Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
RCA 86,380A
CA 02360556 2002-06-27
-1-
ERROR CONCEALMENT PP R TU FOR
A COMPRESSED VIDEO SIGNAL PROCE SING Y TEM
This application is a division of Canadian Serial No. 2,109,520 filed March
12, 1992.
This invention relates to apparatus for decompressing compressed video
data as for recording or transmission.
Over the past two decades significant effort has been committed to the
1 o compression of digitized video signals for purposes of image storage and
transmission. As a result many types of compression techniques have evolved
including the use of discrete cosine transforms, sub-band encoding, pyramid
transforms, intraframe encoding, interframe encoding and combinations of the
above
to name a few. More recently the International Organization for
Standardization has
i s developed a video compression standard for use in video storage
applications, e.g.,
CD-ROM. This proposed standard is described in the document "Coding of Moving
Pictures and Associated Audio", ISO-IEC JTC1/SC2/WG11, MPEG 90!176 Rev. 2;
December 18, 1990. Hereinbelow this system will be referred to as MPEG.
A feature of the MPEG standard is the use of both intraframe and
ao interframe coding techniques in combination with discrete cosine
transforms, run
length encoding and statistical (Huffman) encoding. Intraframe encoding in
general terms involves the encoding of an image frame from a single source
frame to provide sufficient encoded data for reconstruction of an image from
only the intraframe encoded data. Interframe encoding is the generation of
as encoded frame data from, for example, the differences between information
from
a current source frame and a frame predicted from prior frames. As such images
may not be reconstructed from a frame of interframe encoded data without
information from prior frames. The MPEG system incorporates two types of
interframe encoding. The first develops predictive frames (designated P
frames)
3 o from the current frame and a single prior frame. The second develops
bidirectionally
predictive frames (designated B frames) from the current frame and one
RCA 86,380A
CA 02360556 2002-06-27
-2-
or both of the prior and a subsequent frame. For example, assume that frames
occur in a sequence F 1, F2, F3, F4... and that frame F 1 is to be intraframe
encoded (designated I frame), frames F2 and F3 are to be B frame encoded and
frame F4 is to be P frame encoded. The P encoded frame is developed from
s differences between frame F4 and a predicted frame generated from a decoded
version of I frame F1 only. The B encoded frame representing frame F2 (F3) is
developed from differences between frame F2 (F3) and predicted frames
generated from both a decoded version of I frame F1 and a decoded version of
P frame F4. Exemplary circuitry for generating I, B and P encoded frames is
Zo described in "A Chip Set Core for Image Compression", by Alvin Artieri and
Oswald Colavin available prior to April 29, 1991 from SGS-Thomson
Microelectronics, Image Processing Business Unit, 17 avenue des Martyrs-B.P.
217, 38019 Grenoble Cedex France.
An exemplary sequence of I, B and P encoded frames is illustrated in
15 FIGURE 1A. In FIGURE 1A the upper blocks correspond to odd fields of
interlaced image data, and the lower blocks correspond to even fields of
interlaced image data. The MPEG system protocol designates that only the odd
fields respective frames are to be encoded. The exemplary sequence includes 9
frames of I, B and P encoded data which sequences occur cyclically. The
zo amount of encoded data of I frames is significantly greater than the amount
of
encoded data of P frames, and the amount of encoded data of B frames is less
than that of encoded P frames. The number of P frames between I frames and
the number of B frames between P or I and P frames is variable, i.e., it is
user
selectable within certain constraints. Nominally this selection is dependant
2 s upon the channel bandwidth and then image content.
The level of encoding provided by the MPEG protocol (e.g., odd fields
only and a continuous data transfer rate of 1.5 M bitslS) is sufficient to
produce
acceptable images in the computer display environment. However, those
CA 02360556 2001-10-25
3
skilled in the art of television signal processing will readily
recognize that the MPEG protocol as defined will not provide
images of current broadcast quality. It will also be recognized
that minor modifications to the protocol will provide
sufficient data to produce broadcast quality television images
or even HDTV images. These changes include doubling the
number of fields to be encoded as well as increasing the
number of lines per field and the number of pixels per line.
However, even with such modifications to the MPEG protocol,
1 0 certain deficiencies will still exist to preclude acceptable
performance with respect to image reception.
Regarding the TV environment, a first deficiency
of the MPEG system is the timing latency of image production
upon receiver turn on or channel change. An image cannot be
reproduced until an intraframe encoded frame of data is
available to the receiver. For the sequence of encoded frames
shown in FIGURE 1A, in the worst case, image reproduction
has a latency of at least nine frame intervals. A second
deficiency resides in the duration of image corruption due to
2 0 corruption or loss of data in data transmission. That is, if data
for an encoded I frame is lost or corrupted, the images
reproduced during the succeeding eight frames will be in
error, which error may become cumulatively worse over the
interval.
2 5 Independently encoding the odd and even fields
of image data such that intraframe encoded odd fields are
located midway between intraframe encoded even fields can
be utilized to significantly reduce the start-up interval and
image reproduction during channel changes. In addition it
3 0 provides signal information which may be utilized for signal
error concealment.
The present invention is directed to apparatus for
decompressing image data in which odd and even fields of
image data have been independently encoded according to
CA 02360556 2001-10-25
4
respective sequences of intra- and interframe compression
modes.
Apparatus for receiving and decompressing
compressed image data of the type wherein odd and even
fields of image data are independently compressed according
to respective sequences of interframe and intraframe
encoding modes includes circuitry for independently
decompressing said compressed odd and even fields of image
data. The decompressed odd and even fields of image data
are interleaved for display. Further circuitry is included for
substituting odd/even field data for even/odd field data
when valid even/odd field data is not available.
In a particular embodiment, for system start-up
or channel changes, apparatus is included to provide an image
representative signal for a predetermined interval,
corresponding to one of the decompressed odd or even fields
of data (to the exclusion of the other of the odd or even fields
of data) for which the first intraframe encoded field is
received.
Brief Description of the Drawin gs
FIGURES 1 A, 1 B AND 1 C are pictorial
representations of encoded sequences of fieldsof video signal,
useful for describing the invention.
2 5 FIGURE 2 is a block diagram of an exemplary
video signal encoding system embodying the present
invention.
FIGURE 3 is a block diagram of an exemplary
video signal compression apparatus.
3 0 FIGURE 4 is a block diagram of an exemplary
video signal decoding system.
FIGURE S is a block diagram of an exemplary
video signal decompression apparatus.
FIGURE 6 is a pictorial representation of
3 5 compressed signal format.
CA 02360556 2001-10-25
FIGURE 7 is a pictorial representation of a field
error memory useful for describing the invention.
FIGURE 8 is a flow chart of an exemplary process
for generating an image signal error map.
FIGURE 9 is a flow chart of the start-up sequence
of the FIGURE 5 apparatus.
The invention will be described in terms of the
MPEG field/frame protocol, however it should be appreciated
that it is applicable to any encoding format that provides
cyclic sequences of intraframe and interframe compressed
signal.
Referring to FIGURE 1A, the row of boxes
corresponds to respective fields of encoded video signal. Even
and odd numbered boxes correspond to even and odd fields
respectively. The type of encoding applied to the respective
fields (I, B or P) is indicated by the letter above each box. As
indicated above, the sequence of odd fields corresponds to the
MPEG protocol. Adding the even fields to the sequence,
increasing the number of lines per field and the number of
2 0 pixels per line modifies the protocol to provide sufficient
information for television image reproduction.
FIGURE 1 B illustrates an improved coding format,
according to the invention, for reducing image reproduction
latency and concealing signal transmission data loss or
2 5 corruption. In FIGURE 1 B the even fields are encoded
independently of the odd fields and the intraframe encoded
fields are offset by approximately one half the number of
fields in the cyclic sequence. The advantages that flow from
the FIGURE 1 B sequence are as follows. To begin image
3 0 reproduction an I field/frame is required. The sequence of
FIGURE 1 B includes an I field/frame every 9 fields whereas
the sequence of FIGURE 1 A includes an I field/frame only
every 17 fields. Thus the FIGURE 1 B sequence provides signal
entry points at intervals one-half as long as the intervals of
3 5 the sequence of FIGURE 1 A, without increasing the amount of
CA 02360556 2001-10-25
6
coded data. An image may be reproduced from only even
field data or only odd field data, albeit with half vertical
resolution. However, for providing images during channel
scans (sequencing through channels), and at start-up, the
rapidly reproduced lower resolution image is significantly
more acceptable than waiting twice as long for a full
resolution image. Regarding error concealment, assume that
data is lost from a portion of the I fields 1 and 2 of FIGURE
1A. This lost data will affect the reproduction of the images
from fields I-I8, and may induce unacceptable image
artifacts. Consider the loss of an equivalent amount of data
from fields 1 and 2 of the sequence of FIGURE I B. Data lost
from field 2 will only affect the reproduced image
corresponding to field 2 since field 2 is bidirectionally
I 5 predictive encoded. Data lost from the odd I field I has the
potential of affecting all of the odd fields in the sequence, and
thus corrupt all of the frames in the sequence. However, on
detection of lost data in the odd field sequence, data from the
even field sequence may be substituted for display. Such
2 0 substitution will momentarily provide less image resolution,
but this is far more acceptable than corrupted images.
FIGURES 1 A and I B illustrate sequences of fields
as they normally occur (disregarding the type of encoding).
FIGURE IC illustrates a field sequence as it would be
2 5 transmitted in an MPEG system. Recall that, for example,
bidirectionally predictive encoded fields 3 and 5 are
generated in part from I field 1 and P field 7. In order to
decode the B fields 3 and S, I field 1 and P field 7 must have
been previously decoded. Therefore, to facilitate decoding
3 0 and reduce the amount of data storage required in receivers,
the encoded B fields are arranged to follow the occurrence of
I and P fields from which decoding depends. This field
transmission arrangement illustrated in FIGURE IC
corresponds to the coding sequence of FIGURE 1B.
CA 02360556 2001-10-25
7
Apparatus for encoding video signal according to,
e.g., the field format of FIGURE 1C is illustrated in FIGURE 2.
Video signal is provided by a source 10, which may include a
video camera and preprocessing circuitry. The preprocessing
circuitry provides fields of video signal according to an
interlaced scan format and in pulse code modulated format
(PCM). Typically the source 10 will provide luminance, Y, and
chrominance, U and V, color difference signals, but for
purposes of this disclosure they will be referred to jointly as
the video signal. Nominally the luminance and chrominance
signals are independently compressed or encoded and ' then
combined for transmission, but those persons skilled in the
art of video signal compression will be aware of these
techniques and readily be able to implement same.
The video signal from the source 10 is coupled to
a multiplexer 12 which passes even fields of video data to a
first compressor apparatus 16 and odd fields of video data to
a second compressor apparatus 17. The multiplexer 12 is
controlled by a system control circuit 14, which is responsive
2 0 to field interval timing signals provided by the video source
10.
Compressor 16 is conditioned by the control
circuit 14 to compress respective even fields of video data
according to a predetermined sequence of intraframe and
2 5 interframe coding modes, e.g., I, B, P modes. Compressed
video data is applied to a buffer memory 18. Compressed
data from the buffer 18 is coupled to a transport packetizing
circuit 20. The packetizing circuit 20 includes circuitry for
parsing the data into blocks of predetermined amounts of
3 0 data including header information to identify each block as
well as information such as Barker codes for synchronizing
the detection of respective blocks at corresponding receiver
apparatus. The circuit 20 may also include error correction
circuitry for appending error check codes to the data to be
3 5 transmitted. The error correction circuitry may be in the
CA 02360556 2001-10-25
form of a Reed-Solomon error correction encoder. The
transport blocks are coupled to a transmitter 21 which may
be simply a data bus or as complicated as a broadcast
transmitter. In the latter instance the transport blocks of
data may be conditioned to quadrature amplitude modulate
(QAM) a carrier signal for application to a transmission
antenna.
Compressor 17 is conditioned by the system
controller 14 to compress the odd fields of video data
according to a predetermined sequence of intraframe and
interframe encoding modes, e.g., I, B, P. The mode sequence
may be similar to the mode sequence applied to the even
fields, or it may be an alternative sequence. In either event
the mode sequence applied to the odd fields is selected so
that intraframe encoded odd fields occur approximately
midway between intraframe encoded even fields, or vice
versa.
Compressed odd field video data provided by the
compressor 17 is coupled to the transport packetizer circuit
2 0 20, via a buffer memory 19.
The transport packetizer circuit is conditioned by
the control circuit 14 to alternately operate on even fields of
compressed data provided by the buffer 18 and odd fields of
compressed data provided by the buffer 19.
2 5 The buffers 18 and 19 are included because the
amount of compressed data for respective fields differ
according to the compression mode employed and the detail
attendant the image represented by the field of video data.
The differences in amounts of data result in fields of
3 0 compressed data occupying different time intervals, and thus
data output by the compressors 16 and 17 may not occur at
convenient times for interleaving the odd and even fields of
compressed data. The buffers provide accommodation for the
differences in occurrence of the compressed data provided by
3 S the respective compressors.
CA 02360556 2001-10-25
9
The apparatus of FIGURE 2 is shown with first and
second separate compression circuits for compressing the
even and odd fields of data. It should be appreciated that a
single compressor may be employed to perform compression
of both the even and odd fields.
FIGURE 3 illustrates an exemplary compressor
apparatus which may be utilized for compressing both even
and odd fields according to the sequence illustrated in FIGURE
1C. The assumption is made that source fields of video signal
1 0 have been rearranged to occur in the numbered sequence
shown in FIGURE 1C. The compressor provides compressed
data according to I, B, P modes. Intraframe compression
consists of performing discrete cosine transforms over 8 X 8
blocks of pixels and then variable length encoding the
transform coefficients. Predictive compression (P fields)
consists of determining motion vectors which indicate 16 X 16
blocks of pixels from a prior I field which most closely
correspond to 16 X 16 blocks of pixels in the current field. A
predicted field is generated using the motion vectors and data
2 0 from the prior I field, and the predicted field is subtracted
from the current field on a pixel by pixel basis to generate
residues. A discrete cosine transform is then performed on 8
x 8 blocks of the residues. The transform coefficients of the
residues are variable length encoded, and the motion vectors
2 S plus the residue coefficients are non additively combined to
form coded P fields. Bidirectional predictive fields (B) are
formed similarly to the P fields except that the motion vectors
and corresponding residues are associated with both prior
occurring and subsequent fields of video data.
3 0 The apparatus shown only includes the circuitry
required to generate compressed luminance data. Similar
apparatus is required to generate compressed chrominance U
and V data. In FIGURE 3 the memory and storage elements
101, 102, 114 and 115 are each arranged to store an odd field
3 S of data and an even field of data in separate memory sections.
CA 02360556 2001-10-25
When an even (odd) field is being processed the sections of
the respective memory and storage elements designated for
storing even (odd) fields are accessed. In addition there are
elements 104 and 105 designated as elements for computing
S forward and backward motion vectors respectively. Since
whether a motion vector is forward or backward depends
only upon whether the current field is analyzed with respect
to a prior or succeeding field, both elements are realized with
similar circuitry, and in fact both elements 104 and 105
10 alternate on a field/frame basis between generating forward
and backward vectors. The elements 104 and 105 may be
realized using integrated circuits of the type designated STI
3220 MOTION ESTIMATION PROCESSOR available from SGS-
THOMSON MICROELECTRONICS. In order to achieve the
necessary processing rates each of the elements 104 and 105
may comprise a plurality of such integrated circuits operating
simultaneously on different areas of respective images.
Element 109 designated DCT & Quantize performs
the discrete cosine transform ~ and quantization of transform
2 0 coefficients and may be realized using integrated circuits of
the type designated STV 3200 DISCRETE COSINE TRANSFORM
available from SGS-THOMSON MICROELECTRONICS. Element
109 may also be realized with a plurality of such devices
operated in parallel to concurrently process different areas of
2 5 the image.
Even and odd fields occur alternately and
sequentially, and the compressor of figure 3 alternately
compresses odd and even fields. Compression of even and
odd fields is similar except for the relative sequence of intro
3 0 and interframe compression modes. The sequence is
programmed into the controller 116 for both even and odd
field sequences, and communicated to the respective
processing elements via a control bus CB. Since the
compression function is conceptually the same for both even
CA 02360556 2001-10-25
11
and odd field sequences, an explanation of only the even field
compression will be provided below.
Refer to FIGURE 1 C and assume that even field 10
is currently available. Previously occurring even P field 4 has
been snatched and stored in the even field section of the
buffer memory B 101. In addition a previously generated
predicted even field 4 has been stored in the even field
section of one of the buffer storage elements 114 or 115. As
field 10 occurs it is stored in the even field section of the
buffer memory A, 102. In addition field 10 is applied to a
working buffer memory ~ 100. As field 10 occurs, appropriate
blocks of image data are coupled from the memory 100 to the
minuend input of a subtracter 108. During compression of
the I fields the subtrahend input of the subtracter 108 is held
at a zero value so that data passes through the subtracter 10~
unaltered. This data is applied to the DCT and quantizer
element 109 which provides quantized transform coefficients
to elements 110 and 112. Element 112 performs inverse
quantization and inverse DCT transformation of the
2 0 coefficients to generate a reconstructed image. The
reconstructed image is applied via an adder 113 to, and
stored in, the even field section of one of the buffer storage
elements 114 and 115 for use in compressing subsequent B
and P fields. During compression of I frames no information is
2 S added (by adder 1 13) to the reconstructed image data
provided by element 112.
Element 110 performs variable length encoding
(VLC) of the DCT coefficients generated by element 109. The
VLC codewords are applied to a formatter 111 which
3 0 segments the data and appends appropriate header
information to facilitate decoding. Coded data from element
111 is then passed to a further buffer memory (not shown).
The formatter may also be arranged to provide field indicia to
the transport packetizing circuit for generating corresponding
3 S transport block headers. Each of the elements 109, 110 and
CA 02360556 2001-10-25
12
111 are controlled by the system controller 116 to cyclically
perform the appropriate operations at the appropriate times.
After the occurrence and compression of even
field 10 an even field 6 (B) occurs and is loaded into buffer
memory 100. Data from even field 6 is coupled to both of
elements 104 and 105. Element 104, responsive to data from
even field 6 stored in memory 100 and data from even field 4
stored in memory 101, calculates forward motion vectors for
respective blocks of 16 x 16 pixels of image data. It also
provides a distortion signal which is indicative of the relative
accuracy of the respective forward motion vectors. The
forward motion vectors and the corresponding distortion
signals are coupled to an analyzer 106.
Element 105, responsive to data from field 6
stored in memory 100 and data from I field 10 stored in
memory 102, generates backward motion vectors and
corresponding distortion signals which are also coupled to the
analyzer 106. Analyzer 106 compares the distortion signals
against a threshold, and if both exceed the threshold, provides
2 0 both the forward and backward motion vectors as the motion
vector, and also provides a corresponding signal related to the
ratio of the distortion signals. Upon reconstruction predicted
images are generated using both forward and backward
vectors and corresponding field data from which derived. An
2 5 interpolated field is generated from the forward and
backward predicted fields in accordance with the ratio of
distortion signals. If the distortion signals for both the
forward and backward motion vectors are less than the
threshold, the motion vector with the corresponding lesser
3 0 valued distortion signal is selected as the block motion vector.
After the motion vector has been determined, it is
applied to the motion compensated predictor 107 which
accesses the appropriate data block defined by the vector or
vectors from the previously regenerated field 10 or field 4 or
3 5 both, stored in the even field sections of the storage elements
CA 02360556 2001-10-25
13
114 and 115. This data block is applied to the subtrahend
input of the subtracter 108 wherein it is subtracted on a pixel
by pixel basis from the corresponding block of pixel data from
the current field 6 provided by the buffer memory 100. The
differences or residues are then encoded in element 109 and
the coefficients applied to element 110. The corresponding
block vector is also applied to element 110. The motion
vectors are variable length encoded in element 110. The
coded vectors and coefficients are then transferred to the
formatter 111. The encoded B fields are not inverse
quantized and inverse transformed in element 112 since they
are not used for subsequent encoding.
P fields are similarly encoded except that only
forward motion vectors are generated. For example P field 16
is encoded with motion vectors associating corresponding
blocks of I field 10 and P field 16. During encoding of P
fields, element 112 provides corresponding decoded residues
and element 107 provides the corresponding predicted P
field. The predicted field and the residues are added in adder
2 0 1 13 on a pixel-by-pixel basis to generate the reconstructed
field which is stored in the even field section of the one of
storage elements 114 and 116 not containing the even field
information from which the predicted even P field is
generated. The reconstructed and stored even P field is used
2 5 for encoding subsequent even B fields. For both P and B
fields it should be noted that DCT's are performed on a block
basis (e.g., a matrix of 8 x 8 pixels), but motion vectors are
calculated for macroblocks (e.g., a 2 x 2 matrix of luminance
of blocks or a 16 x 16 matrix of pixels).
3 0 FIGURE 4 illustrates an exemplary receiver
apparatus for processing transmitted compressed video signal
occurring as interleaved odd and even fields which have been
independently encoded in sequences of intraframe and
interframe encoding modes. The transmitted signal is
3 5 detected by a detector 40 which may include a tuner, IF
CA 02360556 2001-10-25
14
circuitry and a QAM demodulator. Detector 40 provides a
signal in conformance with the signal provided by the
transport packetizer 20 of FIGURE 2. This signal is coupled to
a transport processing circuit 43. Transport processing circuit
43 includes an error check/correction circuit, which,
responsive to the error check codes appended to the
transmitted signal, corrects signal errors incurred during
transmission. If uncorrectable errors occurred a flag is
generated and communicated to the receiver system
controller 42. The transport processor 43, responsive to
transport header information included in the transport blocks,
identifies odd and even fields of data, and reconfigures the
transmitted signal from transport block format, to a format
which is in conformance with compressed information
1 5 provided by the buffers 18 and 19 of FIGURE 2. The
reconfigured data is coupled to a multiplexer 44. A control
signal, corresponding to the current field type (odd/even) is
provided by the transport processor 43 to condition the
multiplexer 44 to pass odd field data to a decompressor 45
2 0 and even field data to a decompressor 46. The decompressors
45 and 46 perform decompression of the odd and even field
compressed video data respectively, and provide
decompressed video signal to buffer memories 47 and 48.
In this example it is assumed that the compressed
2 5 signal is of the form illustrated in FIGURE 1C, but that the
decompression circuitry 45 and 46 provide decompressed
data reordered in the normal field sequence as per FIGURE
1 B, for example. The reordered fields from buffer memories
47 and 48 are coupled to a multiplexer S 1 which, in the
3 0 steady state with no loss or corruption of data, alternately
couples odd and even fields of data to a video display RAM
52. It is assumed that the display RAM has sufficient storage
capacity to hold one frame of data. Frames of data are
thereafter read from the display RAM in either interlaced or
3 5 non interlaced format for display purposes. The receiver
CA 02360556 2001-10-25
apparatus is controlled by the controller 42 which is
programmed to coordinate the decompression and display of
received video data according to a normal cycle of operations.
Immediately after system turn-on or channel
5 changes effected by user control 41, the system controller 42
initiates a start-up cycle to provide image reproduction as
quickly, as possible. Once image display of a full sequence of
fields of data (a sequence encompassing two successive
intraframe encoded mutually exclusive odd or even fields)
10 has been accomplished, the controller switches to the normal
decompression cycle of operation. At start-up, image
reproduction cannot occur until at least one intraframe
encoded field is received, since reproduction of interframe
encoded fields (P or B) require data from an intraframe
1 5 encoded field. The controller 42 monitors the field types
received responsive to header data provided by the transport
processor 43. The controller precludes display of received
field data until an intraframe encoded field is detected. Its
field type (odd or even) is checked and decompression is
2 0 performed on the successive fields of the same type as the
first occurring intraframe encoded field. Display of fields of
the opposite field type is forestalled until the occurrence of
the first occurring intraframe encoded field of that type,
which occurs a known number of fields after the detection of
2 5 the first intraframe encoded field. At start-up, display of the
first decompressed I frame may be repeated for the number
of frame intervals between an I encoded field and the first B
encoded field following the first P encoded field occurring
after the I encoded field. Note in FIGURE 1 C, if field 10 is the
3 0 first occurring I encoded field, the B fields 6 and 8 cannot be
decoded without the prior P encoded field 4, (which is not
available). The next even field to be displayed after the I
field 10 in the normal display field sequence is field 12 which
occurrs four frame intervals after field 10. Alternatively,
3 5 rather than repeatedly displaying the first occurring I field
CA 02360556 2001-10-25
16
the image display may be inhibited for e.g. four frames
following the occurrence of the first I field.
During channel changes the system may be
conditioned to repeatedly display the last image from the
prior channel until the system has synchronized to display
images from the newly selected channel.
Assume that the first intraframe encoded field is
odd. Successive odd fields are decompressed and provided by
the buffer memory 47 to the multiplexer Sl. At this juncture
several options are available regarding display of the
received data. The first ~is to write the odd fields to the odd
field lines of the display RAM 52, and to set the even field
lines of the display RAM to an, e.g., intermediate gray value
and display the image. A second is to write the odd field data
I 5 to the odd field lines of the display RAM, then read the same
odd field data from the buffer memory 47 a second time, and
write it to the even field lines of the display RAM, and display
the image. The second option will provide a brighter image
than the first option and with apparent greater resolution. A
2 0 third option is to write the odd field data to the odd field lines
of the display RAI~t, then to read the same field from the
buffer memory 47 and apply same to an interpolator 50 vicr a
rnultiplexer 49. The interpolator 50 may be arranged to
generate interpolated lines of data from successive pairs of
2 5 lines of the odd field signal (vertical averaging), thereby
producing pseudo even lines of data which are subsequently
written to the even field lines of the display RAM 52. 'This
option will produce images which have apparent greater
resolution than the second option.
The particular option employed is programmed
into the controller 42, and is part of the start-up cycle. 'fhe
controller, responsive to the data provided by the transport
processor 43, controls the reading of the data from the
appropriate buffer memory 47 or 48 (depending whether the
3 5 first intraframe encoded field is odd or even) and controls the
CA 02360556 2001-10-25
17
switching of the multiplexers 49 and 51. For example,
considering the third option and assuming that the first
intraframe encoded field is odd, the buffer memory 47 is
conditioned to read each field of signal twice, the multiplexer
49 is conditioned to pass signal from buffer memory 47, and
the multiplexer 51 is conditioned to alternately pass fields of
signal from the buffer memory 47 and the interpolator 50.
After a predetermined number of fields have been processed
in this manner the controller switches to the steady state
1 0 control cycle to decompress data from both odd and even
field types.
As indicated earlier, the transport processor may
provide error flags indicative of lost or uncorrectable errors.
To ameliorate potential unacceptable image corruption from
1 S such errors or lost data, the controller may be arranged to
condition the receiver system to substitute noncorrupted
signal. For instance, if the lost or erroneous data occurs in an
intraframe encoded field, the controller may be arranged to
revert to processing similar to the above described option
2 0 three (except that there is no need to wait for a first
intraframe encoded field, assuming that the lost data only
occurred in an odd or an even field; alternatively if data is
lost in both odd and even fields, the controller will revert to
the start-up cycle). If data is lost in a P field the controller
2 5 may again be arranged to condition the system to operate as
per option three. Alternatively, if data is lost from a B field,
the controller may be arranged to condition the system to
replace this data, on a single field basis, or partial field basis,
with interpolated data as per options two or three.
3 0 FIGURE 5 illustrates an exemplary arrangement of
a single decompression apparatus for decompressing both
even and odd field data.
Generally the circuitry of FIGURE 5 is arranged to
decompress video data provided in MPEG-like format. The
3 5 apparatus includes two buffer memories 314 and 316, each of
CA 02360556 2001-10-25
18
which has memory capacity to store an odd field and an even
field of decompressed image data. When odd (even) fields
are being decompressed the odd (even) field portions of the
memories are enabled and vice versa. Decompressed data is
provided by an adder 312, and is applied to a multiplexer
320 and the memories 314 and 316. When B fields are being
decompressed, this image data is passed by the multiplexer
320 from adder 312 to the display RAM 318 (via multiplexer
322). When I or P fields are decompressed;. the
decompressed image data is written to one of the buffer
memories 314 or 316, and passed to the display RAM after
decompression of the subsequently occurring B fields (via
multiplexers 320 and 322). In this way the transmitted field
sequence is reordered to the normal field sequence.
1 S Nominally the multiplexer 322 is conditioned to pass image
data from the multiplexer 320 to the display RAM 318.
During intervals when error concealment is required, the
multiplexer 322 is conditioned to pass image data from a field
memory 324. Error concealment is provided according to the
2 0 aforedescribed option three. As each field of data is provided
by the multiplexer 320 an interpolated field of image data is
generated from the current field and stored in the field
memory 324 for substitution in whole or in part in the next
occurring field. During start-up or channel changes, fields of
2 S data are alternately applied to the display RAM from the
multiplexer 320 and the field memory 324.
Compressed video data from the transport
processor 43 is applied to a buffer memory 300. This data is
accessed by the decompression controller 302 wherein header
3 0 data is extracted to program the controller 302. The variable
length codewords corresponding to DCT coefficients are
extracted and applied to a variable length decoder (VLD) 308
and the variable length codewords corresponding to motion
vectors are applied to the variable length decoder (VLD) 306.
3 S The VLD 308 contains apparatus for performing variable
CA 02360556 2001-10-25
19
length decoding, and inverse run length decoding as
appropriate under the control of the controller 302. Decoded
data from the VLD 308 are applied to an inverse DCT circuit
310 which includes circuitry to inverse quantize the
S respective DCT coefficients and to convert the coefficients to a
matrix of pixel data. The pixel data is then coupled to one
input of an adder 312, the output of which is coupled to the
multiplexer 320 and the buffer memories 314 and 316.
The VLD 306 includes circuitry to decode the
variable length encoded motion vectors under the control of
the controller 302. Decoded motion vectors are applied to a
motion compensated predictor 304. Responsive to the motion
vectors the predictor accesses corresponding blocks of pixels
stored in one (forward) or both (forward and backward) of
the buffer memories 314 and 316. The predictor provides a
block data (from one of the buffer memories) or an
interpolated block of data (derived from respective blocks
from both buffer memories) to a second input of the adder
312.
2 0 Decompression is performed as follows. If a field
of input video data is intraframe encoded there are no motion
vectors and the decoded DCT coefficients correspond to blocks
of pixel values. Thus for intraframe encoded data the
predictor 304 applies a zero value to the adder 312 and the
2 S decoded DCT coefficients are passed unaltered by the adder
312, to the one of the buffer memories 314 and 316 not
storing the last decompressed P field, for decoding
subsequent motion compensated frames (B or P).
If a field of input data corresponds to a forward
3 0 motion compensated P field, the decoded DCT coefficients
correspond to residues or differences between the present e.g.
even field and the lastmost occurring even I or P field. The
predictor 304 responsive to the decoded motion vectors
accesses the corresponding block of I or P field data stored in
3 5 either buffer memory 314 or 316 and provides this block of
CA 02360556 2001-10-25
data to the adder wherein the block of residues are added to
the corresponding block of pixel data provided by the
predictor 304. The sums generated by the adder 312
correspond to the pixel values for the respective blocks of the
S P field, which pixel values are applied to the one of buffer
memories 314 and 316 not storing the I or P field of pixel
data utilized to generate' the predicted pixel data.
For bidirectionally encoded (B) fields the
operation is similar, except that predicted values are accessed
10 from the stored I and P pixel data stored in both buffer
memories 314 and 316 depending upon whether the
respective motion vectors are forward or backward vectors or
both. The generated B field pixel values are applied via the
multiplexes 320 to update the display RAM 318, but are not
15 stored in either of the buffer memories 314 and 316, as B
field data is not utilized for generating other fields of picture
data.
During start-up and channel changes the image
latency may be shortened with insignificant image errors.
20 Assume that field 10 (FIGURE 1C) is the first occurring I field
after start-up or channel change. In order to decompress the
subsequent B fields 6 and 8 a decompressed P field 4 is
required, which of course is not available. However it is
generally assumed that successive fields/frames of image
2 5 data are significantly redundant. Thus decompressed I field
10 should be similar to decompressed P field 4 and may be
substituted therefor. This is accomplished simply by writing
the first decompressed I field to both of the memories 314
and 316. Thereafter B fields 6 and 8 may be decompressed
3 0 and displayed.
The controller 302 is programmed to cycle the
particular processing elements according to the particular
sequence of received odd and even fields. It is also
programmed with start-up and channel change sequences
3 5 which are initiated responsive to control signals provided on
CA 02360556 2001-10-25
21
the control bus CB from the system controller. During start-
up, display is inhibited until the reception of the first I field.
Its type (odd/even) is determined from the header data
provided in the received signal, and responsive to the type
data, only the odd or even fields are provided to the
multiplexer 320 for a predetermined number of field
intervals. During this period the multiplexer 322 is
conditioned to switch at the field rate to provide fields of real
image data provided by the multiplexer 320 interleaved with
interpolated fields of data provided by the field memory 324.
During channel changes the system may be conditioned to
operate in similar fashion, or rather than inhibiting display
until reception of an I field, may provide a frozen display of
data stored in the display RAM from the lastmost received
channel.
The interpolator 323 has been described as a
spatial interpolator which averages lines within a particular
field. It should be appreciated that the interpolator 323 may
be arranged to generate temporally averaged data from the
2 0 current field and a prior field of the same type (odd or even).
Concealment of intermittent data loss, indicated
by the error indications provided by the transport processor
43 may be implemented by simply conditioning the system to
enter the channel change mode of operation. Concealment of
2 5 this type will introduce jerkiness and other display artifacts.
However if the data lost or corrupted occurs in a B field, the B
fields may be substituted with interpolated fields without
significant disturbance to the displayed image.
Refer to FIGURE 6 which illustrates the coding
3 0 hierarchy of the compressed signal. At the top level the
compressed signal is provided as groups of fields GOFi (odd or
even), each of which includes at least one I field. Each field is
divided into slices. A slice includes a plurality of
macroblocks. Each macroblock comprises both luminance ,Y,
3 5 and chrominance ,U,V, data. This data is arranged in blocks
CA 02360556 2001-10-25
22
with each block comprising information from an 8 X 8 array
matrix of image pixels. As such each macro block includes
information from 16 adjacent field lines of the encoded image.
Where appropriate the macroblocks also include motion
vectors and other indicia required for decoding.
For purposes of discussion assume that the
transport packetizer 20 applies error detection codes to the
compressed data on a slice basis so that transmission errors
may be detected limited to at least respective slices. Given
1 0 this data configuration, localized error concealment may be
performed in the receiver on a slice basis. Note however, that
if an error occurs in a slice of an I or P field, this error may
propagate through the remainder of the GOF. Thus when
error concealment is performed, attention must be paid to the
type of field in which the error occurred.
In the receiver the system controller 42 may
include an error memory for storing error indications of
respective slices of respective fields, which error indications
are provided by the transport processor 43. Consider FIGURE
2 0 7 which illustrates an error memory (included within the
system controller 42) having sufficient capacity to store slice
error data for a GOF. Each column of the field error memory
represents slice error data for a particular field in the GOF.
The type of each respective field is indicated by the letters
2 5 I,B,P at the top of each column. Consider column 1 which
contains error data for an I field of the GOF. A "1" in a
memory location indicates that an error was detected in a
corresponding slice and a "0" indicates absence of errors. An
error has been indicated in slice 3. This error may propagate
3 0 through all the remaining fields in the GOF during
decompression. Thus the error indication must be propagated
in the error memory, if proper error concealment is to be
provided. The error indication propagation is indicated by
the arrow in slice 3 from column 1 to column 9. Similarly
3 5 propagation of error indications for P fields is indicated by
CA 02360556 2001-10-25
23
respective arrows in the memory. (The memory bank was
not fully populated with ones and zeroes to avoid confusion.)
Propagation of error indications in the field error
memory may be accomplished with the use of an additional
S working memory designated I/P. In general terms, the I field
slice error data is loaded in both the I field portion of the
field error memory and in the I/P memory. As B field slice
error data is generated it is ORed with the corresponding slice
error data in the I/P memory, and the result. loaded in the
1 0 respective B field portion of the field error memory. As
subsequent P field slice error data is generated, this data is
ORed with corresponding slice date in the I/P memory, the
results are loaded in the respective P field portion of the field
error memory, and the results are also used to replace
1 5 corresponding slice error data in the I/P memory. In this
fashion both I and P field slice error data is propagated
through respective fields of the field error memory for a GOF.
FIGURE 8 shows a flowchart of the process for
generating the error map. Respective error maps are
2 0 generated for both odd and even groups of fields, however
the flow chart shows the process for generating a map for
only an odd or even GOF. Typically this process is initialized
by entering the steady state decompression cycle. Once
initialized the system waits for a field synchronization pulse
2 5 (600), and then reads the field type data in the header of a
transport packet (601 ). The field type is examined (602) to
determine if it is an I encoded field. If it is an I field, a
counter (607) is reset to zero, and slice error indications are
loaded into the I portion of the field error memory (FEM), and
3 0 is ORed with corresponding slice error data in the I/P
memory. The results of ORing are substituted in the I/P
memory (603). Note that the I/P memory is not loaded with
I field slice error data on the occurrence of an I field because
e.g. two subsequent B fields from the previous GOF are yet to
3 5 be decompressed. I field error data is substituted in the I/P
CA 02360556 2001-10-25
24
memory after reception of these B fields. This is
accomplished by counting (607) field synchronization pulses,
and when the appropriate number of fields (608) have
occurred after the reception of an I field, the slice error data
for that I field is read from the FEM to the I/P memory (609).
If the received field is not an I encoded field it is
checked (604) to determine if it is a P encoded field. If it is a
P encoded field, the slice error data is ORed (605) with
corresponding slice error data in the I/P memory, and the
1 0 results are loaded in both the I/P memory and the
appropriate P field portion of the FEM. If the received field is
not a P encoded field, by default it is a B encoded field. The
slice error data is ORed (606) with the corresponding slice
error data in the I/P memory, and the results are loaded in
1 5 appropriate B field locations of the FEM.
The foregoing process for generating error maps
provides error data which permits concealing errors which
propagate temporally in a group of frames and horizontally
within a slice. However, it should be realized, that due to
2 0 image motion, errors may also propagate vertically within
successive fields. Concealment of vertical error propagation
may be accomplished by vertically spreading the slice error
data. For example, if a slice error occurs in slice 3 of e.g. I
field 1, this error indication may also be included in memory
2 5 error locations corresponding to slices 2 and 4. Typically, for
most images, motion occurs predominantly in the horizontal
direction with only slight motion in the vertical direction.
Thus, vertical error propagation may generally be ignored
with only very slight and insignificant consequent image
3 0 corruption.
The error map data is provided to e.g. the
decompression controller 302 and responsive thereto
conditions the multiplexer 322 to substitute interpolated
slices from the prior field to the display RAM 318, for current
CA 02360556 2001-10-25
slices of decompressed image data which have been indicated
to contain errors.
FIGURE 9 illustrates a flow chart of an exemplary
start-up/channel change sequence for the system controller
5 42. During start-up or channel changes flags are generated.
These flags are examined (300) to determine whether start-
up or a channel change is to be effected. If the start-up mode
is requested, display of image data is inhibited and a field
counter (303) is disabled (301 ). Alternatively if a channel
10 change is requested, the display apparatus is conditioned to
repeatedly display image data contained in the display RAM,
and the field counter (303) is disabled. The system waits
(302) for the next field synchronization pulse, and when it
occurs, transport header data is examined (305). The system
15 waits for the occurrence of an I encoded field (306), which
field is examined for its odd/even field type (307). At the
occurrence of an I field the counter (303) is enabled. If the I
field is an even field it is decompressed (308), but not written
to the display RAM until the reception of a predetermined
2 0 number, N, of fields (309). The number N is established to
conform to the delay interval between reception of an I field
and the time required to provide continuous decompressed
fields. After reception of N fields, the decompressed data is
written to the display RAM, and the display mode is enabled
2 S (312). In addition interpolated odd fields, generated from the
decompressed even fields are written to the display RAM.
This mode continues until M fields have been received (31 S),
at which time the steady state decompression mode (316) is
activated. The number M is selected as that number of fields
3 0 after the occurrence of an even (odd) field that valid
decompressed odd (even) field data will be available. For
example, with respect to FIGURE 1C, the number M will be
equal to N+9 fields.
CA 02360556 2001-10-25
26
Similarly if the first detected I encoded field is
odd (307), a similar process (310, 311, 314, 315, 316) is
followed.
The system follows a similar sequence if a channel
change is requested with the only difference being that the
display is not inhibited. At control points (312) or (314),
since the display has not been inhibited, when new data is
written to the display RAM, image change is automatic.
