Graphics Reference
In-Depth Information
weight denominator (LWD) rounding factor may be used before adding the offset.
For the case of bi-prediction, the value to be clipped is calculated as follows (similar
rounding offset is used for uni-prediction case as well):
.P
L0
ŒxŒy
w
0
C
P
L1
ŒxŒy
w
1
C
.o
0
C
o
1
C
1/
LWD/
.LWD
C
1/
(5.10)
In H.264/AVC, weight and offset parameters are either derived by relative distances
between the current picture and the reference distances (implicit mode) or weight
and offset parameters are explicitly signaled (explicit mode) [
6
]. Unlike H.264/AVC,
HEVC only includes explicit mode as the coding efficiency provided by deriving the
weighted prediction parameters with implicit mode was considered negligible.
It should be noted that the weighted prediction process defined in HEVC version
1 was found to be not optimal for higher bit-depths as the offset parameter is
calculated at low precision. The next version of the standard will likely modify the
derivation of the offset parameter for higher bit-depths [
24
].
The determination of appropriate WP parameters in an encoder is outside the
scope of the HEVC standard. Several algorithms for estimating WP parameters have
been proposed in literature. Optimal solutions are obtained when the Illumination
Compensation weights, motion estimation and Rate Distortion Optimization (RDO)
are considered jointly. However, practical systems usually employ simplified tech-
niques, such as determining approximate weights by considering picture-to-picture
mean variation.
5.5
Summary and Conclusions
The inter-picture prediction part of the HEVC video coding standard is not
introducing a revolutionary whole new design. Moreover, it can be seen as a
steady improvement and generalization of all parts known from previous video
coding standards, e.g. H.264/AVC. The motion vector prediction was enhanced with
advanced motion vector prediction based on motion vector competition. An inter-
prediction block merging technique significantly simplified the block-wise motion
data signaling by inferring all motion data from already decoded blocks. When
it comes to interpolation of fractional reference picture samples, high precision
interpolation filter kernels with extended support, i.e. 7/8-tap filter kernels for luma
and 4-tap filter kernels for chroma, improve the filtering especially in the high
frequency range. Finally, the weighted prediction signaling was simplified by either
applying explicitly signaled weights for each motion compensated prediction or just
averaging two motion compensated predictions.
