Coded data generation or conversion – Digital code to digital code converters
Reexamination Certificate
2000-08-16
2002-03-26
JeanPierre, Peguy (Department: 2819)
Coded data generation or conversion
Digital code to digital code converters
C348S403100
Reexamination Certificate
active
06362753
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a video data encoder and method of encoding the video data, more particularly to a technique for raising the compression efficiency in intraframe coding for motion estimation with using discrete cosine transform.
2. Description of the Related Art
Video data encoding (coding) technology has been improved with wide industry participation, and various global standards for video data encoding for multimedia contents have been established. For example, H.263 is an international standard for video data encoding suitable for transmitting video data file, and MPEG (Moving Picture Experts Group) is another one suitable for storing video data files. Each of those techniques employs unique encoding algorithm(s) for appropriate bit rate suitable for the purpose.
For example, a coding technique employed in MPEG-4 being standardized internationally, is hybrid of two data coding techniques, discrete cosine transform (DCT) and estimation with motion compensation (MC). Those coding techniques work as a basic algorithm for video data coding in MPEG-4.
FIG. 6
is a block diagram for explaining typical encoding process in MPEG-4. There are two approaches for video data encoding: intraframe coding and interframe coding. The intraframe coding is a way to compress video data by referencing data in each frame, while the interframe coding refers differences between frames for compression.
The first step of the intraframe coding is to divide video object plane (VOD) into macroblocks by a divider
6
a
. A macroblock is a unit for intraframe coding process.
FIG. 7
shows the structure of a macroblock. As shown in
FIG. 7
, a macroblock contains blocks of data from the luminance component (Y component) and from the chrominance components (Cb and Cr components). The structure shown in
FIG. 7
is based on one of the chrominance formats called “4:2:0 format” which includes four luminance blocks and two chrominance blocks. Each block is subjected to DCT by a discrete cosine transformer
6
d
, and thus transformed blocks are quantized by a quantizer
6
e
. Thus quantized data (DCT coefficient and amplitude) are subjected to variable-length coding by a variable-length coder
6
f
. Finally, resultant codes are output to be transmitted or the like. On the contrary, the quantized blocks are subjected to inverse quantization by an inverse quantizer
6
g
for restoring frame to be used as source frame for next estimation. Further, the DCT coefficient is subjected inverse DCT by an inverse discrete cosine transformer
6
h
. Thus restored VOP is transferred to a frame memory
6
j
via an addition circuit
6
i.
In interframe coding on the contrary, a motion detecting method such as block matching is carried out against a VOP adjoining to a VOP having target macroblock to detect estimative macroblock having least difference from the target macroblock by a motion vector detector
6
b
. A signal representing motion of the estimative macroblock having the least difference is called motion vector. And a frame to which referred for generating the estimative macroblock is called a reference VOP. The motion compensation is carried out against the reference VOP stored in a frame memory
6
j
based on the detected motion vector to obtain an appropriate estimative macroblock. A subtraction circuit
6
c
obtains difference between the target macroblock and the detected estimative macroblock. Thus obtained difference signal is subjected to DCT by the discrete cosine transformer
6
d
, and DCT coefficient is quantized by the quantizer
6
e
. The quantized DCT coefficient is subjected to variable-length coding by the variable-length coder
6
f
together with the motion vector and amplitude.
Details of DCT will now be described. The DCT is one of Fourier Transform. The DCT transforms 2 dimensional image into 2 dimensional frequency, and the frequencies are divided into conspicuous low frequency components and inconspicuous high frequency components. Moreover, it is able to obtain ubiquitous information. The following expression 1 represents 2D DCT on 8×8 blocks, and expression 2 represents inverse DCT.
F
(
u
,
v
)
=
(
1
/
4
)
C
(
u
)
C
(
v
)
∑
i
=
0
7
∑
j
=
0
7
f
(
i
,
j
)
cos
(
2
i
+
1
)
u
π
/
&AutoRightMatch;
16
cos
(
2
j
+
1
)
v
π
/
16
EXPRESSION
1
f
(
u
,
v
)
=
(
1
/
4
)
∑
i
=
0
7
∑
j
=
0
7
C
(
u
)
C
(
v
)
F
(
u
,
v
)
cos
(
2
i
+
1
)
u
π
/
&AutoRightMatch;
&AutoLeftMatch;
&AutoRightMatch;
16
cos
(
2
j
+
1
)
v
π
/
16
EXPRESSION
2
where C(u)=1/{square root over (2)}(u=0)
C(u)=1 (u≠0)
C(v)=1{square root over (2)}(v=0)
C(v)=1(v≠0)
IDCT (i, j): pixel values
DCT (u, v): DCT coefficients
FIG. 8
is a diagram showing a target block X and its peripheral adjoining blocks A, B, and C. Each of squares in each block represents a DCT coefficient. DCT coefficients represented by left upper squares in the blocks are direct current (DC) components, while the others are alternating current (AC) components. MPEG-4 has an advantage over MPEG-2 because MPEG-4 realizes improved efficiency for coding intra-type blocks by adaptively coding both quantized DC and AC components with a predictive coding technique, while MPEG-2 features differential coding to be applied only to DC components in intra-type blocks.
A DCT coefficient memory
6
k
stores quantized DCT coefficients of the reference block (DC components, and AC components for a first horizontal row or a first vertical column) as predicted values. Of the predicted values, values for DC components are subjected to predictive coding and output, while it is determined whether predictive coding should be applied to the AC components or not (a prediction selector
6
m
).
In a case whether the predictive coding should be applied to the AC components, a switch SW
1
is turned on, so that a subtraction circuit
61
calculates difference (prediction error) between the predicted values in the memory
6
k
and the quantized AC components and variable-length coding is carried out by the variable-length coder
6
f
. On the contrary, if the predictive coding is ineffective, the switch SW
1
is turned off, so that the variable-length coding performed by the variable-length coder
6
f
is just applied to the quantized AC components.
An operation of the conventional intraframe coding will now be described.
FIG. 9
is a flowchart for explaining process flow of the conventional intraframe coding. In the following explanation, for example, X(i, j) represents DCT coefficient in the block X (wherein i represents vertical components, j represents horizontal components; i, j=0, 1, 2, . . . , 7).
Input video data (VOP) are divided into unit macroblocks. Each of the macroblocks are subjected to DCT to be quantized (step A
1
). Then, sum of absolute values S
0
of the AC components of the DCT coefficients in the first vertical column of the first horizontal row is calculated (step A
2
). The following expression 3 represents the sum of the absolute values S
0
.
S
0
=&Sgr;|X(i,0)| or &Sgr;|X(0,j)|(i, j=1, 2, . . . , 7) EXPRESSION 3
Then, the sum total (&Sgr;mS
0
) of the sums of absolute values S
0
of the AC components for all blocks (Y, Cb, and Cr components) in the macroblocks is calculated, and the result is stored on a work area in a RAM
8
.
Then, the DC components in the quntized DCT coefficients are coded by the prediction coding. Prediction values of the DC components are adaptively selected based on inclines between horizontal and vertical DC components among blocks around the target block.
The following expression 4 represents prediction value X′ (0, 0
Casio Computer Co. Ltd.
Frishauf, Holtz Goodman, Langer & Chick, P.C.
Jean-Pierre Peguy
LandOfFree
Video data encoder and method of encoding video data does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Video data encoder and method of encoding video data, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Video data encoder and method of encoding video data will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2836708