by Stefan Slivinski, Video Team Manager, LifeSize
In my last article I talked about one of the two components of SVC: temporal scalability. This article takes us to the second component, spatial scalability.
To start, let’s think about the way video compression works. Traditional video compression exploits the fact that very little will change between two consecutive frames in order to reduce the amount of information necessary to represent that video sequence. The same holds true for two frames from the same instant which are at slightly different resolutions.
Spatial scalability uses information from different layers in order to reduce the overall size, so that the combined size of different streams can be much smaller. This use of information between layers is referred to as interlayer prediction and forms the core of SVC. Figure 1 shows a diagram of the interlayer dependencies of spatial scalability.
So, spatial scalability is the ability to have two or more resolutions of the same video sequence within the same stream. This can be achieved by using a container format to combine video streams. Then the overall size of the container will be equal to the sum of the streams within it (i.e. stream 1 + stream 2 + up to stream n).
The difference is that SVC uses identical streams that happen to be at different resolutions.
The three images in Figure 1 represent three single video frames from the same point in time. The arrows represent a dependency between the layers, just like with temporal scalability.
Looking from bottom to top, you can see that in order to decode layer 1, information from layer 0 is needed, and in order to decode layer 2, information is also needed from layer 1. The lowest resolution layer is often known as layer 0 or the base layer and is required by SVC to be fully interoperable with AVC. This means it cannot use any components of SVC – ensuring that decoders capable of decoding only AVC could also decode the base layer of any SVC stream.
In my next blog I’ll talk about how the two components work together – combining temporal and spatial scalability.