Coded data generation or conversion – Quantizer
Reissue Patent
1999-03-31
2002-03-05
Young, Brian (Department: 2819)
Coded data generation or conversion
Quantizer
Reissue Patent
active
RE037568
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an image data compression/expansion technique, and more particularly to an inverse quantizing device for inverse-quantizing a compressed image data.
2. Description of the Prior Art
International Standards Organization (ISO) has proposed using the MPEG SM3 model as a data compression standard for an image in motion.
In image data compression, transform coefficients obtained by DCT(Discrete Cosine Transform)-transforming the image data is compressed into a predetermined amount of data by a quantizer, the operation of which is related to the following equations:
1) In intra mode:
QDC=dc/8
(i,j)=[16*ac(i,j)]//w(i,j)
QAC(i,j)=
(i,j)//g
where g is a quantization step size and g=2*QUANT,W(i,j) is a weight matrix, QAC is a quantized coefficient and QUANT is a quantization level.
2) In non-intra mode:
QAC
(
i
,
j
)
=
ac
(
i
,
j
)
/
(
2
*
QUANT
)
,
QUANT
=
odd
=
[
ac
(
i
,
j
)
+
1
]
/
(
2
*
QUANT
)
,
QUANT
=
even
,
ac
(
i
,
j
)
>
0
=
[
ac
(
i
,
j
)
-
1
]
/
(
2
*
QUANT
)
,
QUANT
=
even
,
ac
(
i
,
j
)
<
0
In the above equations, the DCT coefficients may be divided into dc coefficients having concentrated energy and ac coefficients having a large value of energy dispersion. Also, the quantizer performs quantization in two modes i.e., an intra mode and a non-intra mode. That is, the interrelationship of an input data is detected and if the detected interrelationship is relatively low, the intra mode is performed in which the dc and ac coefficients are quantized in different manners, while if the detected interrelationship is relatively high, the non-intra mode is performed in which the quantization is performed according to the above equation regardless of the dc or ac coefficients.
Inverse quantization is the inverse process of quantization, by which the dc or ac coefficients are calculated in accordance with the values of the quantized coefficients QAC and the quantization level QUANT.
FIG. 1
shows the construction of a conventional decoder adopting the MPEG SM3 standard as described above. According to the decoder of
FIG. 1
, data outputted from header detection section
1
for detecting a header signal from an input data is inputted to VLC(Variable Length Code) decoder
3
through buffer
2
.
VLC decoder
3
converts the input data into a FLC(Fixed Length Code) and the FLC is then inverse-quantized by inverse quantizer
4
. The output of inverse quantizer
4
is inverse-DCT-transformed by inverse DCT section
5
and the output of inverse DCT section
5
is applied to adder
7
. At this time, predictor
6
outputs a predicted value in accordance with the outputs of VLC decoder
3
and adder
7
, and this predicted value is applied to adder
7
to be added to the output of inverse DCT section
5
. Adder
7
outputs a final decoded signal.
The inverse quantizer mentioned as above, however, suffers from disadvantages in that hardware construction thereof is greatly complicated, causing the manufacturing cost to be increased and possibly complicating to a high degree the timing control for real time data processing as well.
SUMMARY OF THE INVENTION
It is an object of the present invention to provide an inverse quantizing device which has a simple structure utilizing memories, and thereby great reductions in manufacturing costs can be achieved.
It is another object of the present invention to provide an inverse quantizing device which enables real time processing of input data and easy control therefor.
In order to achieve the above objects, the inverse quantizing device according to the present invention comprises:
a plurality of memories for receiving input quantized coefficients and quantization levels as their address data and storing resultant data of inverse quantization in groups; and
control means for discriminating the group of the input data to be inverse-quantized in accordance with an input block-type and block-strobe signals and providing a selection control signal for selecting one of said plurality of memories which stores the resultant data of the discriminated group;
whereby the resultant data stored at an address of the selected memory which is assigned by the input quantized coefficients and the quantization levels is provided.
REFERENCES:
patent: 4862173 (1989-08-01), Nishitani
patent: 4933763 (1990-06-01), Chantelou
patent: 5617094 (1997-04-01), Chantelou
Fleshner & Kim LLP
Goldstar Co. Ltd.
Young Brian
LandOfFree
Inverse Quantizer does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Inverse Quantizer, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Inverse Quantizer will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2829290