by Stefan Slivinski, Video Team Manager, LifeSize
Following up on my introductory blog post on SVC, this article will cover temporal scalability, which as I mentioned is one of the two components of SVC (we’ll cover spatial scalability next time). Temporal scalability provides the ability to have multiple frame rates for the same video stream. This isn’t a new feature of SVC, since it is possible to do this with AVC, but SVC simplifies how it’s done and makes it more obvious that there are multiple frame rates. Let’s see what this looks like:
In this figure the colored rectangles represent a single frame from a video sequence (ignore the different colors for the moment). Let’s assume each frame represents 1/60th of a second, and therefore the frame rate of the video sequence is 60 frames per second (fps). The arrows show the dependencies between two frames. So for example in order to decode frame 4, frame 0 needs to be decoded also; therefore frame 0 is a dependency of frame 4.
Now consider frame 1: No one is dependent on it, therefore it could be skipped and all remaining frames could be decoded. You could then also skip all subsequent blue frames, which means you would then only decode half the frames, thus cutting the frame rate by a factor of 2 from 60fps to 30fps.
Now that the blue frames are gone you could skip the orange frames since no one is dependent on them either. This would again cut the frame rate in half resulting in 15fps. You could then skip the gray frames and be left with 7.5fps. This would mean that depending on what frames you chose to decode, this one stream could have either 60fps, 30fps, 15fps or 7.5fps. That’s one video stream with four different frame rates, and that in essence is temporal scalability.
Next time on the “SVC Demystified” series, we’ll define spatial scalability.
Stefan Slivinski is the Manager of the Video Team at LifeSize, a division of Logitech. His team designs and develops all of the video algorithms that go into LifeSize’s current and next generation video communications equipment. His responsibilities over his ten-year career have spanned design and development of embedded video compression algorithms. Prior to joining LifeSize in 2005, he was at UB Video, developing video compression codecs for many of the major video conferencing OEM providers.