Prediction Coding - Advanced Video Coding Systems - page 66

0, 2, 12 or 24, the corresponding prediction for the chroma prediction mode would be
1(DC), 4(horizontal), 3(vertical) or 2(bilinear). When the final selected chroma pre-
dictionmode is ModeChroma and the predictionmode from luma is PredChroma ,
if the PredChroma is not the above four modes or ModeChroma is less than
PredChroma , the chroma prediction mode ModeChroma will be coded directly
into the bitstream as intra _ chroma _ pr ed _ mode . Otherwise, if ModeChroma
is the same as PredChroma , intra _ chroma _ pr ed _ mode will be coded as 0; if
ModeChroma is larger than PredChroma , intra _ chroma _ pr ed _ mode will be
coded as ModeChroma -1.

4.3 Interprediction

4.3.1 Prediction Partition

Compared to the spatial intraprediction, interprediction focuses on exploiting the
temporal correlation between the consecutive pictures to reduce the temporal redun-
dancy. InAVS2, in addition to traditional P andB frame, F frame is defined as a special
P frame which can utilize more than one forward prediction hypothesis. More specif-
ically, P frame is the forward predicted frame, while B frame is a bi-predicted frame
which can ultilize forward, backward, bi-prediction and symmetric prediction. For
each CU in an interframe, as mentioned above, inter PU sizes can be selected from
any of the following sizes: 2N

× × × × × ×

2N, 2N N, N 2N, 2N nU, 2N nD, nL 2N,

nR 2N. Both skip/direct mode and normal motion estimation mode are allowed

for 2N

×

2N PU. However, only motion estimation mode is allowed for other PU
partitions.

×

4.3.2 Motion Vector Prediction and Motion Vector Coding

Motion vector prediction (MVP) plays an important role in interprediction, which
can reduce redundancy among motion vectors of neighboring blocks and thus save
large numbers of bits for coding motion vectors. In AVS2, four different prediction
methods are adopted, as tabulated in Table 4.4 . Each of them has its unique usage.
Spatial motion vector prediction is used for spatial derivation of Skip/Direct mode
in F frame and B frame. Temporal motion vector prediction and spatial-temporal
combined prediction are used for temporal derivation of Skip/Direct mode in B
frame and that in F frame. In other cases, median prediction is used.
With motion vector prediction, the motion vector is coded into the bitstream as
the MVP index and motion vector difference (MVD). The MVD, in quarter-pel
precision, may still have much redundancy. To further save reduce bits of coding
motion vectors, a progressive motion vector resolution adaptation method is adopted
in AVS2, called progressive motion vector resolution (PMVR) (Ma et al. 2013 ). In