Cryptography Reference
In-Depth Information
mismatch between the reference frames at encoder and decoder. Thus, even
if the prediction residues are correctly received by decoder, the output video
will still contain errors.
In summary, closed-loop prediction has reduced e
ciency and thus pro-
duces more prediction residues, which degrade coding e
ciency. On the con-
trary, open-loop prediction provides better e
ciency but creates mismatch
errors. Both prediction e
ciency and mismatch errors influence the quality
of the output video. Thus, the loop configuration is used as an encode option
for the trade-off between prediction e
ciency and mismatch error.
TheUpdateStep
In MCTF, the update step performs low-pass filtering along the motion tra-
jectory. It provides better coding e
ciency by removing both temporal noise
and aliasing. In addition, using the update step, the fluctuation of decoded
quality between even frames and odd frames can be reduced in open-loop
configuration [11].
However, the gain of coding e
ciency and subjective quality comes at the
cost of extra complexity. For the update step, the derivation of motion vector
needs intensive branch instructions. Moreover, the motion compensation for
adjacent high-pass frames demands extra memory bandwidth. In addition,
the buffer management for reference frames becomes more complex.
In [11], the role of the update step in MCTF is investigated. The studies
indicate that disabling the update step at decoder does not degrade the visual
quality and the PSNR performance of the decoded video while comparing with
the case of enabling the update step. Specifically, enabling the update step
at encoder reduces the fluctuation in quality between even and odd frames.
However, disabling the update step at decoder side does not show significant
differences in terms of PSNR or of subjective quality. As a result, the update
step is adopted as an encoder option in SVC. The MCTF at decoder becomes
a purely predictive structure and reduces the complexity of decoder by 50%.
Comparison of MCTF and traditional GOP structure
In SVC [3], MCTF is a key technology for temporal scalability. In addition,
it has been shown in [12] that MCTF also provides higher coding e
ciency
than the traditional GOP structure.
For a better understanding, Fig. 3.8 depicts the comparison of rate-
distortion performance with different GOP structures. As shown, the MCTF
with closed-loop configuration outperforms the traditional IBBPBBP
structure in all cases. Moreover, the closed-loop configuration provides better
rate-distortion performance than the open-loop one. In particular, with open-
loop configuration, the MCTF may have worse performance than traditional
IBBPBBP structure. An example is shown in Fig. 3.8(d).
