Information Technology Reference
In-Depth Information
2 Related Work
The main task of a transcoder is to convert from a source format to another one; this
task must be executed as efficiently as possible though. Consequently, transcoding
techniques focus on improving the second stage by using information gathered during
the first one. The key issues to manage are the time reduction and the quality - bitrate
penalty. Transcoding algorithms between traditional video coding standards are more
suitable to be used because both formats keep many features whereas DVC is more
different.
Firstly, in 2005 [6] WZ coding was proposed as a candidate in a transcoding scene,
however it only introduced the idea and the benefits of this new transcoding paradigm to
support low cost video communications but it did not offer a practical implementation.
The first WZ transcoder architecture was presented in 2008 by Peixoto et al. in [3]. In
this approach the authors designed a WZ to H.263 transcoder to support mobile
communications. This transcoder makes a mapping between WZ and H.263 GOPs,
including some Motion Estimation (ME) refinement for P and B slices.
In our previous work, we proposed the first transcoding architecture between WZ
and H.264 [4] available in the literature. This work improves the H.264 ME using the
Motion Vectors (MV) calculated in the Side Information (SI) to reduce the H.264
searching area with negligible RD impact. Nevertheless, in the previous approach we
employed a WZ GOP size of 2 to be transcoded to a H.264 GOP size of 2 which
means transcoding from K - WZ - K DVC pattern to I - P - I H.264 pattern. This
solution is not very useful in a real implementation due to the high bitrate generated
because one of every two frames is an I-frame. Moreover the refinement technique is
improved for P frames and generalized for longer H.264 patterns. Other
improvements of this approach with respect to the previous one is related to the DVC
implementation, as, in present work, it is based on VISNET-II project [7] which is
more realistic than the architecture used in [4] because it implements lossy key frame
coding, on-line correlation noisy modeling and do not use the ideal procedure call at
decoder for the stopping criterion. In other words, this work extends the approach
presented in [4] to a more realistic GOP size and format implementation. Moreover,
this work is a generalization to support whatever GOP size or format incoming from
DVC stage to be transcoded to I11P GOP pattern using a low complexity algorithm.
3 Proposed Video Transcoder
In a real scenario, video transcoding should be able to convert efficiently different
patterns. For this reason, the main aim of the proposal is to provide an architecture
which supports practical GOP patterns making efficiently the transcoding process
through exploiting the information that the WZ decoding algorithm provide in order
to reduce the H.264 encoding time on the ME process.
Mobile-to-mobile communications need to execute low complexity algorithms at
both ends. In the proposed architecture the source employs the DVC encoders and the
destination employs the H.264 decoder, so terminal devices support the lower
complexity parts in both paradigms (as Figure 1 shows). On the other hand, the
Search WWH ::
Custom Search