Page 97 - ITU Journal, ICT Discoveries, Volume 3, No. 1, June 2020 Special issue: The future of video and immersive media

ITU Journal: ICT Discoveries, Vol. 3(1), June 2020

leading to a higher partitioning flexibility. In and AV1 support a cross-component prediction
addition, 16 wedge prediction types, similar to mode. In this mode, the chroma samples are
geometric PUs of VVC, are available. predicted from co-located reconstructed luma
samples.
One specific feature of VVC is the support in intra
slices of separate luma and chroma partitioning, The reference and prediction filtering processes
which allows co-located luma and chroma Coding used in HEVC are also more elaborated in VVC and
Tree Blocks to be independtly partitioned. AV1. In particular VVC introduces a position-
dependent prediction combination, while AV1
defines a recursive filtering-based intra predictor.

2.3 Inter coding
Inter prediction has been significantly enhanced in
the recent video coding solutions by introducing the
affine motion model on top of the regular
translational model, along with more efficient
coding of motion vectors and higher motion
accuracy.
HEVC specifies three inter prediction types which
can be unidirectional (one motion vector) or
bidirectional (two motion vectors) using the
pictures available in the Decoded Pictures Buffer
(DPB): Advanced Motion Vector Prediction (AMVP),
Merge, and Skip. In AMVP, both motion information
(motion vectors, reference pictures) and the
prediction residual are signalled. In merge mode,
only the residual is signalled and motion
information is derived from a list of most probable
candidates. The Skip mode is similar to the Merge
mode for the motion, but no residual is transmitted.
Fig. 3 and Fig. 4 illustrate the increased number of
coding modes supported by VVC compared to HEVC,
as well as the impact of the new partitioning. The

same color code is used on both figures with
Fig. 2 – Partitions different color tints for VVC new inter-coding
modes: orange for AMVP, green for Merge and blue
2.2 Intra coding for Skip.
Intra prediction in HEVC is based on 33 angular VVC adds affine motion prediction to AMVP, Merge,
predictors plus planar and DC modes, all predicted and Skip modes. The regular Merge and Skip modes
from reference samples in the causal spatial are enhanced with MMVD (Merge with Motion
neighbourhood of the coded block (top line, left Vector Difference), ATMVP (Advanced Temporal
column). In VVC, 93 angular predictors, plus planar Motion Vector Prediction) and GPM (Geometric
and DC modes are specified. Furthermore, the Partition Merge). CIIP (Combined Inter Intra
predictors can be computed from an extended Prediction) is added to Merge mode. Moreover the
neighbourhood. New matrix-based prediction motion information can be refined at the decoder to
modes are also inserted for luma, leading to enhance the prediction per pixel at constant bit rate
30 additional modes. EVC is close to HEVC with of the motion information. It can be noticed on the
30 angular predictors, plus planar and DC modes. figures that the amount of Intra modes (red color)
AV1 specifies 56 angular predictors plus DC and for VVC has decreased compared to the HEVC case
4 additional modes. as a result of improved Inter coding (Intra has
higher bit rate).
For chroma samples, in addition to the conventional
directional, planar and DC prediction modes
commonly supported by the four standards, VVC

© International Telecommunication Union, 2020 75