Digital Signal Processing Reference
In-Depth Information
where
N
is the codebook size and
S
is the number of codeword pixels. As for the
HVQ scheme, the total bit rate is
n
R
=
log (
NNM
)
×
,
(9)
2
ci
bi
i
=
1
where
b
N
is the number of blocks which are decoded by the codebooks
C
,
n
is
the total number of the layers and
M
is the total pixels in the original image.
Here for clear illustration, we give an example shown in Fig. 1 to demonstrate the
calculation of the total bit rate in different layers.
Fig. 1.
Blocks and sub-blocks of the image
codewords.
As shown in Fig. 1, there are three different layers to form this image. So the total bit
rate can be calculated as
Let us suppose all of the layers in Fig. 1 are reconstructed by
5
N
=
2
ci
R
=×+×+×
(1
5
2
5
8
5) / (16
×
8)
=
0.430 bpp.
5
Therefore the code length of every layer is
. For the same length of
the codewords in every layer, the choice of the codebooks in different layers becomes
ambiguous. In order to solve this problem, 2 more bit information
log (2 )
=
5
bit
2
is
added in the code length to indicate the exact codebook. And thus (9) can be rewritten
as
(log
4
=
2
bit
)
2
n
R
=
log (
Nt NM
+
)
×
,
(10)
2
ci
bi
i
=
1
where
t
is the extra bit information,
s
is the number of the
(0
≤≤
t l s
log
)
2
equal value in
c
N
. By using (10), the bit rate of the proposed HVQ can be calculated
as
R
=× + +× ++× +
1(5 2) 2(5 ) 8(5 0 ]/ 168)
×
=
0.461 bpp.
