How it works
To give you more dimension on 3D, here’s some background how the conversion technology works at YouTube. Since last September we’ve been constantly improving the underlying technology, which now uses several techniques:
- We use a combination of video characteristics such as color, spatial layout and motion to estimate a depth map for each frame of a monoscopic video sequence
- We use machine learning from the growing number of true 3D videos on YouTube to learn video depth characteristics and apply them in depth estimation
- The generated depth map and the original monoscopic frame create a stereo 3D left-right pair, that a stereo display system needs to display a video as 3D
With this broader knowledge of 3D conversion, we then apply cloud computing scalability to make conversion possible across even more videos on YouTube. Breaking up a video into tiny chunks of data and processing them in parallel on Google’s cloud infrastructure lets us process these videos, while still producing the quality you expect.
We’d love to hear your feedback and other 3D features you’d like to see. With 4D, 5D and 6D around the corner there’s lots more we can do!
Deb Mukherjee, technical staff, and Chen Wu, software engineer, recently watched "YouTube Rewind 2011" in 3D!