Method of architecture for converting MPEG-2 4: 2: 2

Method of architecture for converting MPEG-2 4: 2: 2
US006259741B1
(12) United States Patent
(10) Patent N0.:
(45) Date of Patent:
Chen et al.
(54)
US 6,259,741 B1
Jul. 10, 2001
METHOD OF ARCHITECTURE FOR
Van Dusen et al, “MPEG2 4:2:[email protected]—From Concept to an
CONVERTING MPEG-2 4:2:2-PROFILE
BITSTREAMS INTO MAIN-PROFILE
BITSTREAMS
Implementation”, International Broadcasting Convention,
Publication No. 428, IEEE Sep. 1996.*
Assuncao et al, “A Frequency—Domain Video Transcoder
for Dynamic Bit—Rate Reduction of MPEG—2 Bit Streams”,
IEEE Trans. on Circuits and Systems for Video Technology,
vol. 8, No. 8, Dec. 1998*
(75) Inventors: Xuemin Chen; Limin Wang; Ajay
Luthra; Robert O. Eifrig, all of San
Diego, CA (US)
(73) Assignee: General Instrument Corporation,
Horsham, PA (US)
ISO/IEC JTC1/SC29/WG11N1159,
Amendment 2, Jan. 1996, pp. 1—24.
(*)
ISO/IEC JTC1/SC29/WG11N2125, ISO/IEC 13818—2 Pro
posed Draft Amendment 5, Mar. 1998, pp. 1—5.
Notice:
Subject to any disclaimer, the term of this
patent is extended or adjusted under 35
U.S.C. 154(b) by 0 days.
(51)
(52)
(58)
Bjork, Niklas et al., “TRANSCODER ARCHITECTURES
Electronics, vol. 44, No. 1, Feb. 1998, pp. 88—98.
Feb. 18, 1999
Int. Cl.7 ..................................................... .. H04N 7/12
US. Cl. ................ ..
375/240.26; 375/240.29
Field of Search ....................... .. 375/240.03, 240.21,
375/240.25, 240.26, 240.29; 348/405.1,
419.1, 424.1, 425.1, 425.3, 441, 449, 453;
382/235, 234, 238, 244; 386/109, 111
(56)
References Cited
5,218,435
5,260,808
8/1996 Koppelmans et a1.
5,808,570
9/1998
*
Bakhmutsky
382/238
.........
. . . . ..
6,141,447 * 10/2000 Linzer et a1.
6,144,698 * 11/2000
341/65
382/236
Poon et a1. ......................... .. 375/240
FOREIGN PATENT DOCUMENTS
WO 97/47128
WO 99/51036
1/1996 (EP) .............................. .. HO4N/9/64
11/1997 (EP) .............................. .. H04N/7/95
4/1997 (W0).
10/1999 (WO) ............................ .. HO4N/1/5O
OTHER PUBLICATIONS
Horne et al, “Study of the Characteristics of the MPEG2
4:2:2 Pro?le—Application of MPEG2 in Studio Environ
ment”, IEEE Trans. on Circuits and Systems for Video
Technology, vol. 6, No. 3, Jun. 1996*
4:2:2P 4:2:2 305
BITSTREAM
Primary Examiner—Vu Le
(74) Attorney, Agent, or Firm—Barry R. Lipsitz
(57)
ABSTRACT
the pre- and-post-conversion bitstreams, including quantizer
6/1993 Lim et a1. ..................... .. 375/240.16
11/1993 Fujii .................... ..
358/458
5,544,266 *
* cited by examiner
A system for converting the color format of a digital video
bitstream. The system accounts for the alloWable formats of
U.S. PATENT DOCUMENTS
O 692 915
0 805 592
13818—2
FOR VIDEO CODING,”IEEE Transactions on Consumer
(21) Appl. No.: 09/252,135
(22) Filed:
ISO/IEC
WM
31
0
precision level, and Whether luma and chroma data have
separate quantization matrices, or share a common quanti
zation matrix. In a particular implementation, an MPEG—2
4:2:2 P bitstream having a color format of 4:2:2 or 4:2:0 is
converted to a MP bitstream having a color format of 4:2:0.
Coding efficiencies are achieved by using the luma quanti
zation matrix to re-quantize the chroma data, and re-using
luma motion vectors for performing motion compensation of
the chroma data. Further efficiencies can be achieved by
representing a 4:2:2 reference picture in a 4:2:0 format for
converting inter coded frames, and changing the position of
a pixel downsizing ?lter and clip function. Adjustment of the
quantization precision is provided as required. Atranscoding
function can also be achieved.
26 Claims, 6 Drawing Sheets
330
lNTRA/INTER
m
MODE
iiiiiiii *‘l
Fri-111M677
‘
‘
FOR LUMA
,
,,,,,,,, ,3
RATE CONTROL
SIGNAL
582\t iiiiiiiiiiiiiiiiii 1,
US 6,259,741 B1
1
2
METHOD OF ARCHITECTURE FOR
CONVERTING MPEG-2 4:2:2-PROFILE
BITSTREAMS INTO MAIN-PROFILE
BITSTREAMS
the High-1440 Level, With a maXimum of 1440 piXels per
line, and the High Level, With a maXimum of 1920 piXels per
line.
Furthermore, a 4:2:2 pro?le, also referred to as 4:2:2 P,
has recently been developed, Which accommodates both
10
4:2:2 and 4:2:0 color formats. See ISO/IEC 13818-2
Amendment 2, MPEG-2 4:2:2 Pro?le at Main Level, Janu
ary 1996; ANSI/SMPTE 308 M, SMPTE STANDARD for
television—MPEG-2 4:2:2 Pro?le at High Level, 1997; and
ISO/IEC 13818-2 Proposed Draft Amendment 5, 4:2:2 Pro
15
?le at High Level, March 1998, each of Which is incorpo
rated herein by reference.
The 4:2:2 pro?le is intended for professional video appli
cations Where ease of editing of compressed video and
multiple-generation encoding/decoding of video are impor
BACKGROUND OF THE INVENTION
The present invention provides a system for converting
the format of a digital video bitstream. The invention is
particularly suitable for converting a MPEG-2 digital video
data from a 4:2:2 Pro?le format to a Main Pro?le format.
The folloWing acronyms and abbreviations are used:
4:2:2P—4:2:2 Pro?le;
CBP—Coded Block Pattern;
DCT—Discrete Cosine Transform;
tant requirements. The primary applications targeted by this
HDTV—High De?nition Television;
pro?le are:
HL—High Level;
Storage
IDCT—Inverse DCT;
Editing and creation of visual effects
Video tape or disk recording for professional use
MB—Macroblock
MC—Motion Compensation;
ML—Main Level;
MP—Main Pro?le;
MV—Motion Vector;
QDC—QuantiZed Direct Current;
SDTV—Standard De?nition Television;
VBV—Video Buffer Veri?er;
VLC—Variable Length Coder; and
VLD—Variable Length Decoder.
Avideo image is de?ned by a number of picture elements,
(contribution quality)
Studio post-production of high-quality video sequences
Ef?cient transmission for storage and distribution of con
25
tribution quality video
4:2:2 P can provide higher video quality, better chroma
resolution and alloWs a higher bit-rate (at Main Level(ML),
up to 50 Mbit/s) than Main Pro?le (e.g., [email protected]). In
particular, 4:2:2 P provides separate quantiZation matrices
for luma and chroma data. In studio applications, very high
also knoWn as pixels or pels. A pixel, Which is the smallest
element of a raster scan line in the image, has an associated
color space. For eXample, in a YCrCb color space, Y is a 35
quality video and ITU-R 601 4:2:2 video format are often
needed for ease of chroma keying and other special effects.
Because of the requirement of ease of editing, more frequent
INTRA pictures are necessary, Which also results in high
coding bit-rates. 4:2:2 P permits all I-picture encoding. This
luminance component, and Cr and Cb are color difference
enables fast recovery from transmission errors and can
components. Various sampling formats have been de?ned,
including 4:4:4, 4:2:2, and 4:2:0. For eXample, With a 4:2:2
simplify editing applications.
format, a macroblock has four 8><8 Y blocks, tWo 8><8 Cr
blocks and tWo 8><8 Cb blocks. With this format, the
better quality image than MP With the same color format
since 4:2:2 P alloWs greater quantiZation precision. 4:2:2 P
further alloWs the high bit rates required to maintain high
Even the 4:2:2 P With a 4:2:0 color format can provide a
sampling frequencies for the Y, Cr and Cb components may
be 13.5 MHZ, 6.75 MHZ and 6.75 MHZ, respectively.
quality While using only I-picture coding. 4:2:2 P also alloWs
With a 4:2:0 format, a macroblock has four 8><8 Y blocks,
one 8><8 Cr block and one 8><8 Cb block.
Moreover, various digital video coding standards have
been developed for coding video data including, in
particular, the MPEG-2 standard, de?ned in ISO/IEC
13818-2 MPEG-2 Video (ITU-R H.262), 1995.
45
through multiple generations of encode/decode as local TV
stations add local programming information and commer
cials to video before it gets distributed to consumers for
reception at their homes, e. g., via a cable television netWork.
MPEG-2 designates several sets of constrained param
eters using a tWo-dimensional ranking order. One of the
Moreover, With analog TV, multiple generations of encode/
dimensions, called the “pro?le” series, speci?es the coding
decode can result in signi?cant picture quality losses. On the
other hand, 4:2:2 P can preserve high quality after multiple
generations of encoding/decoding. In the case of multiple
features supported. The other dimension, called “level”,
speci?es the picture resolutions, bit rates, and so forth. that
can be accommodated. The most important pro?le-level
combination is called Main Pro?le at Main Level, or
55
generations Without picture manipulation or change in pic
ture coding type betWeen generations, the quality from the
4:2:2 P coder remains nearly constant after the ?rst genera
[email protected] [email protected] supports a 4:2:0 color subsampling
tion. Use of picture manipulation or change in picture coding
ratio.
In addition to the Main Pro?le, other pro?les eXist. For
eXample, the Simple Pro?le is similar to the Main Pro?le but
has no B-pictures. The SNR Scaleable Pro?le adds SNR
scalability to the Main Pro?le. The Spatially Scaleable
Pro?le adds spatial scalability to the SNR Scaleable Pro?le.
The High Pro?le adds a 4:2:2 color format capability to the
Spatially Scaleable Pro?le.
The Main Level is de?ned for CCIR 601 video, While the
Simple Level is de?ned for Standard Intermediate Format
(SIF) video. Additionally, tWo higher levels for HDTV are
the use of P- and B-picture coding types, Which can further
improve quality or reduce bit rate for the same quality.
Furthermore, in a typical TV broadcast chain, video goes
type betWeen generations causes some degradation in qual
ity. Nevertheless, the resulting quality is acceptable for a
broad range of applications.
TV studios typically produce “contribution quality”
video, Which usually is ITU-R 601 source video, or 4:2:2
[email protected] (or 4:2:2 [email protected]) compressed video. This video is
65
then encoded and transmitted as “distribution quality” TV.
Digital television and HDTV in North America have
adopted MPEG-2 [email protected] and [email protected] as their video
coding standard. Both [email protected] and [email protected] encode and
US 6,259,741 B1
4
3
transmit only 41210 format distribution quality video. MP
Data corresponding to the chroma data With the second
chroma format provided by the ?ltering step is transformed
provides a common quantization matrix for luma and
chroma data. For example, the aggregate data rate for a
from a pixel domain to a transform domain, then quantiZed,
then inverse quantiZed and then inverse transformed to
HDTV system, Which includes compressed video, com
pressed audio, conditional access, and an auxiliary data
channel, is around 18 to 20 Mbits/s.
provide data for the second motion compensation process
ing.
Accordingly, there is a need for an ef?cient system to
convert a 41212 P pre-compressed contribution quality bit
stream to a MP distribution quality bitstream.
To reduce coding complexity, the system should alloW
reuse of motion vectors, avoid the need to change MB
The method may include the further steps of recovering a
?rst quantiZation precision level from the pre-conversion
bitstream, and if the ?rst quantiZation precision level is
10
greater than a maximum alloWed precision level of the
coding types, and use only a single MC unit for processing
chroma blocks. The system should further provide the
second format of the post-conversion bitstream:
capability to perform a decimate-?ltering process on residue
chroma-blocks in the INTER coded MBs.
luma transform data from the pre-conversion bitstream, and
(iii) re-quantiZing data corresponding to the recovered DC
luma transform data according to the loWered quantiZation
precision level.
The system should provide a simple approach to replace
the chroma quantization matrix during format conversion it
15
if appears in the 41212 P bitstreams.
The system should re?ne the CBP for chroma blocks.
Alternatively, When the pre-conversion bitstream com
prises inter coded images, and the recovered chroma data
The system should also provide adjustment of the MPEG
DC coef?cient precision variable, intraidciprecision, as
required to conform to the coding standard used. For
example, the maximum alloWed intraidciprecision level
has a ?rst chroma format that corresponds to the ?rst format
for a MP bitstream is loWer than that for a 41212 P bitstream.
The system should provide a combined 41212 to 41210
converter and a normal transcoder to perform 41212 P to MP 25
transcoding With a format conversion.
The system should avoid the need to fully decode the
of the pre-converted bitstream, the method includes the
further steps of: ?ltering the recovered chroma data to
provide residue chroma data that has a second chroma
format that corresponds to the second format of the post
conversion bitstream, recovering motion vectors associated
With the luma data from the pre-conversion bitstream, and
using the recovered luma motion vectors to perform motion
compensation processing of data corresponding to the resi
due chroma data for the inter-coded images to provide the
data for the re-compressing step. Here, the motion compen
sation processing uses the second chroma format, so only
processed bitstream during the format conversion.
The present invention provides a system having the above
and other advantages.
SUMMARY OF THE INVENTION
The invention relates to approaches to converting the
format of a digital video bitstream, such as for converting a
41212 P pre-compressed contribution quality bitstream to a
MP distribution quality bitstream.
A method for converting a pre-conversion bitstream hav
loWering
the ?rst quantiZation precision level, (ii) recovering DC
one motion compensator is required. The errors introduced
are generally not problematic since chroma errors are less
detectable than luma errors in the resulting image.
35
ing a ?rst format to a post-conversion bitstream having a
The ?rst format may be the MPEG 41212 Pro?le format,
and the second format may be the MPEG Main Pro?le
format.
A coded block pattern of the pre-conversion bitstream
second format, includes the steps of: at least partially
decompressing the pre-conversion bitstream to recover
chroma data therein in a pixel domain, recovering quanti
may be modi?ed for use in the post-conversion bitstream.
Zation matrix data associated With luma data from the
conversion bitstream may use the same macroblock coding
pre-conversion bitstream, and re-compressing data corre
sponding to the recovered chroma data. The re-compressing
includes re-quantiZing of the data corresponding to the
type.
Furthermore, the re-quantiZation step may be responsive
recovered chroma data according to the recovered luma
quantiZation matrix to provide the post-conversion bit
Moreover, the pre-conversion bitstream and the post
45
stream.
to improve coding ef?ciency of the post-conversion bit
When the pre-conversion bitstream comprises inter coded
stream.
images, and the recovered chroma data has a ?rst chroma
format that corresponds to the ?rst format of the pre
converted bitstream, the method includes the further steps
of: recovering motion vectors associated With the luma data
A corresponding apparatus is also presented.
BRIEF DESCRIPTION OF THE DRAWINGS
from the pre-conversion bitstream, using the recovered luma
FIG. 1 illustrates a ?rst approach to 41212 P to MP
motion vectors to perform ?rst motion compensation pro
cessing of the recovered chroma data for the inter-coded
images, Where the ?rst motion compensation processing
using the ?rst chroma format, ?ltering the chroma data after
the ?rst motion compensation processing to provide chroma
to a rate control signal for setting a bit rate of the post
conversion bitstream.
The recovered luma quantiZation matrix may be modi?ed
bitstream converting in accordance With the present inven
55
tion.
FIG. 2 illustrates a second approach to 41212 P to MP
bitstream converting in accordance With the present inven
second format of the post-conversion bitstream, and using
tion.
FIG. 3 illustrates chrominance block bitstream conversion
the recovered luma motion vectors to perform second
from a 41212 P 41212 bitstream to a MP 41210 bitstream in
motion compensation processing of the chroma data With the
second chroma format to provide the data for the
accordance With the present invention.
FIG. 4 illustrates a simpli?ed chrominance-block bit
data that has a second chroma format that corresponds to the
re-compressing step.
stream conversion from a 41212 P 41212 bitstream to a MP
With the second chroma format, one chroma block is
provided for at least every tWo chroma blocks in the ?rst
chroma format (e.g., tWo blocks for 41212 and one block for
41210 bitstream in accordance With the present invention.
FIG. 5 illustrates an approximated chrominance-block
41210).
65
bitstream conversion from a 41212 P 41212 bitstream to a MP
41210 bitstream in accordance With the present invention.
US 6,259,741 B1
5
6
FIG. 6 illustrates a chrominance-block bitstream conver
sion from a 41212 P 41210 bitstream to a MP 41210 bitstream
TABLE 1
in accordance With the present invention.
4:2:[email protected] and [email protected]
FIG. 7 illustrates a luminance-block bitstream conversion
from a 41212 P 41212 or 41210 bitstream to a MP 41210
bitstream When intraidciprecision=11 in accordance With
the present invention.
FIG. 8 illustrates a combination of a 41212 to 41210
converter and a transcoder in accordance With the present
invention.
4:2:[email protected]
[email protected]
Chroma format
4:2:2 or 4:2:0
4:2:0
Pro?le and level
identi?cation
10000101
1001000
Repeat ?rst ?eld
Constrained as [email protected],
With tWo additional
Table 8-7 in MPEG-2
Video
10
constraints for pictures
With more than 512 lines:
DETAILED DESCRIPTION OF THE
INVENTION
The invention relates to approaches to converting a 41212
P pre-compressed contribution quality bitstream to a MP
B pictures shall have no
repeat ?rst ?elds if the
15 Intra DC precision
Upper bounds for
sample density
distribution quality bitstream.
20
ing an MPEG-2 41212 [email protected] (or 41212 [email protected]) bit stream to
[email protected] (or [email protected]) bit stream can be a cascaded 41212
720 samples/line,
576 lines/frame,
Upper bound for
luma sample rate
Upper bound for
608
11,059,200
lines/frame
pixels/sec.
for 25
10,368,000 pixels/sec.
50 MB/sec.
15 Mbits/sec.
bit rate
Maximum VBV
Buffer size
P (@ML or @HL) or MP transcoder and a 41212 to 41210
converter With rate-control. The following transcoding cases
are of particular interest: 41212 [email protected] to [email protected], 41212
[email protected] to [email protected], and 41212 [email protected] to [email protected]
Transcoding from HL to ML is very common, e.g., from
HDTV format to SDTV format. HoWever, transcoding from
ML to HL generally is not of interest since it results in a
higher bit rate but does not improve picture quality. If one
8,9 or 10 bits
720 samples/lines, 608
lines/frame, 30 frame/sec.
(512 lines/frame for 30 Hz, 30 frames/sec
In general, a MPEG-2 transcoder is an instrument Which
converts a pre-compressed MPEG-2 bit stream into another
MPEG-2 bit stream at a neW rate. A transcoder for convert
frame rate is 25 Hz.
8,9,10 or 11 bits
9,437,184 bits
Quantization Tables Separate luminance and
chrominance quantization
tables
both luminance and
chrominance
Maximum number
unconstrained for 4:2:2,
4,608
of bits in a
macroblock
4,608 for 4:2:0
25
30
Wants a higher resolution from an ML bitstream, the bit
TABLE 2
stream can be decoded, and post-processing interpolation
can be used to enlarge the image.
4:2:[email protected] and [email protected]
There are tWo different approaches, as shoWn in FIGS. 1
and 2, to achieve 41212 P to MP transcoding.
In the Figures, like-numbered elements correspond to one
another.
4:2:[email protected]
[email protected]
Chroma format
4:2:2 or 4:2:0
4:2:0
Pro?le and level
identi?cation
10000010
1000100
Repeat ?rst ?eld
Constrained as [email protected],
With tWo additional
Table 8-7 in MPEG-2
Video
35
FIG. 1 illustrates a ?rst approach to 41212 P to MP
bitstream transcoding in accordance With the present inven
tion. Here, When a pre-compressed 41212 P bitstream is input,
constraints for pictures
40
With more than 1152 lines:
B pictures shall have no
a 41212 to 41210 bitstream converter 110 partially decodes the
41212 bitstream and re-assemble the results to provide a MP
bitstream. A conventional 41210 transcoder 120 is then used
repeated ?rst ?elds if
to generate a neW MP compressed bit stream at a neW rate. 45
Arate control function 130 controls the rate of the bitstream
Intra DC precision
Upper bounds for
sample density
Upper bound for
luma sample rate
Upper bound for
FIG. 2 illustrates a second approach to 41212 P to MP
bitstream transcoding in accordance With the present inven
tion. Here, When a pre-compressed 41212 P bitstream is input,
1920 samples/line,
1152 lines/frame,
62,668,800 pixels/sec.
62,668,800 pixels/sec.
300 MB/sec.
80 Mbits/sec.
47,185,920 bits
9,781,248 bits
Quantization Tables Separate luminance and
bitstream converter 220 to generate a MP compressed bit
chrominance quantization
stream at a neW rate, under the control of a rate control
Maximum number
55 of bits in a
macroblock
format is up-bounded by 4,608 bits, but is unconstrained for
bitstream converter 110, 220.
The 41212 [email protected] format is an extension to the [email protected]
format in many Ways. The key ML features are summarized
in Table 1. Similarly, 41212 [email protected] is an extension to
[email protected] in many Ways. The key HL features are summa
rized in Table 2.
8,9 or 10 bits
1920 samples/line, 1152
lines/frame, 60 Frame/sec.
bit rate
50 Maximum VBV
Buffer size
a 41212 P transcoder 210 is cascaded With a 41212 to 41210
a macroblock for the 41212 P 41212 chroma format.
The present invention focuses on the 41212 to 41210
the frame rate is 25 Hz.
8,9,10 or 11 bits
60 Frames/sec.
that is output by the transcoder 120.
function 230.
With rate control, special attention has to be paid to the
maximum number of bits in each macroblock since the
maximum number of bits in a macroblock for 41210 chroma
1,835,008 bits
The sample table for
60
65
tables
unconstrained
4,608 for 4:2:0
The sample table for
both luminance and
chrominance
4,608
Without changing the coded macroblock type in the 41212
P-bitstreams, the differences betWeen 41212 P and MP given
in Tables 1 and 2 result in the possible changes to syntax
listed in Table 3.
Note that the macroblock type, given by a VLC coded
term macroblockitype, indicates a number of different
characteristics of a macroblock, including:
Whether quantiscaleicode is present in the bitstream
(macroblockiquant),
US 6,259,741 B1
8
7
Whether forward or backward motion compensation is
TWo Weighting matrices are used for 41210 data —one for
used (macroblockimotioniforWard and macroblocki
intra macroblocks (luma and chroma), and the other for
non-intra macroblocks (luma and chroma). For 41212 data,
four matrices are used (intra luma, intra chroma, non-intra
luma, and non-intra chroma). Thus, different matrices can be
motionibackWard, respectively),
Whether codediblockipattern is present in the bitstream
(macroblockipattern),
used for luminance and chrominance data. Each matrix has
a default set of values Which may be overWritten by doWn
loading a user-de?ned matrix.
Whether the macroblock is intra-coded (macroblocki
intra),
Whether spatialitemporaliWeighticode is present in the
For the Weighting matrices W[W] [v] [u], W takes the
bitstream
(spatialitemporaliWeighticodei?ag), and
the permitted spatialitemporaliWeighticlasses.
10
values 0 to 3 indicating Which of the matrices is being used.
Table 4 summariZes the rules governing the selection of W.
TABLE 4
TABLE 3
Possible Syntax Changes
Selection of W
15
4:2:0
Elements of
syntax change Syntax Change Action
A
In Sequenceiheader( ), re-set vbvibufferisizeivalue=
1,835,[email protected] (9,781,[email protected]) and re-set bitiratei
value.
In Sequenceiextension( ), re-set pro?leiandileveli
indication=1 100 [email protected] (O 100 [email protected]).
intra blocks
20 (macroblockiintra = 1)
non-intra blocks
4:2:2
luma
chroma
luma
chroma
0
0
0
2
1
1
1
3
(macroblockiintra = 0)
In Pictureiheader( ), re-compute and re-setivbvidelay.
B
In Sequenceiextension( ), re-set chromaiformat=
C
In Pictureicodingiextension( ), re-set intraidci
precision=8,9, or 10(programmable).
Set chromai420itype=progressiveiframe;
In QuantiMatrixiextension( ), re-set load-chromai
“4:210”
D
For intra blocks, the DC value of a tWo-dimensional array
25
treated differently than the AC values. The DC value is
inverse quantiZed to produce the DCT transform DC coef
?cient array value F“[0] [0], according to the relationship:
quantiserimatrix=0 and remove chromaiintrai
F“[0] [0]=intraidcimult><QF[0]
quantiserimatrix[64].
E
In Quantimatrixiextension( ), re-set
30
loadichromainoniintraiquantiserimatrix=0 and
F
fullness just before decoding picture n, and R is the bit rate.
In general, the vbvidelay and vbvibufferisize re-setting
35
require a quantiZation feedback loop to avoid over- or
under-?oW of the VBV buffer and the actual rate buffer.
The neW coded block pattern (CBP) for the chroma block
of a MB in the MP-bitstream is determined by the
re-quantiZed chrominance data. The CBP indicates the num
“Sequence header”, “Sequence extension”, “Picture
Header”, “Picture coding extension”, and “Quant matrix
extension”, disclosed in the MPEG-2 standard.
Note that if the original intraidc precision=11 bits, it Will
Moreover, vbvidelay is re-computed from vbvidelay
(n)=Bd(n)/R, Where B d(n) is the decoder (VBV) buffer
remove chromainoniintraiquantiserimatrix[64].
Change Coded Block Patterns for each marcoblock.
The syntax change actions disclosed in Table 3 can be
understood further With reference to the syntaxes for
of quantiZed DCT transform coef?cients, QF[0] [0], is
ber of coded blocks for each macroblock, e.g., eight coded
40
be re-set to 8,9, or 10-bits for compatibility With the MP (see
blocks for 41212 color format, or six coded blocks for 41210
color format. For example, for a 41212 to 41210 conversion,
Tables 1 and 2).
The MPEG parameter intraidciprecision is 2-bit integer
the presence of tWo rather than four chroma blocks. No
de?ned to have the binary value 00, 01, 10 or 11 to designate
a precision of 8, 9, 10 or 11 bits, respectively. The parameter
syntax change F may involve changing the CBP to indicate
change to the number of luma blocks (e.g., four blocks)
45 occurs.
intraidciprecision indicates values for the multiplier,
intraidcimult, used for inverse quantiZing of DC coef?
cients of intra coded blocks. Namely, intraidcimult=8, 4,
To perform the 41212 P to MP conversion, the folloWing
key syntax elements need to be parsed from the 41212 P
bitstream:
2 or 1 for intra dc precision values of 00, 01, 10 or 11,
a=(intraidciprecision==11)?1:0;
b=(chromaiformat==“412:2”)?1:0;
respectively.
Generally, for inverse quantization, a tWo-dimensional
array of coef?cients, QF[v] [u], is inverse quantiZed to
c=(loadichromaiintraiquantiserimatrix==1)?1:0; and
produce the reconstructed DCT coefficients. This process is
d=(loadichromainoniintraiquantiserimatrix==1)
essentially a multiplication by the quantiZer step siZe. The
quantiZer step siZe is modi?ed by tWo mechanisms. Namely,
55
‘21:0.
The above statements use a C-language like pseudo-code
a Weighting matrix is used to modify the step siZe Within a
block, and a scale factor is used so that the step siZe can be
syntax. For example,
“a=(intraidciprecision==11)? 1:0”means “a” is assigned
modi?ed at the cost of only a feW bits (as compared to
encoding an entire neW Weighting matrix).
the value “1” if “intraidciprecision== 1” is true.
Appropriate inverse quantiZation arithmetic is performed
on the transform coef?cients in the bitstream to be converted
As mentioned, intraidciprecision must have a value of
8, 9 or 10 for MP. Accordingly, it must be reset if intrai
using quantiseriscaleicode, an unsigned, non-Zero integer
dciprecision==11 (decimal 11, or 1110).
OtherWise, “a” is assigned the value “0”.
in the range 1 to 31 that indicates the quantisation scale
The term chromaiformat is a tWo-bit integer indicating
factor, and Weighting matrices W[W] [v]
The resulting
coef?cients, F“[v] [u], are saturated to yield F‘[v] [u], and
the chrominance format of the bitstream to be converted.
then a mismatch control operation is performed to give the
?nal reconstructed DCT coef?cients, F[v]
65
Namely, chromaiformat 01 (binary) indicates a 41210
format, and chromaiformat=10 indicates a 41212 format.
Recall that the 41212 P accommodates both the 41212 color
US 6,259,741 B1
10
format (b=1) and 41210 color format (b=0). The 41210 format
of 41212 P is also an important case. It differs from ML even
TABLE 5-continued
though both of them have the same resolution since, With
41212 P, both luma and chroma can have different quantiZa
tion matrices. In contrast, With ML, both luma and chroma
use the same quantization matrix. Experiments shoW that
coding With the separate luma and chroma matrices can
Cases of Syntax Changes and Texture Converting Process
Syntax changes
provide a better quality image.
Conversion
Case
abcd (see Table 3)
Process
13
1101 A, B, C, E, F
FIGS. 3-5, 7
The term loadichromaiintraiquantiserimatrix is a
apply chromai
intraiquantiseri
10
matrix
14
1110 A, B, C, D, F
FIGS. 3-5, 7
apply chromainoni
intraiquantiseri
The term chromaiintraiquantiserimatrix is a list of
sixty-four, non-Zero 8-bit unsigned integers used for quan
tiZing intra chroma transform values.
The term loadichromainoniintraiquantiserimatrix is
a one-bit ?ag Which is set to “1” if chromainoniintrai
quantiserimatrix folloWs. If it is set to “0”, there is no
change in the values that shall be used. If chromaiformat is
15
20
25
30
With a higher precision, more bits are needed for trans
Options
mission. Moreover, both full pel and half pel motion vectors
vbvidelay and
40
vbvisize
1
0001 A, E
FIG. 6
apply chromai
2
0010 A, D
FIG. 6
matrix
apply chromainoni
intraiquantiseri
intraiquantiseri
for 41212 P to MP transcoding.
45
matrix
FIG. 6
FIGS. 3-5
FIGS. 3-5
matrix
0110 A, B, D, F
FIGS. 3-5
If the chroma Q-matrices (chromaiintraiquantiseri
50
apply chromainoni
quantiserimatrix) are applied in the re-quantiZation of the
matrix
0111 A, B, D, E, F
1000 A, C
FIGS. 3-5
FIGS. 6, 7
9
1001 A, C, E
FIGS. 6, 7
change intraidci
precision
55
change intraidci
apply chromai
stream in a scan order that is converted into the tWo
intraiquantiseri
dimensional Weighting matrix W[W] [u] [v] used in the
matrix
1010 A, C, D
FIGS. 6, 7
change intraidci
precision
60
1011 A, C, D, E
FIGS. 6, 7
12
1100 A, B, C, F
FIGS. 3-5, 7
inverse quantiZer as discussed above.
FIG. 3 illustrates chrominance block bitstream conversion
apply chromainoni
from a 41212 P bitstream to a MP bitstream in accordance
intraiquantiseri
With the present invention. FIG. 3 provides a complete block
diagram for a converter for converting tWo (possible)
chrominance blocks in a 41212 P bitstream (e.g., pre
matrix
11
chrominance blocks in accordance With the present inven
tion.
When the quantiZation matrices (Q-matrices) are pro
vided at a format converter, they are encoded in the bit
precision
10
matrix and/or chromainoniintraiquantiserimatrix) are
applied in the 41212 P-bitstream, the luma Q-matrices
(lumaiintraiquantiserimatrix and/or lumainoniintrai
intraiquantiseri
7
8
If the re-quantiZation process is applied in the format
conversion, intraidciprecision=8 should be used if it is
desired to reduce the number of coded bits.
apply chromai
intraiquantiseri
6
can be re-used. Since at least the half pel MVs are generated
from reconstructed pictures, a change in intra dc precision
could yield errors (e.g., drift), so minimiZing such a change
during format conversion could reduce the possible errors
changes
0011 A, D, E
0100 A, B, F
intraidciprecision=10 (the highest alloWed value for 41210
MP), but it can be over-Written to 8 or 9 by the user if
desired.
Conversion
0101 A, B, E, F
converting process, the motion vectors are re-used for the
MP bitstreams. Since intraidciprecision can be re-set to 8,
9, or 10 bits, there is a trade-off betWeen saving bits on
the accuracy of re-used motion vectors. Therefore, the
method to adjust intraidciprecision is to set the default
35
3
4
For the case of intraidciprecision=11 bits in a 41212 P
quantization of intra-DCT DC components and preserving
Cases of Syntax Changes and Texture Converting Process
5
luminance blocks are used for re-encoding the chrominance
blocks.
re-quantiZation of the chroma components). To simplify the
TABLE 5
0000 A
accordance With the present invention, if intraidci
precision-11 bits, only chrominance blocks need to be
bitstream, re-quantiZation of intra DCT DC luma compo
nents is required for the converting process (along With
rate Q-matrix for chroma. If it does, the chroma matrix is
replaced by a luma matrix.
0
change intraidci
precision
re-encoded, and no motion-estimation needs to be per
sion process. For example, for case=0, abcd=0000 (i.e., a=0,
b=0, c=0 and d=0). For the options to apply a Q-matrix, this
is determined for 41212 P to MP transcoding/converting by
Process
FIGS. 1, 2, 4,
5, 7
formed since motion vectors already determined by the
Table 5 de?nes the sixteen possible cases for the conver
detecting Whether or not the received bitstream has a sepa
1111 A, B, C, D, E, F
In the texture converting process of a format converter in
The term chromainoniintraiquantiserimatrix is a list
of sixty-four, non-Zero 8-bit unsigned integers used for
quantiZing non-intra chroma transform values.
matrix
15
“41210” (i.e., 01), this ?ag takes the value “0”.
abcd (see Table 3)
change intraidci
precision
10), the ?ag takes the value “0”.
Case
change intraidci
precision
one-bit ?ag Which is set to “1” if chromaiintraiquantiseri
matrix folloWs. If it is set to “0”, there is no change in the
values that shall be used. If chromaiformat is “141210” (i.e.,
Syntax changes
Options
change intraidci
precision
change intraidci
precision
65
conversion bitstream) into one chrominance block in a MP
bitstream (e.g., post-conversion bitstream). The converter
300 receives a 41212 P bitstream at a VLD function 305,
US 6,259,741 B1
11
12
Which provides MV data for ?rst and second chroma
function is subtracted. At DCT function 350, the data from
motion-compensation functions, chroma MC(1) 320 and
the adder 345 (comprising the current image 41210 piXel data
chroma MC(2), respectively.
for an intra coded image, or comprising difference 41210
piXel data corresponding to the difference betWeen the
current image and the reference image for an inter coded
image) is transformed to DCT coef?cients, and quantized at
a re-quantizer (Q) 355.
The re-quantization level Q1 is shoWn as corresponding to
Information regarding intra or inter mode status of the
received data is provided to sWitches 325, 327 and 370.
For intra mode data, the sWitch 325 is activated to pass a
null “0” signal to an adder 330, the sWitch 327 is activated
to pass a null “0” signal to an adder 345, and the sWitch 370
the inverse quantization level Q_1. HoWever, the
is activated to pass a null “0” signal to an adder 367. For
inter mode data, the sWitch 325 is activated to pass reference
10
image data from the chroma MC(1) function 320 to the
adder 330, and the sWitch 327 is activated to pass reference
image data from the chroma MC(2) function 325 to the
adder 345.
The chroma MC(1) function 320 is the motion
(Q2), based on a suitable control signal from a rate control
function, to provide transcoding of the received bitstream,
15
compensation unit used for the 41212 chroma data, While the
chroma MC(2) function 325 is the motion-compensation
unit used for 41210 chroma data. In accordance With the
present invention, luma MV data is used for motion com
pensation of the 41210 chroma MP data at the chroma MC(2)
function 325 (as Well as for motion compensation of the
41212 chroma 41212 P data at the chroma MC(1) function
re-quantization level may be adjusted to a different level
Wherein the MP bitstream is provided at a different rate than
the 41212 P bitstream.
See FIG. 8 for a full transcoder implementation.
The quantized data is provide to a VLC function 395 to
obtain the MP bitstream for transmission or other process
ing. For eXample, a transcoder may be concatenated With the
format converter 300.
20
The quantized data is processed at an inverse quantizer
360 and an IDCT function 365. The recovered current image
piXel data is then provided to an adder 367, Where it is
summed With either a null signal (for intra mode data) or
reference image data from the chroma MC(2) function 325
320). This avoids the need for separate motion estimation
processing, including searching in a reference frame, for the
25
(for inter mode data) according to the sWitch 370. The sWitch
370 is responsive to intra/inter mode information provided
to it (e.g., from the VLD 305).
The output from the adder 367 is clipped at a clip function
375 and provided to the chroma MC(2) function 325.
30
from the adder 330 is provided to a clip function 335 to clip
the data betWeen minimum and maXimum values if
In the converter 300, the dotted-line function units and
paths are conditional (i.e., are used in some cases). For
eXample, a QDC unit 380 may be used to perform the
necessary, e.g., in the range [0,255].
The clipped data is provided to a ?lter 340, Which
user setting, or if intraidciprecision=111O. Speci?cally,
41210 chroma MP data.
The data output from the VLD function 305 includes
quantized transform coefficient data from a current image
(e.g., frame). This data is provided to an inverse quantizer
Qfl 310, and to an IDCT function 315. The resulting piXel
domain data is provided to the adder 330, and the output
re-quantization of DC chroma coef?cients if requested by a
includes a ?eld-based vertical ?lter and a 211 vertical 35
doWnsampling ?lter. For eXample, the vertical ?lter taps
may be {—16,0,79,130,79,0,—16} for the top-?eld, While the
?lter taps are {1,7,7,1} for the bottom-?eld. The ?ltered
results are clipped into the range [0,255].
Note that the chroma MC(1) function 320 Will modify the
MVs received from the VLD 305 using 211 doWnsampling
of the horizontal
components of the MVs. Similarly, the
chroma MC(2) function 325 Will modify the MVs received
from the VLD 305 using 211 doWnsampling of both the
horizontal
and vertical (y) components of the MVs.
DC//2 for intraidciprecision=10. “DC” denotes the value
of the unquantized DC DCT coef?cient. “//” denotes integer
division With rounding to the nearest integer. Half-integer
40
A function 382 is provided for setting a neW intraidci
previously.
A chroma Q-matriX function 385 recovers the chroma
45
The clipped data is also provided to the chroma MC(1)
Note that some function units are embedded in other units.
For eXample, differential encoding and decoding of the
quantized DCT DC coefficients are considered to be part of
block in the original (received bitstream) tWo Cb or Cr
blocks of a MB having piXel values.
The ?ltered current image data, having the 41210 color
format, is provided to the adder 345, Where either a null
signal or reference image data from the chroma MC(2)
quantization matriX for use by the inverse quantizer 310. A
luma Q-matriX function 390 recovers the luma quantization
matriX for use by the re-quantizer 355 in re-quantizing the
41210 MP chroma coef?cients. Optionally, the luma Q-matriX
can be modi?ed for coding reasons, e.g., to improve coding
ef?ciency by adjusting the values in the matriX in a manner
that should be apparent to those skilled in the art. The
re-quantizer 355 also re-quantizes DC luma coefficients
When intraidciprecision=111O.
55
If no quantization matriX is present in the receives
bitstream, a default, such as the default MPEG quantization
matriX, may be used. The default matriX may be stored
locally at the converter 300, for eXample.
reference pictures.
A CBP function 315 may be used to change the CBP of
the outgoing MP bitstream. The luma Q-matriX function 390
is one factor Which could modify the CBP. Other factors are
an increase in the re-quantization level, and only one ?eld
values are rounded aWay from zero.
precision level (e.g., 8—10 bits) if required, as discussed
function 320.
VLC block 395 and VLD block 305, respectively. Inverse
quantization Q 1-1 at block 310 includes the saturation
process With the range [—2048,+2047] for each coef?cient
from the inverse quantization arithmetic. IDCT block 315
includes the saturation With the range [-256, 255] for each
inverse transformed value. The MC blocks, chroma MC(1)
320 and chroma MC(2) 325, include memories for the
QDC(8 bits)=DC//8 for intraidciprecision=8, QDC(9
bits)=DC//4 for intraidciprecision=9, and QDC(10 bits)=
To balance cost vs. performance trade-off, tWo simpli?ed
60
chrominance-block bitstream converters are provided neXt
in FIGS. 4 and 5.
FIG. 4 illustrates a simpli?ed chrominance-block bit
stream conversion from a 41212 P bitstream to a MP bit
stream in accordance With the present invention. In a con
verter 400, only one motion-compensation unit, namely the
65
chroma MC(2) function 325, is used. This saves Z/3 of the
frame-buffer memories compared With the converter 300 of
FIG. 3. The ?lter 340 in FIG. 4 is only applied to chromi
US 6,259,741 B1
13
14
nance blocks re-constructed from the IDCT function 315,
video bitstream. The system accounts for the alloWable
and no clipping is applied in the ?ltering process.
However, While the converter 400 provides signi?cant
formats of the pre- and—post-conversion bitstreams, includ
ing quantizer precision level, and Whether luma and chroma
reductions in complexity, errors are introduced by this
data have separate quantization matrices, or share a common
simpli?cation. Three error sources are:
quantization matrix. In a particular implementation, an
1. removing of chrominance MC(1) and representing the
41212 chroma reference picture in (a) the 41210 format and (b)
quantized by the neW luma quantizer matrix;
2. changing the position of the ?lter and clip function
blocks, and
3. integer operations of the ?ltering process.
Since the entire process only involves chrominance, the
MPEG-2 41212 P bitstream having a color format of 41212 or
41210 (i.e., the pre-conversion bitstream) is converted to a
MP bitstream having a color format of 41210 (i.e., the
post-conversion bitstream).
1O
introduced errors are likely very small since chroma errors
are less likely to be perceived in an image.
If, in addition to the above approximations, the errors
generated by quantization and clipping can be ignored, the
chrominance-block bitstream conversion process given by
15
the converter 400 of FIG. 4 can be approximated by the
Coding ef?ciencies are achieved by using the luma quan
tization matrix to re-quantize the chroma data, and re-using
luma motion vectors for performing motion compensation of
the chroma data.
Further ef?ciencies can be achieved by representing a
41212 reference picture in a 41210 format for converting inter
coded frames.
Further efficiencies can be achieved by changing the
converter 500 of FIG. 5.
FIG. 5 illustrates an approximated chrominance-block
position of a pixel doWnsizing ?lter and clip function.
bitstream conversion from a 41212 P bitstream to a MP
format pre-conversion bitstream.
Simpli?cations can also be made for a 41212 P 41210
bitstream in accordance With the present invention. In the
An efficient conversion for luma blocks is disclosed When
converter 500, a motion-compensation unit, namely the
chroma MC(3) function 510, is applied to difference (or
the quantization precision of the pre-conversion bitstream is
residue) images. By combining the IDCT function 315, ?lter
340, and DCT function 350 into one single unit 520, the
25
converter 500 reduces complexity. HoWever, such a struc
ture could introduce color drift.
For the case of 41212 P With 41210 color format bitstreams,
the converter can be further simpli?ed as shoWn in FIG. 6.
stream.
Although the invention has been described in connection
With various speci?c embodiments, those skilled in the art
Will appreciate that numerous adaptations and modi?cations
may be made thereto Without departing from the spirit and
FIG. 6 illustrates a chrominance-block bitstream conver
sion from a 41212 P bitstream With a 41210 color format to a
scope of the invention as set forth in the claims.
What is claimed is:
MP bitstream in accordance With the present invention. For
the converter 600, an approximation can be made for a
loW-cost implementation by removing the three dash-line
function units: IDCT 315, clip function 335, and DCT
not compatible With that of the post-conversion bitstream.
A transcoding format converter is also disclosed for
achieving a different bit rate in the post-conversion bit
1. A method for converting a pre-conversion bitstream
35
function 350.
having a ?rst format to a post-conversion bitstream having
a second format, comprising the steps of:
at least partially decompressing the pre-conversion bit
For luminance blocks, the only needed conversion is the
possible DC precision change 382 shoWn in FIG. 7.
stream to recover chroma data therein in a pixel
FIG. 7 illustrates a luminance-block bitstream conversion
recovering quantization matrix data associated With luma
data from the pre-conversion bitstream; and
re-compressing data corresponding to the recovered
chroma data;
said re-compressing including re-quantizing of the data
corresponding to the recovered chroma data according
to the recovered luma quantization matrix to provide
said post-conversion bitstream.
2. The method of claim 1, Wherein said pre-conversion
bitstream comprises inter coded images, and said recovered
domain;
from a 41212 P 41212 or 41210 bitstream to a MP 41210
bitstream When intraidciprecision=111O in accordance
With the present invention. The converter 700 includes the
function 382 for setting a neW intraidciprecision value
When the value is out of bounds, or based on a user input.
The various cases of the texture converting processes are 45
summarized in Table 5.
FIG. 8 illustrates a combination of a 41212 to 41210
converter and transcoder in accordance With the present
invention. The 41212 to 41210 converter 400 given in FIG. 4
chroma data has a ?rst chroma format that corresponds to
can be combined With a normal transcoder to provide a 41212
said ?rst format of said pre-converted bitstream, comprising
P to MP transcoder 800. In the transcoder/converter 800,
the further steps of:
recovering motion vectors associated With the luma data
re-quantization function (Q2) 855 provides a different quan
tization level than Q1., e.g., in response to a rate control
from the pre-conversion bitstream;
signal. A corresponding inverse quantization function Q2-1
860 is provided. The rate control signal may be generated by
55
a rate control function in a conventional manner to achieve
a desired bit rate for the MP bitstream.
Functions 810, 830, 850 and 867 correspond to functions
310, 330, 350 and 367, respectively. Functions 855‘, 860‘
motion compensation processing of the recovered
chroma data for the inter-coded images;
said ?rst motion compensation processing using said ?rst
chroma format;
?ltering the chroma data after said ?rst motion compen
sation processing to provide chroma data that has a
second chroma format that corresponds to said second
and 365‘ correspond to functions 855, 860 and 365, respec
tively.
A luma MC function 820 is also provided to provide
motion compensation of the luma data in the received
bitstream. Note that the MVs are re-used to avoid the need
for motion estimation.
Accordingly, it can be seen that the present invention
provides a system for converting the color format of a digital
using the recovered luma motion vectors to perform ?rst
65
format of said post-conversion bitstream; and
using the recovered luma motion vectors to perform
second motion compensation processing of the chroma
data With the second chroma format to provide said
data for said re-compressing step.
US 6,259,741 B1
15
16
13. The method of claim 1, Wherein:
the recovered luma quantization matriX is modi?ed to
3. The method of claim 2, wherein:
With said second chroma format, one chroma block is
provided for at least every tWo chroma blocks in said
?rst chroma format.
4. The method of claim 2, Wherein:
data corresponding to the chroma data With the second
improve coding ef?ciency of said post-conversion bit
stream.
14. An apparatus for converting a pre-conversion bit
stream having a ?rst format to a post-conversion bitstream
chroma format provided by said ?ltering step is trans
having a second format, comprising:
formed from a piXel domain to a transform domain,
means for at least partially decompressing the pre
then quantized, then inverse quantized and then inverse
conversion bitstream to recover chroma data therein in
transformed to provide data for said second motion
a piXel domain;
means for recovering quantization matriX data associated
With luma data from the pre-conversion bitstream; and
means for re-compressing data corresponding to the
recovered chroma data, including means for
re-quantizing the data corresponding to the recovered
chroma data according to the recovered luma quanti
zation matriX to provide said post-conversion bit
compensation processing.
5. The method of claim 1, comprising the further steps of:
recovering a ?rst quantization precision level from the
pre-conversion bitstream; and
if said ?rst quantization precision level is greater than a
maXimum alloWed precision level of the second format
of the post-conversion bitstream:
(i) loWering the ?rst quantization precision level,
(ii) recovering DC luma transform data from the pre
conversion bitstream, and
(iii) re-quantizing data corresponding to the recovered
stream.
15. The apparatus of claim 14, Wherein said pre
conversion bitstream comprises inter coded images, and said
DC luma transform data according to the loWered
quantization precision level.
recovered chroma data has a ?rst chroma format that cor
25
6. The method of claim 1, Wherein said pre-conversion
bitstream comprises inter coded images, and said recovered
further comprising:
means for recovering motion vectors associated With the
chroma data has a ?rst chroma format that corresponds to
luma data from the pre-conversion bitstream;
said ?rst format of said pre-converted bitstream, comprising
the further steps of:
?ltering the recovered chroma data to provide residue
means for using the recovered luma motion vectors to
perform ?rst motion compensation processing of the
chroma data that has a second chroma format that
corresponds to said second format of said post
conversion bitstream;
recovered chroma data for the inter-coded images;
35
recovering motion vectors associated With the luma data
compensation processing to provide chroma data that
using the recovered luma motion vectors to perform
has a second chroma format that corresponds to said
motion compensation processing of data corresponding
second format of said post-conversion bitstream; and
to the residue chroma data for the inter-coded images to
means for using the recovered luma motion vectors to
provide said data for said re-compressing step;
perform second motion compensation processing of the
45
chroma data With the second chroma format to provide
said data for said re-compressing means.
16. The apparatus of claim 15, Wherein:
With said second chroma format, one chroma block is
provided for at least every tWo chroma blocks in said
?rst chroma format.
17. The apparatus of claim 15, Wherein:
data corresponding to the chroma data With the second
chroma format provided by said ?lter is transformed
from a piXel domain to a transform domain, then
55
quantized, then inverse quantized and then inverse
transformed to provide data for said second motion
compensation processing.
18. The apparatus of claim 14, further comprising:
means for recovering a ?rst quantization precision level
from the pre-conversion bitstream; and
means for:
loWering the ?rst quantization precision
level, (ii) recovering DC luma transform data from the
stream use the same macroblock coding type.
12. The method of claim 1, Wherein:
said re-quantization step is responsive to a rate control
signal for setting a bit rate of the post-conversion
bitstream.
said ?rst motion compensation processing using said ?rst
chroma format;
a ?lter for ?ltering the chroma data after said ?rst motion
from the pre-conversion bitstream; and
Wherein said motion compensation processing uses said
second chroma format.
7. The method of claim 6, Wherein:
With said second chroma format, one chroma block is
provided for at least every tWo chroma blocks in said
?rst chroma format.
8. The method of claim 1, Wherein:
said ?rst format comprises a 4:2:2 color format, and said
second format comprises a 4:2:0 color format.
9. The method of claim 1, Wherein:
said ?rst format comprises an MPEG 4:2:2 Pro?le format,
and said second format comprises an MPEG Main
Pro?le format.
10. The method of claim 1, Wherein:
a coded block pattern of the pre-conversion bitstream is
modi?ed for use in the post-conversion bitstream.
11. The method of claim 1, Wherein:
the pre-conversion bitstream and the post-conversion bit
responds to said ?rst format of said pre-converted bitstream,
pre-conversion bitstream, and (iii) re-quantizing data
65
corresponding to the recovered DC luma transform data
according to the loWered quantization precision level, if
said ?rst quantization precision level is greater than a
US 6,259,741 B1
17
18
maximum allowed precision level of the second format
of the post-conversion bitstream.
19. The apparatus of claim 14, Wherein said pre
conversion bitstream comprises inter coded images, and said
21. The apparatus of claim 14, Wherein:
said ?rst format comprises a 412:2 color format, and said
second format comprises a 4:2:0 color format.
22. The apparatus of claim 14, Wherein:
said ?rst format comprises an MPEG 412:2 Pro?le format,
recovered chroma data has a ?rst chroma format that cor
responds to said ?rst format of said pre-converted bitstream,
and said second format comprises an MPEG Main
Pro?le format.
further comprising:
a ?lter for ?ltering the recovered chroma data to provide
residue chroma data that has a second chroma format
that corresponds to said second format of said post
10
23. The apparatus of claim 14, Wherein:
a coded block pattern of the pre-conversion bitstream is
modi?ed for use in the post-conversion bitstream.
conversion bitstream;
24. The apparatus of claim 14, Wherein:
the pre-conversion bitstream and the post-conversion bit
means for recovering motion vectors associated With the
luma data from the pre-conversion bitstream; and
mans for using the recovered luma motion vectors to 15
stream use the same macroblock coding type.
25. The apparatus of claim 14, Wherein:
perform motion compensation processing of data cor
responding to the residue chroma data for the inter
coded images to provide said data for said
re-compressing means;
Wherein said motion compensation processing uses said
said re-quantiZation means is responsive to a rate control
signal for setting a bit rate of the post-conversion
bitstream.
26. The apparatus of claim 14, Wherein:
the recovered luma quantization matrix is modi?ed to
second chroma format.
20. The apparatus of claim 19, Wherein:
improve coding ef?ciency of said post-conversion bit
With said second chroma format, one chroma block is
provided for at least every tWo chroma blocks in said
?rst chroma format.
stream.
25
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement