Information Technology Reference
In-Depth Information
Resolution layer N
RoI
-
↑
RoI
-
Resolution layer 1
↑
P slices
Reconstructed
thumbnail video
Fig. 2
The thumbnail video constitutes a base layer and is coded with H.264/AVC using
I, P and B pictures. The reconstructed base layer video frames are upsampled by a suitable
factor and used as prediction signal for encoding video corresponding to the higher resolution
layers. Higher resolution layers are coded using P slices.
With the above-mentioned coding scheme, the thumbnail is transmitted continu-
ously. As shown in Fig. 1, the video player can display it to aid navigation. More-
over, the thumbnail can be used for error concealment, in case parts of the RoI do not
arrive in time. Ideally, the video delivery system should react to the client's changing
RoI with as little latency as possible. The described coding scheme enables access
to a new region, with an arbitrary zoom factor, during any frame-interval instead of
having to wait for the end of a GOP or having to transmit extra slices from previous
frames. The coding scheme described above uses H.264/AVC building blocks, but
it is neither AVC-compliant nor SVC-compliant.
Compliance with State-of-the-Art Video Compression Standards.
Current
video compression standards provide tools like slices but no straightforward method
for spatial random access since their main focus has been compression efficiency of
full-frame video and resilience to losses. SVC supports both slices as well as spa-
tial resolution layers. Alas, SVC allows only single-loop decoding whereas upward
prediction from intercoded base-layer frames implies multiple-loop decoding, and
hence is not supported by the standard. If the base layer frame is intercoded, then
SVC allows predicting the motion-compensation residual at the higher-resolution
layer from the residual at the base layer. However, interframe prediction dependen-
cies across slices belonging to a high-resolution layer hamper spatial random access.
Note that the motion vectors (MVs) can be chosen such that they do not point outside
slice boundaries. Also note that instead of SVC, AVC can be employed separately
for the high-resolution layers and the MVs can be similarly restricted to eliminate
inter-slice dependencies. However, this is very similar to treating the slices as sep-
arate video sequences. An obvious drawback is the redundancy between the high-
resolution slices and the base layer. A second drawback is that after RoI change, a
newly joined slice can only be decoded starting from an intracoded frame. However,
