believe their technology is at such a level that they can approach Hollywood to dramatically reduce the time and cost of making theatrical-quality conversions. Other companies are already in this space, most notably JVC and DDD. After seeing a demo recently, it seems the main differentiating feature is that it actually works as advertised without as many artifacts. And, it even works in real-time.

I know, the purists will say that converted 3D content will never look as good as native 3D content. But, we have seen some horrendous ‘native’ content too. The result of either approach ultimately depends on the skill of the artists involved. The challenge with automated conversion is that the machine is making artistic decisions, not a person, and is doing so very rapidly in the case of real-time.

We were invited to BitAnimate’s offices for an exclusive demonstration of their technology prior to their planned demonstrations in a private suite at CES in January. They provided a side-by-side demonstration of their software with the built-in 2D-to-3D conversion on the Samsung 3DTV (based on the DDD software). The source material was Jurassic Park and Troy.

We tried to look at a lot of converted content while researching our 2010 Real-Time 2D-to-3D Conversion report and developed a flow-chart of how most conversion algorithms operate. There is a lot going on in a very short period of time; analyzing the scene to figure out what is in it, creating a depth map using visual and/or motion cue, constructing the 3D image — filling in as many missing detail as possible and finally, some post-processing to make sure the colors and gamma are the same for each eye. And you do this about 30 times per second — whew.

BitAnimate considers their conversion process to be a trade secret so they wouldn’t reveal too many details. When asked directly about their process, Behrooz Maleki, President of BitAnimate used an automotive analogy; "While you could describe two cars as having an 8 cylinder engine, two doors, a transmission and four tires - you don’t know if it is a Ferrari or a Pontiac." So, perhaps our flow chart is fairly accurate, but how they do each step is the secret sauce.

Maleki’s experience in video process goes back to his days at InFocus in the late 1990s. While there, he developed a chip to do deinterlacing and image process to compete with the market leading solution from Faroudja. "The Farouda chip was $230, but our solution was a $15 chip — and it looked better in a side-by-side demo," commented Maleki.

The same approach is being applied to 2D-to-3D conversion. Maleki says that not a lot of hardware is needed to implement his algorithms. The graphics cards on modern PCs and the video processing cores in TVs should be adequate, he says.

In addition, they are developing a new 3D web site that will potentially offer a 3D conversion service. Upload your 2D content and get back streaming 3D content. This could go to your iPhone or PC (using Silverlight). BitAnimate uses their own 3D player, which allows the user to choose the output format for the 3D platform they are viewing the content on.

It seems logical that if 2D-to-3D conversion can be done very well there should be a market for it. BitAnimate seems to be another example of a small company with a good technology having the potential to raise the bar on everyone.

By Dale Maunu, Display Daily