Information Technology Reference
In-Depth Information
coding the difference between the predicted and original pictures. The principle of
Motion Compensated Prediction is shown in Fig. 10.
೨⟎੍᷹
ⴕ੍᷹
ᐔ㕙੍᷹
Previous-sample prediction
Matrix prediction
Plane prediction
S
i
㧙
n
S
i
㧙
n
㧙
1
S
i
㧙
n
Scanning line
S
i
㧙
l
S
i
S
i
㧙
1
S
i
S
i
㧙
1
S
i
ᩏ✢
P
i
㧩
S
i
㧙
1
P
i
㧩
( S
i
㧙
1
㧗
S
i
㧙
n
) / 2
P
i
㧩
S
i
㧙
1
㧗
S
i
㧙
n
㧙
S
i
㧙
n
㧙
1
੍᷹ᑼ
Prediction formula
P
i
㧦੍᷹୯
n
㧦 ࠗࡦᒰߚࠅߩ↹⚛ᢙ
:
Prediction value
Number of pixels per line
㧦੍᷹ߔߴ߈↹⚛
Predicted pixel
㧦੍᷹ߦ↪ࠆ↹⚛㧔ᣢߦ╓ภൻߐࠇߚ↹⚛㧕
Reference pixel
Fig. 9
Examples of Intra Frame Prediction
䈀
Current frame
䈁
䈀
Prediction error
䈁
㽲
Divide current frame into
block
x
n
(
i, j
)
Transmit
prediction
error
picture
and
motion vector
㽳
Calculate distortion (Sum
of Absolute Difference:
D
L
) between
x
L
n
(
i, j
) and
x
L
n
䋭1
(
i, j
) of current and
previous reference frames,
respectively.
Motion-
compensated
Prediction
(MCP)
Motion vector
䈀
current frame
䈁
D
L
=
㰿
|
x
n
(
i, j
)
䋭
x
L
n
䋭
1
(
i, j
) |
i
,
j
㽴
Find the block with
minimum distortion in the
previous reference frame
and obtain motion vector
which indicates
translational movement
motion vector detection at
fractional-pel accuracy
䈀
previous reference frame
䈁
Fig. 10
Principle of Motion Compensated Prediction
Motion Compensated Prediction can reduce the energy of the residual signal
compared with the simple difference between frames. Fig. 11 shows an example
that compares the simple difference signal between frames and the Motion Com-
pensated Prediction difference signal. It is clear that the difference signal de-
creases dramatically when Motion Compensated Prediction is utilized.
Fig. 12 shows an example of the signal characteristics of the original, Intra
Frame Prediction, Inter Frame Prediction and Motion Compensated Prediction
pictures of a HDTV picture with their entropy and signal power. It is shown that
signal power decreases sharply when Motion Compensated Prediction is utilized.
