Information Technology Reference
In-Depth Information
software. Many proposals for Evidence collection of MPEG used the tools ex-
tracted from KTA. Therefore, it can be said that KTA at present is a benchmark
for the improvement of video coding efficiency. The following tools have been in-
corporated into the KTA software and have been found to yield gains in coding
efficiency.
(1)
Extension of bit depth
(2)
1/8-pel accuracy Motion Compensated Prediction
(3)
Prediction direction adaptive transform
(4)
Adaptive prediction error coding
(5)
Adaptive quantization matrix
(6)
Adaptive quantization rounding
(7)
RD optimal quantization
(8)
Adaptive interpolation filter
(9)
High precision interpolation filter
(10)
Adaptive loop filter
The block diagram of encoder with these tools is shown in Fig. 55.
Adaptive quantizing matrix
Adaptive quantizing rounding
RD optimal quantization
Prediction direction adaptation transform
Adaptation prediction error coding
Extension of bit
length
㧙
Transform / Quantize
Inv.Quantize
/ Inv.Transform
Adaptive
interpolation
filter
High
precision
interpolation
filter
1/8 pel
accuracy MC
㧗
Intra
Prediction
Entropy coding
Motion
compensation
Loop filter
Adaptive loop filter
(Block based / Quad-tree based)
Reference frame
memory
Motion
prediction
Fig. 55
Improvement points of AVC/H.264
6.3.1.1 Improvement of Coding Performance of KTA
Some new coding tools that have been incorporated into the KTA software that
provide improvements in video coding efficiency for motion compensated predic-
tion, transform, loop filter, and Intra prediction, are reviewed below.
(1) Motion Compensated Prediction
When generating the motion compensation signal of sub-pixel accuracy, a
suitable filter can be chosen among several candidates and the interpolation
