Image analysis – Image compression or coding – Adaptive coding
Reexamination Certificate
1997-11-10
2002-12-31
Johnson, Timothy M. (Department: 2623)
Image analysis
Image compression or coding
Adaptive coding
C382S251000
Reexamination Certificate
active
06501858
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention relates to a quantizer table controller of quantizer/inverse-quantizer in an image compression and expansion apparatus.
2. Description of the Prior Art
FIG. 8
is a block diagram showing a conventional image compression and expansion apparatus. An operation of the first embodiment is explained below. First of all, in the coder portion, image data, for example, component image P
xy
(x, y=0,1,2,3 . . . 7) having an 8-bit width, is inputted from the image data input terminal
1
. The inputted image data is transmitted to DCT (Discrete Cosine Transformer)
2
. DCT
2
carries out a two dimensional Discrete Cosine Transform on the divided 8×8 picture element block P
xy
. As a result of the two dimensional Discrete Cosine Transform, 64 (=8×8) coefficients S
uv
are obtained. The 64 coefficients obtained are then rearranged from a serial order to a zigzag order in a zigzag transformer
23
and transmitted to the quantizer
3
. The 64 coefficients are quantized in the quantizer
3
by different step sizes at every coefficient location using the quantization table 4. The 64 quantized coefficients are transmitted to Huffman encoder
5
. Huffman encoder
5
carries out a coding operation according to Huffman coding system using the encoding table 6 and the encoded data are outputted from the output terminal
7
in units of several bytes (16-bits wide, for instance).
Next, in the decoder portion, the encoded data are inputted into an input terminal
8
in units of several bytes unit. The inputted encoded data are transmitted to Huffman decoder
9
. In Huffman decoder
9
, the data is then decoded to r
uv
by Huffman decoding system using the encoding table 6 and then transmitted to the inverse-quantizer
10
. Huffman decoded coefficients are inverse-quantized to S
uv
in the inverse-quantizer
10
per 64 coefficients using the quantization table 4 rearranged from a zigzag order into a serial order for every 8×8 image block at an inverse zigzag transformer
24
, and then transmitted to the inverse discrete cosine transformer (IDCT)
11
.
The inverse discrete cosine transformer
11
carries out a two dimensional Inverse Discrete Cosine Transform for every 8×8 picture element block. As a result of Inverse Discrete Cosine Transform, reconstruction image data P
xy
for every 8×8 picture element block are obtained and then the image data P
xy
are outputted from the image data output terminal
12
as a component image having an 8 -bit width.
Detailed operation is explained below. An 8×8 component image P
xy
(x, y=0, 1, 2, 3 . . . 7) is transformed using a two dimensional Discrete Cosine Transform at DCT
2
and the following coefficient S
uv
is obtained from formula (1):
S
uv
=
1
4
C
u
C
v
∑
x
=
0
7
∑
y
=
0
7
(
P
xy
-
L
s
)
cos
(
2
x
+
1
)
u
π
16
cos
(
2
y
+
1
)
v
π
16
(
1
)
where x, y=picture element position within block u, v=location of Discrete Cosine Transform coefficient
Cu
,
Cv
=
1
/
2
:
u
,
v
=
0
=
1
:
others
Ls
=
128
:
bit
accuracy
of
Pxy
=
8
bits
=
2048
:
bit
accuracy
of
Pxy
=
12
bits
Discrete Cosine Transform coefficient S
uv
is obtained by the two dimensional Discrete Cosine Transform. Discrete Cosine Transform coefficient S
uv
, as shown in
FIG.9
, comprises S
00
(DC coefficient) and the rest of S
01
~S
77
(AC coefficient). S
00
has a maximum value and other values of AC coefficient are very small compared with S
00
.
A more detailed explanation about the quantizer
3
and the inverse-quantizer
10
of the invention is given below. Discrete Cosine Transform coefficients obtained as above are divided using values Q
uv
of quantization table 4 in the quantizer
3
. In other words, a quantized Discrete Cosine Transform coefficient r
uv
is calculated as follows.
r
uv
=round (S
uv
/Q
uv
)
Where, the round function is a function which assigns an operation result of S
uv
/Q
uv
to a nearest integer number. Therefore, a quantization table value Q
uv
is determined such that the quantization table value Q
uv
becomes large where the two dimensional order uv increases. As a result, most of the AC coefficients become zero where the two dimensional order uv is large. Thus, the component image of 8×8 picture elements is transformed by two dimensional Discrete Cosine Transform and quantized in order to compress the component image, which greatly decreases the transmitted bits as a result.
On the other hand, in the decoder portion, a compressed picture signal inputted from the input terminal
8
is decoded in Huffman decoder
9
and transformed to quantized Discrete Cosine Transform coefficients r
uv
. Quantized Discrete Cosine Transform coefficient r
uv
is multiplied, i.e. inverse-quantized, by value Q
uv
of quantization table 4 in the inverse-quantization
10
.
Discrete Cosine Transform coefficient S
uv
is obtained by carrying out the inverse-quantization by the following formula:
S
uv
=r
uv
×Q
uv
Discrete Cosine Transform coefficient S
uv
thus obtained is transformed to a component image P
xy
of 8×8 picture elements by a two dimensional Inverse Discrete Cosine Transform shown in formula (2):
P
xy
=
1
4
∑
u
=
0
7
∑
v
=
0
7
C
u
C
v
S
uv
cos
(
2
x
+
1
)
u
π
16
cos
(
2
y
+
1
)
v
π
16
+
L
s
(
2
)
where x, y=picture element position within block u, v=location of Discrete Cosine Transform coefficient
Cu
,
Cv
=
1
/
2
:
u
,
v
=
0
=
1
:
others
Ls
=
128
:
bit
accuracy
of
Pxy
=
8
bits
=
2048
:
bit
accuracy
of
Pxy
=
12
bits
Since the quantizer/inverse-quantizer in a conventional image compression and expansion apparatus is constructed as explained above, a table value had to be updated because every different processing or expansion processing needs a different quantization table value. When updating a quantization table, it is difficult to update a quantization table without stopping the system. Therefore, it is necessary to stop the system before updating the quantization table in the conventional art.
SUMMARY OF THE INVENTION
It is an object of the present invention to provide an image compression and expansion apparatus which changes values of a quantization table of quantizer/inverse-quantizer in appearance by carrying out an operation at each processing step without actually writing new quantization table values in a quantization table.
It is a further object of the present invention to provide an image compression and expansion apparatus wherein values of register can be controlled from outside a CPU, which may easily change many kinds of tables.
It is a further object of the present invention to provide an image compression and expansion apparatus wherein the compressibility of image may be easily changed locally in the display by always monitoring the compression data and by changing a scaling factor according to various kinds of conditions such as a block unit and a block line unit.
It is a further object of the present invention to provide an image compression and expansion apparatus wherein the compressibility can be changed such that it is rough at the end of a frame and fine at the center portion of the frame.
According to a further aspect of the present invention, there is provided an image compression and expansion apparatus which comprises a register for setting necessary values in response to an outside signal; a data processing unit for carrying out an operation between values set into the register and values in the quantization
Johnson Timothy M.
Mitsubishi Denki & Kabushiki Kaisha
Oblon & Spivak, McClelland, Maier & Neustadt P.C.
LandOfFree
Image compression and expansion apparatus using a... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image compression and expansion apparatus using a..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image compression and expansion apparatus using a... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2949654