The present invention deals with a method for encoding
video data streams with a view to enhancing the
reliability and the quality of reception of these
streams in a wireless digital network, governed by a
5 communication protocol for example of WIFI type. The
streams concerned are sent in multicast mode and
originate from at least one transmitter broadcasting
them to receiving devices, in this case situated within
a relatively restricted perimeter, such that they can
10 operate with a communication protocol as mentioned,
designed to operate in a more local environment. The
receiving devices are then provided with at least one
software application making it possible to process and
view the information received.
15
More particularly, the video data effected by the
invention are based on groups of pictures (GOP)
conventionally used in video coding (in particular in
the MPEG standards for the compression, the
20 decompression, the processing and video and audio
coding) to define the order of arrangement of the
images with internal coding and with predictive coding
allowing the generation of a visible image in a video
stream. The streams processed in the context of the
25 invention also comprise equally conventional
transmission error correction means, the correction
being done in this case by redundancy and being
performed via algorithms known as AL-FEC (application
level forward error correction) algorithms implementing
30 a matrix function.
Although the method of the invention can have many
practical applications, since it very generally
optimizes the transmission of streams of pictures in
35 real and/or quasi-real time, one of the applications
envisaged is the virtually instantaneous broadcasting
of events of sports meeting or show type, within the
perimeter of an enclosed space where said event takes
2
place. The quality of the video streams transmitted to
the client devices primarily allow the fluid and good
quality viewing of the video pictures, but also make it
possible to enrich the viewing through processes
5 capable of conferring an added value on the streams
received, for example a repetition and therefore a
reviewing of certain sequences, if necessary in slow
motion or even with a zoom effect. Such processes would
be almost meaningless if the quality of the video
10 streams was not assured.
The video streams are usually transmitted in the form
of successive information packets which obey a
predetermined subdivision for the sending thereof,
15 these packets very generally comprising the pictures to
be transmitted, encoded for the transmission, and
additional information which, in case of reception
problems, is used for reconstruction algorithms to try
to restore the initial pictures.
20
As such, the method of the invention applies to video
streams composed in practice of successive data packets
G, each packet G comprising, as is known, a first set M
containing native video data to be transmitted (the
25 pictures of the video stream encoded) and a second set
containing redundancy data obtained by a suitable
algorithm. According to the invention, the first set of
data is composed from all the data M of at least one
group of pictures (GOP).
30
The subdivision of the information packets to be
transmitted in video streams therefore relies,
innovatively according to the invention, on particular
groups of pictures developed and defined in the context
35 of standards organizing the architecture of the video
data packets to be transmitted. The computer system
charged with sending the video streams, and
consequently the software driving same, consequently
3
perform a subdividing of the information which is based
on these groups of pictures, or GOP, formatted in a way
that is standard in the video coding world, identifying
a specific architecture based on particular coded
5 pictures. The data M may come under a single group of
pictures GOP, of which they then contain all the data,
or two or more groups of pictures GOP, with this same
condition of including all their data in one and the
same packet.
10
Still according to the invention, the size of all the
pictures of the group of pictures is added to the data
M of the native video stream, in an area juxtaposing
each picture, which, for example occupies two data
15 bytes per picture.
Furthermore, according to one possibility, an index
table containing the absolute position of each picture
of the group of the pictures can be added to the data
20 M, said table being able to be located in an area for
example situated at the start of the series of coded
pictures to be transmitted.
In practice, according to the invention, each packet G
25 is then organized as K first data blocks of
predetermined size L and N redundancy blocks of size L,
K being equal to M/L, the next integer value
immediately higher than the result of the division M/L
being assigned to K if M/L is not an integer number.
30 The portion of block corresponding to the difference
(KxL – M), which exists only if M/L is not an integer
number, is then filled by data identified as being
stuffing data, for example zeros.
35 Preferably, according to the invention, for each of the
blocks K and N, the information is added, comprising:
- the number of the group of pictures (GOP),
incremented for each new group of pictures;
4
- the value of K, incremented from 1 to K for each
block K;
- the value of N, beginning at K+1 and ending at
N+K;
5 - the value of L; and
- the number of the block (from 1 to K+N).
The length of each of the blocks K and N then becomes
L+ n bytes that the coding of the above information
10 requires. The existence of all these additional data,
combined with the subdivision into blocks K + N of the
native video information, itself based on a standard of
video coding (one or more GOP), makes it possible to
considerably enhance the control of the transmission of
15 the video streams. The reconstruction operations to be
implemented given the assumption of a corrupted
transmission of the video streams are clearly
facilitated by the novel architecture proposed.
20 Moreover, the invention relates also to a video stream
organized according to the method applied above and
comprising a succession of coded data packets G each
composed of a first set of native video data to be
transmitted and of a second set of error correction
25 redundant data, characterized in that the first set
contains all the data of at least one group of pictures
(GOP).
This type of video stream, structured according to the
30 method of the invention, makes it possible to guarantee
a reception in a quality that is sufficient first of
all for the viewing of the streams transmitted to be
simply comfortable, then also for any subsequent
processes applied to the signals transmitted to be able
35 to be based on an information platform sufficient for
the performance thereof.
5
The invention in reality culminates in organizing the
information, that is to say the native video streams to
be sent, so as to allow an optimization of the response
of the system to defects in transmission of the video
5 signal, based on a sufficiently structured and
information-rich architecture.
It will now be described in more detail, with reference
to the attached figures, representing an exemplary
10 implementation of the method of the invention showing
different possible steps for the generation of
optimized video streams, for which:
- figure 1 represents a diagram schematically
representing the organization of a group of
15 pictures as used in the context of the invention;
- figure 2 illustrates, for a same data packet, the
addition to the native video stream of the size of
each coded picture of the group of pictures;
- figure 3 shows the size M of the picture packet
20 when an index table is added to the data;
- figure 4 represents the subdivision into K blocks
of length L of the data packet M, with the
possible addition of stuffing when K x L > M;
- figure 5 illustrates the addition of N blocks due
25 to the encoding by redundancy (FEC); and
- figure 6 finally shows the addition of additional
parameters enriching the information to be
transmitted and facilitating the subsequent
implementation of the data recovery, if
30 appropriate.
Referring to figure 1, the first set of data, forming
the native video stream to be transmitted before
encoding of the transmission error correction function,
35 is composed from all the data of a group of pictures.
This data organization structure is well known in the
field of video coding, since it is in particular
defined by the MPEG standards. These groups of
6
pictures, or GOP, define the nature and the order in
which the pictures resulting from several particular
types of codings are arranged. The grouping of these
pictures in fact forms a particular series, this series
5 being repeated periodically to constitute the encoding
of a video stream. There are thus pictures with
internal coding I and pictures with predictive coding,
the latter being able to be of two types: the pictures
P with predictive coding based on a past image, and the
10 pictures B with bidirectional predictive coding. The
pictures P contain information on difference, by motion
prediction, with a picture I or a past image P. The
pictures B contain information on difference with the
past and future pictures I or P within a group of
15 pictures.
For information, the pictures I or P can be used as
reference pictures, although this is not generally the
case with the pictures B. The visible pictures are, in
20 short, generated from coded pictures contained in a
group of pictures, which is also expressed as a number
of coded pictures, which are constructed at the time of
encoding so as to guarantee the video stream transfer
rate.
25
As appears in particular in figure 1, the group of
pictures N (GOP N) begins with a reference picture IN,
followed by pictures P and B in a particular order
which defines said group of pictures. The video stream
30 to be transmitted comprises this group of pictures N,
followed by a group of pictures N+1 of which the
initial picture IN+1 is the new reference picture,
preceded by a group of pictures N-1 and so on. The set
of these groups of pictures GOPN forms the encoded video
35 stream. For the present invention, the basis of the
subdivision of the video data packets to be sent
consists of these groups of pictures. Figure 1
7
represents only one thereof, but there can be two or
more thereof in the video data packets M.
Figure 2 illustrates another feature of the method of
5 the invention, which enhances the coding of the video
stream by adding the size of all the pictures I, P, B
of the group of pictures in an area juxtaposing each of
these pictures. The extra addition of an index table,
appearing at the start of the video data packet in
10 figure 3, further pads the encoded video information,
the whole forming a first encoded set of size M.
This length or size M is the starting point of a
calculation targeting a subdivision into K blocks of
15 length L (even in figure 4), L being a constant and K
an integer number. When the result of the division of M
by L is not an integer, the immediately higher integer
value is then assigned to K. In most cases, K x L > M.
Given this hypothesis, as represented in figure 4, a
20 stuffing is applied to obtain K complete blocks of
length L. This simply involves adding zeros to the
block K, which are identified by the program as
stuffing value.
25 Hitherto, the subject has been video coding, that is to
say the coding of pictures aiming for the transmission
of the video stream and ultimately allowing the
generation of a visible image after decoding. Figures 5
and 6 add the concept of correction of any transmission
30 errors. As mentioned previously, error correction is
performed in this case by redundancy, by using an
algorithm that is known per se of AL-FEC (application
layer forward error correction) type. The error
correction encoding implements a matrix function which
35 culminates in adding N extra data blocks, also of
length L. These are the blocks K+1 to K+N appearing in
these figures 5 and 6.
8
According to a final operation specific to the
invention, aiming to further optimize the processing
possibilities, information is added for each of the
blocks K and N, comprising in particular but not
5 exclusively:
- the number of the group, incremented for each new
group of pictures (GOP);
- the value of K, incremented from 1 to K for each
block K;
10 - the value of N, beginning at K+1 and ending at
N+K;
- the value of L; and
- the number of the block (from 1 to K+N).
15 This complementary information in practice better
organizes the information, that is to say the set of
video streams to be sent, including the redundancy
blocks, so as to allow an optimization of the response
of the system to video signal transmission defects. It
20 in fact provides data on the blocks taken individually,
in order to better trace the errors and locate them in
the data packets.
The invention is of course not limited to the examples
25 described and explained with reference to the figures,
but it encompasses the variants and versions which fall
within the scope of the claims.

The present invention deals with a method for encoding
video data streams with a view to enhancing the
reliability and the quality of reception of these
streams in a wireless digital network, governed by a
5 communication protocol for example of WIFI type. The
streams concerned are sent in multicast mode and
originate from at least one transmitter broadcasting
them to receiving devices, in this case situated within
a relatively restricted perimeter, such that they can
10 operate with a communication protocol as mentioned,
designed to operate in a more local environment. The
receiving devices are then provided with at least one
software application making it possible to process and
view the information received.
15
More particularly, the video data effected by the
invention are based on groups of pictures (GOP)
conventionally used in video coding (in particular in
the MPEG standards for the compression, the
20 decompression, the processing and video and audio
coding) to define the order of arrangement of the
images with internal coding and with predictive coding
allowing the generation of a visible image in a video
stream. The streams processed in the context of the
25 invention also comprise equally conventional
transmission error correction means, the correction
being done in this case by redundancy and being
performed via algorithms known as AL-FEC (application
level forward error correction) algorithms implementing
30 a matrix function.
Although the method of the invention can have many
practical applications, since it very generally
optimizes the transmission of streams of pictures in
35 real and/or quasi-real time, one of the applications
envisaged is the virtually instantaneous broadcasting
of events of sports meeting or show type, within the
perimeter of an enclosed space where said event takes
2
place. The quality of the video streams transmitted to
the client devices primarily allow the fluid and good
quality viewing of the video pictures, but also make it
possible to enrich the viewing through processes
5 capable of conferring an added value on the streams
received, for example a repetition and therefore a
reviewing of certain sequences, if necessary in slow
motion or even with a zoom effect. Such processes would
be almost meaningless if the quality of the video
10 streams was not assured.
The video streams are usually transmitted in the form
of successive information packets which obey a
predetermined subdivision for the sending thereof,
15 these packets very generally comprising the pictures to
be transmitted, encoded for the transmission, and
additional information which, in case of reception
problems, is used for reconstruction algorithms to try
to restore the initial pictures.
20
As such, the method of the invention applies to video
streams composed in practice of successive data packets
G, each packet G comprising, as is known, a first set M
containing native video data to be transmitted (the
25 pictures of the video stream encoded) and a second set
containing redundancy data obtained by a suitable
algorithm. According to the invention, the first set of
data is composed from all the data M of at least one
group of pictures (GOP).
30
The subdivision of the information packets to be
transmitted in video streams therefore relies,
innovatively according to the invention, on particular
groups of pictures developed and defined in the context
35 of standards organizing the architecture of the video
data packets to be transmitted. The computer system
charged with sending the video streams, and
consequently the software driving same, consequently
3
perform a subdividing of the information which is based
on these groups of pictures, or GOP, formatted in a way
that is standard in the video coding world, identifying
a specific architecture based on particular coded
5 pictures. The data M may come under a single group of
pictures GOP, of which they then contain all the data,
or two or more groups of pictures GOP, with this same
condition of including all their data in one and the
same packet.
10
Still according to the invention, the size of all the
pictures of the group of pictures is added to the data
M of the native video stream, in an area juxtaposing
each picture, which, for example occupies two data
15 bytes per picture.
Furthermore, according to one possibility, an index
table containing the absolute position of each picture
of the group of the pictures can be added to the data
20 M, said table being able to be located in an area for
example situated at the start of the series of coded
pictures to be transmitted.
In practice, according to the invention, each packet G
25 is then organized as K first data blocks of
predetermined size L and N redundancy blocks of size L,
K being equal to M/L, the next integer value
immediately higher than the result of the division M/L
being assigned to K if M/L is not an integer number.
30 The portion of block corresponding to the difference
(KxL – M), which exists only if M/L is not an integer
number, is then filled by data identified as being
stuffing data, for example zeros.
35 Preferably, according to the invention, for each of the
blocks K and N, the information is added, comprising:
- the number of the group of pictures (GOP),
incremented for each new group of pictures;
4
- the value of K, incremented from 1 to K for each
block K;
- the value of N, beginning at K+1 and ending at
N+K;
5 - the value of L; and
- the number of the block (from 1 to K+N).
The length of each of the blocks K and N then becomes
L+ n bytes that the coding of the above information
10 requires. The existence of all these additional data,
combined with the subdivision into blocks K + N of the
native video information, itself based on a standard of
video coding (one or more GOP), makes it possible to
considerably enhance the control of the transmission of
15 the video streams. The reconstruction operations to be
implemented given the assumption of a corrupted
transmission of the video streams are clearly
facilitated by the novel architecture proposed.
20 Moreover, the invention relates also to a video stream
organized according to the method applied above and
comprising a succession of coded data packets G each
composed of a first set of native video data to be
transmitted and of a second set of error correction
25 redundant data, characterized in that the first set
contains all the data of at least one group of pictures
(GOP).
This type of video stream, structured according to the
30 method of the invention, makes it possible to guarantee
a reception in a quality that is sufficient first of
all for the viewing of the streams transmitted to be
simply comfortable, then also for any subsequent
processes applied to the signals transmitted to be able
35 to be based on an information platform sufficient for
the performance thereof.
5
The invention in reality culminates in organizing the
information, that is to say the native video streams to
be sent, so as to allow an optimization of the response
of the system to defects in transmission of the video
5 signal, based on a sufficiently structured and
information-rich architecture.
It will now be described in more detail, with reference
to the attached figures, representing an exemplary
10 implementation of the method of the invention showing
different possible steps for the generation of
optimized video streams, for which:
- figure 1 represents a diagram schematically
representing the organization of a group of
15 pictures as used in the context of the invention;
- figure 2 illustrates, for a same data packet, the
addition to the native video stream of the size of
each coded picture of the group of pictures;
- figure 3 shows the size M of the picture packet
20 when an index table is added to the data;
- figure 4 represents the subdivision into K blocks
of length L of the data packet M, with the
possible addition of stuffing when K x L > M;
- figure 5 illustrates the addition of N blocks due
25 to the encoding by redundancy (FEC); and
- figure 6 finally shows the addition of additional
parameters enriching the information to be
transmitted and facilitating the subsequent
implementation of the data recovery, if
30 appropriate.
Referring to figure 1, the first set of data, forming
the native video stream to be transmitted before
encoding of the transmission error correction function,
35 is composed from all the data of a group of pictures.
This data organization structure is well known in the
field of video coding, since it is in particular
defined by the MPEG standards. These groups of
6
pictures, or GOP, define the nature and the order in
which the pictures resulting from several particular
types of codings are arranged. The grouping of these
pictures in fact forms a particular series, this series
5 being repeated periodically to constitute the encoding
of a video stream. There are thus pictures with
internal coding I and pictures with predictive coding,
the latter being able to be of two types: the pictures
P with predictive coding based on a past image, and the
10 pictures B with bidirectional predictive coding. The
pictures P contain information on difference, by motion
prediction, with a picture I or a past image P. The
pictures B contain information on difference with the
past and future pictures I or P within a group of
15 pictures.
For information, the pictures I or P can be used as
reference pictures, although this is not generally the
case with the pictures B. The visible pictures are, in
20 short, generated from coded pictures contained in a
group of pictures, which is also expressed as a number
of coded pictures, which are constructed at the time of
encoding so as to guarantee the video stream transfer
rate.
25
As appears in particular in figure 1, the group of
pictures N (GOP N) begins with a reference picture IN,
followed by pictures P and B in a particular order
which defines said group of pictures. The video stream
30 to be transmitted comprises this group of pictures N,
followed by a group of pictures N+1 of which the
initial picture IN+1 is the new reference picture,
preceded by a group of pictures N-1 and so on. The set
of these groups of pictures GOPN forms the encoded video
35 stream. For the present invention, the basis of the
subdivision of the video data packets to be sent
consists of these groups of pictures. Figure 1
7
represents only one thereof, but there can be two or
more thereof in the video data packets M.
Figure 2 illustrates another feature of the method of
5 the invention, which enhances the coding of the video
stream by adding the size of all the pictures I, P, B
of the group of pictures in an area juxtaposing each of
these pictures. The extra addition of an index table,
appearing at the start of the video data packet in
10 figure 3, further pads the encoded video information,
the whole forming a first encoded set of size M.
This length or size M is the starting point of a
calculation targeting a subdivision into K blocks of
15 length L (even in figure 4), L being a constant and K
an integer number. When the result of the division of M
by L is not an integer, the immediately higher integer
value is then assigned to K. In most cases, K x L > M.
Given this hypothesis, as represented in figure 4, a
20 stuffing is applied to obtain K complete blocks of
length L. This simply involves adding zeros to the
block K, which are identified by the program as
stuffing value.
25 Hitherto, the subject has been video coding, that is to
say the coding of pictures aiming for the transmission
of the video stream and ultimately allowing the
generation of a visible image after decoding. Figures 5
and 6 add the concept of correction of any transmission
30 errors. As mentioned previously, error correction is
performed in this case by redundancy, by using an
algorithm that is known per se of AL-FEC (application
layer forward error correction) type. The error
correction encoding implements a matrix function which
35 culminates in adding N extra data blocks, also of
length L. These are the blocks K+1 to K+N appearing in
these figures 5 and 6.
8
According to a final operation specific to the
invention, aiming to further optimize the processing
possibilities, information is added for each of the
blocks K and N, comprising in particular but not
5 exclusively:
- the number of the group, incremented for each new
group of pictures (GOP);
- the value of K, incremented from 1 to K for each
block K;
10 - the value of N, beginning at K+1 and ending at
N+K;
- the value of L; and
- the number of the block (from 1 to K+N).
15 This complementary information in practice better
organizes the information, that is to say the set of
video streams to be sent, including the redundancy
blocks, so as to allow an optimization of the response
of the system to video signal transmission defects. It
20 in fact provides data on the blocks taken individually,
in order to better trace the errors and locate them in
the data packets.
The invention is of course not limited to the examples
25 described and explained with reference to the figures,
but it encompasses the variants and versions which fall
within the scope of the claims.
9
I/We Claim:
1. A method for encoding video data streams based on
groups of pictures (GOP) used in video coding to
5 define the order of arrangement of the images with
internal coding and with predictive coding
allowing the generation of a visible image,
including means for correcting transmission errors
by redundancy, for transmission over a wireless
10 network governed by a local data communication
protocol, said video stream being composed of
successive data packets G, each packet G
comprising a first set M containing the video data
to be transmitted, and a second set containing
15 redundancy data obtained by a redundancy
algorithm, characterized in that the first set of
data M is composed from all the data of at least
one group of pictures (GOP).
20 2. The method for encoding video data streams as
claimed in the preceding claim, characterized in
that the size of all the pictures of the group of
pictures is added to the data M, in an area
juxtaposing each image.
25
3. The method for encoding video data streams as
claimed in one of the preceding claims,
characterized in that an index table containing
the absolute position of each picture of the group
30 of pictures is added to the data M.
4. The method for encoding video data streams as
claimed in one of the preceding claims,
characterized in that each packet G is organized
35 as K first data blocks of predetermined size L and
N redundancy blocks of size L, K being equal to
M/L, the next integer value immediately higher
10
than the result of the division M/L being assigned
to K if M/L is not an integer number.
5. The method for encoding video data streams as
5 claimed in the preceding claim, characterized in
that the portion of block corresponding to the
difference (KxL – M), which exists if M/L is not
an integer number, is filled by data identified as
being stuffing data, for example zeros.
10
6. The method for encoding video data streams as
claimed in one of claims 4 and 5, characterized in
that, for each of the blocks K and N, the
information is added, comprising:
15 - the number of the group of pictures (GOP),
incremented for each new group of pictures;
- the value of K, incremented from 1 to K for
each block K;
- the value of N, beginning at K+1 and ending at
20 N+K;
- the value of L; and
- the number of the block (from 1 to K+N).
7. Video streams organized according to the method of
25 the preceding claims and comprising a succession
of data packets G each composed of a first set of
native video data to be transmitted and of a
second set of error correction redundant data,
characterized in that the first set contains all
30 the data of at least one group of pictures (GOP).

I/We Claim:
1. A method for encoding video data streams based on
groups of pictures (GOP) used in video coding to
5 define the order of arrangement of the images with
internal coding and with predictive coding
allowing the generation of a visible image,
including means for correcting transmission errors
by redundancy, for transmission over a wireless
10 network governed by a local data communication
protocol, said video stream being composed of
successive data packets G, each packet G
comprising a first set M containing the video data
to be transmitted, and a second set containing
15 redundancy data obtained by a redundancy
algorithm, characterized in that the first set of
data M is composed from all the data of at least
one group of pictures (GOP).
20 2. The method for encoding video data streams as
claimed in the preceding claim, characterized in
that the size of all the pictures of the group of
pictures is added to the data M, in an area
juxtaposing each image.
25
3. The method for encoding video data streams as
claimed in one of the preceding claims,
characterized in that an index table containing
the absolute position of each picture of the group
30 of pictures is added to the data M.
4. The method for encoding video data streams as
claimed in one of the preceding claims,
characterized in that each packet G is organized
35 as K first data blocks of predetermined size L and
N redundancy blocks of size L, K being equal to
M/L, the next integer value immediately higher
10
than the result of the division M/L being assigned
to K if M/L is not an integer number.
5. The method for encoding video data streams as
5 claimed in the preceding claim, characterized in
that the portion of block corresponding to the
difference (KxL – M), which exists if M/L is not
an integer number, is filled by data identified as
being stuffing data, for example zeros.
10
6. The method for encoding video data streams as
claimed in one of claims 4 and 5, characterized in
that, for each of the blocks K and N, the
information is added, comprising:
15 - the number of the group of pictures (GOP),
incremented for each new group of pictures;
- the value of K, incremented from 1 to K for
each block K;
- the value of N, beginning at K+1 and ending at
20 N+K;
- the value of L; and
- the number of the block (from 1 to K+N).
7. Video streams organized according to the method of
25 the preceding claims and comprising a succession
of data packets G each composed of a first set of
native video data to be transmitted and of a
second set of error correction redundant data,
characterized in that the first set contains all
30 the data of at least one group of pictures (GOP).