Digital Signal Processing Reference
In-Depth Information
FIGURE 14.37
Original image.
After normalization, as shown in
Table 14.8
, the DC coefficient is reduced to 75, and a few small
AC coefficients exist, while most are zero. We can encode and transmit only nonzero DCT coefficients
and omit transmitting zeros, since they do not carry any information. They can be easily recovered by
resetting coefficients to zero during decoding. By this principle we achieve data compression.
As shown in
Table 14.8
,
most nonzero coefficients reside in the upper left corner. Hence, the order
of encoding for each value is based on the zigzag path in which the order is numbered, as in
Table 14.9
.
According to the order, we record the nonzero DCT coefficients as a JPEG vector, shown as
JPEG vector
:
½
75
L
1
L
1
0
L
1
3 2 EOB
where “EOB”
¼
end of block coding. The JPEG vector can further be compressed by encoding the
difference of DC values between subblocks, in
differential pulse code modulation
(DPCM), as dis-
cussed in Chapter 11, as well as by run-length coding of AC values and Huffman coding, which both
belong to lossless compression techniques. We will pursue this in the next section.
Table 14.6
DCT Coefficients for the Subblock Image in
Table 14.5
1198
10
26
24
5
16
0
12
8
6
3
8
0
0
0
0
0
3
0
0
8
0
0
0
0
0
0
0
0
0
0
0
0
4
0
0
0
0
0
0
0
0
1
0
0
0
0
0
10
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Search WWH ::
Custom Search