Image analysis – Image compression or coding – Adaptive coding
Reexamination Certificate
2000-06-16
2003-12-09
Couso, Jose L. (Department: 2621)
Image analysis
Image compression or coding
Adaptive coding
Reexamination Certificate
active
06661926
ABSTRACT:
FIELD OF THE INVENTION
The invention concerns a picture compression process in which each picture or picture macroblock is subjected to a coding chosen from among several types of coding.
It relates more particularly to the MPEG2 compression process. Although the invention is not limited to this type of compression, in the description hereafter, we shall refer mainly to this type.
BACKGROUND OF THE INVENTION
Recalled below is the principle of such compression and, at this juncture, the coding types which must be selected for each macroblock will be indicated.
In the MPEG2 standard, it is possible to start from a picture containing in progressive mode 576 rows of 720 points each. In interlaced mode, this picture is composed of two frames each of which comprises 288 rows, also of 720 points each.
Each picture is split up into macroblocks, each of which is formed by a square of 16×16 luminance points. Each macroblock is thus formed of 4 square blocks of 8×8 luminance points. With each of these 4 luminance blocks there are associated (in the 4.2.0 format) two chrominance blocks each of which has 8×8 points, one of the blocks representing the colour difference or red chrominance signal Cr and the other block representing the colour difference or blue chrominance signal Cb. In the 4.2.2 format, with each luminance macroblock there are associated four 8×8 chrominance blocks, 2 blocks for the blue chrominance and 2 blocks for the red chrominance. There is also a 4.4.4 format in which each of the luminance and chrominance components comprises 4 blocks of 8×8.
Represented in
FIG. 1
are four 8×8 luminance blocks, with reference
10
within the set and 8×8 chrominance blocks
12
and
14
for the blue and red chrominances respectively, the set illustrating a macroblock in the 4.2.0 norm.
Each block is coded by using a transformation denoted DCT which is a discrete cosine transformation which makes it possible to transform a luminance or chrominance block into a block of coefficients representing spatial frequencies. As may be seen in
FIG. 2
, a source block
16
is transformed into a block
18
of 8×8 coefficients. The upper left corner
20
of the block
18
corresponds to the zero spatial frequencies (mean luminance value of the block) and, onwards of this origin
20
, the horizontal frequencies increase towards the right, as represented by the arrow
22
, while the vertical spatial frequencies increase from top to bottom, as represented by the arrow
24
.
For each macroblock, it is necessary to choose the coding type: either “intra” or “inter”. Intra-coding consists in applying the DCT transformation to a source block of the picture, while inter-coding consists in applying the DCT transformation to a block representing the difference between a source block and a predicted block, or prediction block, of a preceding or following picture.
The choice depends in part on the type of pictures to which the macroblock belongs. These pictures are of three types: the first type is the so-called I or intra type, for which the coding is intra for all the macroblocks.
The second type is of P or prediction type; in pictures of this type, the coding of each macroblock can be either intra or inter. In the case of inter-coding of a macroblock of a picture of P type, the DCT transformation is applied to the difference between the current macroblock of this picture P and a prediction macroblock arising from the preceding I or P picture.
The third type of picture is called B or bidirectional. Each macroblock of such a picture type is either intra-coded or inter-coded. Inter-coding consists also in applying the transformation to the difference between the current macroblock of this B picture and a prediction macroblock. This prediction macroblock may arise either from the preceding picture or from the following picture or from both at once (bidirectional prediction), it being possible for the so-called preceding or following prediction pictures to be of I or P type only.
Represented in
FIG. 3
is a set of pictures forming a group called a GOP (Group Of Pictures) which comprises twelve pictures, namely an I picture followed by eleven B and P pictures according to the following succession: B, B, P, B, B, P, B, B, P, B, B.
In the case of predicted pictures (that is to say those deduced from other pictures), motion estimation followed by motion compensation are applied to the macroblock to be coded. This is because, between two pictures, the macroblock may be situated at different locations by reason of the inter-picture and inter-frame motions. The effect of motion compensation is to compute the prediction macroblock according to a given mode of interpolation (commonly called the prediction mode); this macroblock will actually serve as prediction for the current macroblock in inter-picture mode for a given coding mode. Hereinafter, this prediction macroblock and, by the same token this coding mode, will be retained or rejected depending on the decisions taken within the procedure for computing the choice of the coding mode.
Moreover, in the case of interlaced scanning, for which each picture is formed of two successive frames, an odd frame and an even frame, it is necessary to determine whether the DCT transformation should be performed progressively or individually on each frame. This is because, depending on the motion of the picture or the structure of this picture, the result of the coding may be different depending on whether the transformation is performed on the picture or on each frame.
This choice is represented by
FIGS. 4
a
and
4
b
. Represented in
FIG. 4
a
is a macroblock
28
of an interlaced picture formed of rows
30
1
,
30
3
, . . .
30
15
of an odd frame and of rows
30
2
,
30
4
, . . . ,
30
16
of an even frame.
FIG. 4
a
corresponds to a DCT transformation performed on the picture; each of the four blocks of the macroblock
28
is transformed without rearranging the rows. Thus, the coding is performed on the four blocks
28
1
,
28
2
,
28
3
,
28
4
forming the macroblock
28
and the transformation is performed on rows
30
1
to
30
8
for blocks
28
1
and
28
2
and on rows
30
9
to
30
16
for blocks
28
3
and
28
4
.
On the other hand,
FIG. 4
b
represents a transformation performed separately for the odd and even frames. Blocks
32
1
and
32
2
correspond to the odd frame and blocks
32
3
and
32
4
to the even frame. Thus, block
32
1
comprises rows
30
1
,
30
3
, . . . ,
30
15
, while blocks
32
3
and
32
4
comprise rows
30
2
,
30
4
, . . . ,
30
16
.
Represented in
FIG. 5
is a chart in block form representing the various operations to be performed in respect of the picture compression or video compression. Each digitized picture is applied to an input of a facility
40
which performs the separation into 8×8 blocks and these 8×8 blocks are transmitted to a facility
42
for selecting between the intra-coding and the inter-coding. If the coding chosen is intra, the block is transmitted to the DCT transformation facility
44
. If the coding is inter, the block is subjected to a subtraction by a subtractor facility
46
which takes the differences between the block itself and a prediction block delivered by a time prediction facility
48
.
After the DCT transformation
44
, a quantization
50
is performed and the quantized coefficients thus obtained are coded according to a VLC coding of variable or fixed length
52
. The coefficients thus coded obtained at the output of the coder
52
are directed to a buffer memory
54
whose output constitutes the coding output
56
. To avoid saturation and drying up of the buffer memory
54
, regulation
60
is performed which modifies the quantization
50
.
To be able to perform the time prediction, the output of the quantization facility
50
is linked to the input of a facility
62
for inverse quantization Q
−1
whose output is applied to the input of a facility
64
performing the inverse cosine transformation DCT
−1
. The output of the facili
Ruellou Pierre
Thoreau Dominique
Couso Jose L.
Eriksen Guy H.
Kurdyla Ronald H.
Thomson Licensing S.A.
Tripoli Joseph S.
LandOfFree
Picture compression process, especially of the MPEG2 type does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Picture compression process, especially of the MPEG2 type, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Picture compression process, especially of the MPEG2 type will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3146451