FR2930106A1 - Pre-encoded image sequence i.e. video, transmitting method for e.g. mobile telephone, involves selecting coding units such that sum of rates associated to selected units is equal to average rate, and transmitting images from units - Google Patents

Abstract

The method involves estimating an average rate corresponding to a pass band of a communication network over a current period, and identifying a set of images associated to the current period. A set of coding units of the set of images is selected such that a sum of transmission rates associated to the selected coding units is equal to the estimated average rate, where the images are coded along the coding units respectively at the transmission rates. The set of images is transmitted from the selected coding units during the current period. Independent claims are also included for the following: (1) a device for transmitting a sequence of pre-encoded images along a hierarchical coding, between a server and a client on a communication network (2) a server of a communication network for transmitting a sequence of pre-encoded images along a hierarchical coding (3) a computer program comprising instructions to perform a method for transmitting a sequence of pre-encoded images along a hierarchical coding, between a server and a client on a communication network.

Description

The present invention relates to a method of transmission over a communication network of a sequence of pre-encoded images. Correlatively, it also relates to a transmission device on a communication network of a sequence of pre-encoded images.

In general, the present invention relates to the field of streaming video over a communication network between a server and a client. It deals in particular with the transmission of video data taking into account the variations of available bandwidth on the communication network. It applies more particularly to the transmission of a pre-encoded image according to a hierarchical coding, each image being encoded according to several coding units associated respectively with different transmission rates.

This type of coding is notably implemented by the SVC standard (acronym for "Scalable Video Coding"). The SVC standard is an extension of the H.264 compression standard. This SVC extension notably makes it possible to encode each image 25 according to coding layers having different levels of spatial resolution and / or different temporal frequencies. Thus, from a single data stream encoded according to the SVC standard, it is possible to extract sub-data streams corresponding to lower spatial resolutions and / or lower temporal frequencies. As part of the transmission of a sequence of pre-encoded images (corresponding to a broadcast of video also called "video streaming") between a server and a client of a communication network, the SVC standard presents important advantages when it comes to adapting to the conditions of the communication network, and in particular to take into account the variation of the available bandwidth on each transmission period.

However, when the image sequences have been pre-encoded without significant constraints of the hierarchy level, that is to say that the various available coding layers have very large transmission rate deviations, it is difficult to to adapt to the bandwidth constraints of the communication network.

When an estimate of an average rate corresponding to the bandwidth of the communication network over a transmission period is implemented, the most immediate solution consists in broadcasting coding units of the images associated with the transmission period, of which the cost related to the transmission period is strictly less than or equal to the estimated average flow rate. However, when a bandwidth control algorithm of the AIMD (Additive Increase / Multiplicative / Decrease) type is used, the loss of a packet results in a halving of the estimated bandwidth of such a bandwidth. so that the estimate of the bandwidth is very detrimental for the quality of the transmitted image sequence. It has been proposed new bandwidth estimation algorithms taking into account a distortion rate model in order to optimize the overall quality of the broadcast video signal (J.'s Media-and TCP-Friendly congestion control for scalable video streams). Yan, K. Katrinis, M. May, and B. Plattner, IEEE Transaction On Multimedia, 8 (2) April 2006). This algorithm makes it possible to obtain a finer estimate of the available bandwidth on the communication network. However, it does not solve the quality differences observed during the decoding of the image sequence, related to the threshold effects associated with the average rate available on the communication network for each transmission period.

A mechanism for adapting the quality of the transmission of a video signal has also been proposed in the article "Layered quality adaptation for internet video streaming" by Mr. Handley, R. Rejaie and D. Estrin, in ACM SIGCOMM , 1999. This mechanism consists in adding or removing hierarchical layers of the video data stream in order to achieve a long-term quality adaptation while using a bandwidth calculation algorithm to react rapidly to the fluctuations of the communication network. Poor adaptation is absorbed by using buffers at the client of the communication network.

This technique is attractive for smoothing bandwidth while maintaining consistent video quality over time, but requires the use of buffers. The object of the present invention is to improve the state of the art and to propose a transmission method making it possible to make the maximum use of the available bandwidth of a communication network on which a sequence of pre-encoded images is broadcast. of the SVC type. The present invention aims for this purpose according to a first aspect a method of transmission on a communication network of a sequence of images pre-encoded according to a hierarchical coding, each image being encoded according to several coding units associated respectively with data rates of different transmission, comprising the following steps: estimating an average rate corresponding to the bandwidth of the communication network over a current period; identifying a set of images associated with said current period. According to the invention, the transmission method further comprises the following steps: selecting a set of coding units of the set of images so that the sum of the bit rates associated with the selected coding units is substantially equal to the estimated average flow rate; and transmitting the set of selected coding units during the current period.

Thus, the transmission method according to the invention makes it possible to take into account the estimated average available bit rate from the bandwidth of the communication network and to adapt the bit rate of the transmitted image sequence to this average bit rate.

The transmission method makes it possible to respect a flow constraint during the transmission period during which this flow constraint is valid, by selecting the coding units that can be transmitted in this transmission period so as to respect, in particular average, the available flow over this period of time.

This solution thus makes it possible to make the most of the authorized bandwidth for the transmission of a pre-encoded picture sequence according to a hierarchical coding. In practice, at the selection stage, the set of coding units comprises a first subset of coding units associated with a first subset of images, the bit rate associated with a coding unit of the first subset of images respectively. first subassembly being strictly less than the average bit rate, and a second subset of coding units associated with a second subset of pictures, the bit rate associated with a coding unit of the second subset being respectively greater than at the average flow. Thus, the transmission method makes it possible to respect, on average, the average flow rate constraint available over the current period, while entailing in a few images a temporary overshoot of the average flow rate and on other images a flow rate below this average flow rate; Note that this transmission method is particularly simple to implement since it does not require the addition or change of the encoder or the hierarchical decoder. Preferably, at the selecting step, the selected coding unit associated with an image of the first subset has a rate as close as possible to the average rate of the current period. The transmission method thus makes it possible to select the coding units whose rate best corresponds to the estimated average rate of the communication network for the current period, without however exceeding it, only a second subset of images being transmitted with one unit. coding belonging to a layer of higher hierarchy, that is to say higher flow, while ensuring the current period the respect of the estimated average flow. According to a practical embodiment, the selection step comprises the following sub-steps: determining the transmission position of a second subset of images during a preceding period; and selecting the second subset of images of the current period at the beginning of the current period when the second subset of images is transmitted at the end of the preceding period and at the end of the current period when the second subset of images is transmitted at the beginning of the previous period.

Thus, the choice of the coding units of the second subset of images is carried out in such a way as to ensure good continuity at the border of the transmission periods in order to obtain a smoothing in the transmission of the coded data. This type of transmission is particularly well suited when the image sequence is of the IPPP type, that is to say when each image has about the same importance in the video data stream to be transmitted. According to another embodiment, the images of the second subset of images are reference images used in the prediction of subsequent images of said sequence of pre-encoded images.

This embodiment is particularly well suited when the sequence of pre-encoded images to be transmitted comprises predictive images and hierarchical images. According to an advantageous characteristic of the invention, in the step of estimating an average bit rate, the estimation of the bandwidth is carried out on the basis of a reception report notably comprising information on the loss rate, a round-trip time and variance of transit times.

The use of this reception report to estimate the bandwidth of a communication network is particularly easy to implement in the context of a communication network using an RTCP protocol accompanying the RTP protocol (acronym for Real Time Protocof ') In order to maintain the coherence of the estimated average flow rate and thus to avoid unwanted fluctuations, in the step of estimating the average flow rate, the loss of the coding units selected in said selection step and whose associated bit rate is greater than the estimated average bit rate is ignored in a loss rate calculation, so failure to temporarily respect the estimated average bit rate for some images does not affect the estimate of available bandwidth on In practice, the transmission method applies to the real-time transmission of a sequence of pre-encoded images between a server and a network. communication network client. It finds particular interest when the sequence of pre-encoded images is encoded according to coding layers having different spatial resolutions and / or temporal resolutions, and for example according to the SVC standard. Correlatively, the present invention relates to a transmission device on a communication network of a sequence of pre-encoded images according to a hierarchical coding, each image being encoded according to several coding units associated respectively with different transmission rates, comprising means for estimating an average rate corresponding to the bandwidth of the communication network over a current period; means for identifying a set of images associated with said current period. According to the invention, the transmission device further comprises: means for selecting a set of coding units of the set of images so that the sum of the bit rates associated with the selected coding units is substantially equal to the estimated average flow rate; and means for transmitting the set of coding units selected during the current period. This transmission device has characteristics and advantages similar to those of the transmission method described above. According to another aspect, the present invention aims to protect a server of a communication network adapted to transmit a sequence of pre-encoded images according to a hierarchical coding, comprising means adapted to implement the transmission method according to the present invention. 'invention. This server has characteristics and advantages similar to those described above in relation to the transmission method according to the invention. The present invention relates to a computer program loadable in a computer system, the program comprising instructions for implementing the transmission method according to the invention, when the program is loaded and executed by a computer system. Other features and advantages of the invention will become apparent in the description below. In the accompanying drawings, given by way of nonlimiting examples: FIG. 1 is an example of a sequence of pre-encoded images according to a hierarchical coding implementing the SVC standard; FIG. 2 is a block diagram illustrating a communication network adapted to implement the transmission method according to the present invention; FIG. 3 is an algorithm illustrating the transmission method 30 according to one embodiment of the invention; FIG. 4 is an algorithm detailing the selection step of FIG. 3 according to one embodiment of the invention; and FIGS. 5a and 5b illustrate the implementation of the selection step as detailed in FIG. 4. A particular video stream on which the compliant transmission method will be described will first be described with reference to FIG. The invention can be advantageously applied. FIG. 1 illustrates a sequence of pre-encoded images according to a hierarchical coding that can be implemented by the SVC standard. In particular, the sequence of pre-encoded images is encoded according to coding layers having in particular spatial resolutions and different temporal resolutions. In this particular example, from a video sequence of 720x576 format, originally composed of 60 frames per second, it is possible to encode, and therefore to decode, a lower spatial resolution whose size is here by example equal to 360x288. This lower image size typically corresponds to a screen resolution of an electronic personal assistant also known by the acronym PDA (Persona / Digital Assistant). Similarly, it is possible to code, and therefore to decode, for the same spatial resolution, different dyadic temporal representations. In this example, the basic temporal representation of 60 Hz can be declined in lower temporal representation of 30 or 15 Hz. As well illustrated in Figure 1, these different temporal representations are available for any spatial resolution. The number of temporal representations and spatial resolutions may be chosen by the user at the time of encoding. Thus, each image is encoded according to several coding layers of variable spatial and temporal resolutions, whose associated transmission rate is thus also variable. Thus, the SVC standard makes it possible to assign a variable bit rate to each image of a sequence and thus to recreate a hierarchy in terms of image quality at the decoding level.

Note that the example illustrated in Figure 1 shows ratios of two between each spatial resolution (720x576 for a television screen resolution, 360x288 for a screen resolution of an electronic personal assistant and 180x144 for a resolution of screen of a mobile phone).

However, the SVC standard is not limited to multiple spatial resolutions of 2. In addition, the aspect ratio of each image can also be changed. The coding units, consisting of compressed data sets, thus correspond to the different images that are transmitted over a communication network, and for example from a server device to a client device on which the decoding of the image is carried out. view. As a nonlimiting example, for each coding layer of an image characterized by its resolution and its frequency, the associated bit rate for the transmission of this coding layer over a communication network is given below. The encoding units are the building blocks of the layers and are associated with at least one layer. Each coding unit has a cost. For each transmission period, the cost of each coding unit reported over the transmission period makes it possible to associate a transmission rate with each coding unit. Layer Resolution Frequency in Hz Rate in kbits / s 0 180x144 1.8750 23.00 1 180x144 3.7500 36.21 2 180x144 7.5000 54.21 3 180x144 15.0000 73.93 4 360x288 1.8750 118.00 5 360x288 3.7500 166.21 6 360x288 7.5000 228.21 7 360x288 15.0000 302.25 8 360x288 30.0000 363.25 9 720x576 1.8750 481.00 10 720x576 3.7500 660.32 11 720x576 7.5000 883.32 12 720x576 15.0000 1139.09 13 720x576 30.0000 1411.09 14 720x576 60.0000 1642.09 It is noted that according to the SVC standard, the encoded video data stream is not coded according to quality hierarchy levels: transmission is fixed when a temporal resolution and a spatial resolution are selected. It is thus noted that, thanks to the SVC standard, it is possible, from a data stream, to extract a sub-stream, whose transmission rate is lower, corresponding to reduced image sizes thus requiring a lower decoding power.

This sub-stream is thus more easily decodable by a low capacity device such as a mobile phone. However, it is noted that the layers (consisting of coding units) exhibit significant spatial / temporal resolution in terms of flow rate, so that it is difficult to finely adapt to the constraints of the bandwidth of the communication network on which the video data stream is to be transmitted. We will now describe with reference to Figure 2, a client / server application of a communication network adapted to implement a transmission method according to the invention.

This is to transmit a sequence of images pre-encoded in real time between a server and a client of a communication network. Thus, in the context of the invention, the communication apparatus, called server 200, has an encoded digital signal that it wishes to transmit to the remote communication device 300, called client, via a communication network 201.

The structure of the server 200 and the client 300 is substantially identical and comprises in particular a central unit 202, 302 (called the CPU in the drawing) which executes the instructions relating to a program. In particular, the central unit 202 of the server 200 is adapted to load and execute the program containing instructions for the implementation of the transmission method which will be described below. The program instructions are stored in a ROM 203, 303 or other storage elements. When powering on, the programs stored in a non-volatile memory, for example the read-only memory (ROM), are transferred to RAM (acronym for the English term Random Access Memory) 204, 304 which will then contain the executable code of the invention as well as registers for storing the variables necessary for the implementation of the program.

Alternatively, the program or programs could be received to be stored identically to that previously described through the communication network 201. More generally, information storage means, readable by a computer or by a microprocessor, integrated or not to the device, possibly totally or partially removable, stores at the server 200 a program implementing the data transmission method according to the invention. The communication bus 210, 310 allows the communication between the various elements included in the server 200 or the client 300. The representation of the bus 210, 310 is not limiting and in particular the central unit 202, 302 is able to communicate instructions to any element of the server 200 or client 300, directly or through another element. The server 200 and the client 300 further comprise a hard disk 206, 306 used by the central unit 202, 302 in a conventional manner over the communication bus 210, 310. The server 200 and the client 300 furthermore comprise an interface communication 205, 305 connected to the communication network 201 which is capable of transmitting digital data in the context of the implementation of the invention. The server 200 thus integrates all the means of a device for transmitting a pre-encoded video signal adapted to implement the transmission method which will be described below. In particular, this server incorporates means for estimating an average bit rate corresponding to the bandwidth of the communication network 201 over a given period of time, means for identifying a set of images to be transmitted over this current period, means for selecting a set of coding units associated with this set of images and transmission means at the communication interface 205 of the sets of selected coding units to the client 300. The client 300 also comprises a decoding program of a video signal adapted to be implemented by the processor 302.

This client device further comprises interface elements with the user, including a keyboard 311 and a screen 312 for viewing the video signal after transmission and decoding. It will be noted that the transmission method is implemented entirely on the server 200.

In the application described below, the communication protocols implemented on the communication network 201 are known protocols and used in video broadcasting applications (also called video streaming in English). These RTP, RTCP and RTSP protocols are implemented between the server 200 and the client 300. In particular, the server 200 and the client 300 use the Real Time Streaming Protocol (RTSP) to control the display of a video sequence . With this protocol, the client 300 can specify the video to be transmitted to a remote server by specifying its URI (acronym for the English term Uniform Resource Identifier). The characteristics of the video to be transmitted are also transmitted from the server 200 to the client 300 in order to know the main video parameters at the decoder.

Other useful information can also be communicated between the server and the client using the RTSP protocol. On the other hand, an underlying protocol RTP (Real Time Protocol) is also used to transmit in real time coded data of the video sequence stored on the server 200. An associated protocol RTCP is used to exchange control messages. This RTCP protocol is a protocol that accompanies the RTP protocol to measure its performance. These two protocols include useful fields for reconstituting the stream, identifying the transported information and thus controlling the arrival of the packets transmitted at the level of the client 300. In this context, a periodic transmission of control packets, in the form of reports, is performed in an RTP session. In particular, an RR (Receiver Report) may be transmitted. This RR report includes receiver statistics for the transmitter on the quality of transmission, including loss rate information, round-trip time (RTT), and transit time variance also known as jitter. This information is important because it makes it possible to adjust the redundancy according to the losses and to adapt the buffer 20 at the level of the client 300 as a function of the jitter. This reception report RR also makes it possible to estimate the bandwidth of the communication network. Referring now to FIG. 3, the transmission method in accordance with the invention implemented on a server 200 as illustrated above and communicating via a communication network 201 implementing the recalled transmission protocols will now be described with reference to FIG. above. This transmission method firstly comprises a reception step E30 of an RR report in which, as indicated above, loss rate, round trip time and jitter information are given.

An estimation step E31 of the average rate corresponding to the bandwidth of the communication network is implemented from the reception report RR. Note that this estimate of the average rate is performed for a given period so that the estimated bandwidth is valid at least for this period of time. In video transmission applications, RTT round-trip times can be as high as several hundred milliseconds. Thus, the estimation of an average rate is carried out over a current period equal for example to 200 ms. In order to estimate the average rate for the current period, it is possible to use known methods such as the TRFC algorithm which makes it possible to evaluate the available bit rate as a function of the state of the network. This algorithm is detailed in particular in the article Equation-based congestion control for unicast applications, S. Floyd, M. Handley, J. Padhye and J. Widmer, SIGCOMM 2000, August 2000. One can also use the algorithm AIMD which allows to increase by linear growth the available bandwidth if there is no congestion of the communication network and decrease by a multiplicative decrease the bandwidth when congestion arrives. It will be noted that the current period on which the estimation of an average flow rate is carried out may be variable according to the reception frequency of the RR reports. For each current period on which an average rate has been estimated, the transmission method then comprises an identification step E32 of a set of images associated with this current period. In practice, for a period of the order of 200 ms and a video sequence of 60 Hz, the estimate of the bandwidth over the current period is valid for a dozen images.

This identification step E32 amounts to identifying the coding units associated with a set of images to be transmitted over the current period.

The transmission method then comprises a selection step E33 of a set of coding units. This selection step E33 makes it possible to select for the current period the most appropriate hierarchy levels.

According to the invention, this selection is made in such a way that the sum of the bit rates associated with the selected coding units is substantially equal to the estimated average bit rate over the current period. It is thus a question of respecting on average the constraint of flow on the interval of time considered.

This solution achieves the maximum bandwidth allowed on the current period. This behavior of the transmission algorithm implemented on the server is of the best effort type. During the identification step E32 of a set of images associated with the current period, this can be done by detecting the temporal frequency of the video sequence and marking the start of the data transmission. It is thus easy to identify the coding units that correspond to a time interval by analyzing the different data of the pre-encoded stream. Once the coding units selected at the end of the selection step E33, these selected coding units are transmitted during the current period to a transmission step E34. In practice, this transmission step E34 consists in encapsulating the selected coding units in RTP packets for transmission to the client 300. A test step E35 makes it possible to know whether the end of the video sequence to be transmitted has been reached. . If not, step E30 is repeated and the following steps E31 to E34 over a subsequent period. Otherwise, the transmission process is terminated. An exemplary embodiment of the selection step E33 described above with reference to FIG. 3 will now be described with reference to FIGS. 4, 5a and 5b. In practice, in order to respect on average the available flow over a period of time Ti, one selects for a first subset of images m; a first subset of coding units such that the bit rate associated with each coding unit of this first subset is strictly less than the average bit rate DE; of the considered period Ti and for a second subset of images n; a second subset of coding units such that the rate associated with each coding unit of this second subset is strictly greater than the average rate DE ;. Thus, the set of images Ni identified for the period Ti is divided into two subsets of images m ;, n ;. The sum of the bit rates associated with the coding units selected for the two subsets m ;, n; is equal to the estimated average flow rate DE; over the given period Ti, although for the images of the second subset of images n; the transmission rate of the selected coding units temporarily exceeds the average rate DE; over the period Ti. FIGS. 5a and 5b illustrate this selection of images in two cases of evolution of the bandwidth available on the communication network. Thus, Figure 5a shows an illustration of the selection of the images when the flow rate estimate is increasing and Figure 5b illustrates the selection of the images when the flow rate estimate is decreasing. Thus, for each period TI, T2, T3, we estimate the average available flow DE1, DE2, DE3 materialized by a solid line in the figures. These flow rates DE; correspond to the result of the flow estimation made for example by the TFRC algorithm for a time interval Ti. Two levels of hierarchy and their associated bit rate DCo and DCA have also been illustrated for the image sequence to be transmitted. By way of example, these two levels of hierarchy can be representative of the layer 8 (DCo equal to 363.25 kbits / s) and of the layer 13 (DCA equal to 1411.09 kbits / s) illustrated in the table described. previously indicating different levels of possible hierarchies for a given video sequence. Note that between these two layers, there is a significant difference in transmission rate.

Thus, when the average value of estimated flow DE; is large but less than the DCA rate (for example, when DE is 1000 kbit / s), a traditional transmission method would only transmit the hierarchy levels associated with the lower DCo rates, so that the available bandwidth would be largely under -Used. The vertical lines in FIGS. 5a and 5b respectively correspond to the flow rates associated with the images Ni identified to be transmitted during the time interval Ti. It will be appreciated that the coding units of an image have been shown to have an equivalent transmission rate during a time interval T1. This flow rate may however be different from one image to another. The selection step E33 consists in selecting for a first subset of images m; coding units whose transmission rate is lower than the average rate DE; valued. In practice, the coding unit associated with an image of the first subset m is selected; having a flow rate as close as possible to the average flow rate DE; of the relevant period Ti. However, it can be seen that these coding units associated here with the DCo bit rate do not use all the available bandwidth. The present invention allows by the selection of the second subset n; with higher resolution coding units to optimize the use of the available bandwidth. Thus, a second subset of images n; complementary to the first subset of images m; transmitted with coding units having a transmission rate DC ;, greater than the average available rate DE; of the second subset n; being determined in such a way that the sum of the rates of the selected coding units transmitted during the period Ti remains, however, on average equal to the value of the average rate DE ;. More precisely, the sum DIS; transmission rates of the selected images in the second subset n; can be calculated as: DIS; = n; x (DC1 - DCo) This DIS flow; corresponds to the complementary transmission rate transmitted because of the selection for the second subset of images n; higher level coding units. This DIS flow; must not exceed the available supplementary budget BT; during the period Ti, which is equal to: BT; = Ni x (DE; - DCo) This supplementary budget BT; available therefore corresponds to the difference between the lower level layer DCo and the average available flow rate DE; over the period Ti, available bit rate that would not be used and that would be lost in a conventional process of video data transmission depending on the available bandwidth. Thus, in the selection step E33, the following formula must be respected: DIS; BT; that is, n; x (DCA - DCo) Ni x (DE, - DCo) In practice, the number of images n; is selected by choosing n; the largest possible, satisfying the inequality above. Furthermore, during the selection step E33, it is advantageous to be able to choose the higher level coding units, associated with the second subset of images n ;, so as to ensure good continuity at the time boundary. Ti in order to achieve a smoothing in sending coded data. In this regard, FIG. 4 illustrates a first embodiment of this optimized selection step.

In practice, a determination step E331 makes it possible to know the transmission position of the second subset of images n; _1 during a preceding period T; _I. This determination step E331 makes it possible to determine whether the higher hierarchical level coding units transmitted to the previous period T; _1 were selected at the beginning or at the end of the period T; _I. If in the preceding period T; _I, the selection of the second subset of images n; _1 is transmitted at the end of the preceding period T; _I, the selection step E332 is adapted to select the second subset of images n; of the current period Ti at the beginning of this current period Ti. On the other hand, if the second subset of images n; _1 is transmitted at the beginning of the preceding period T; _I, a selection step E333 is adapted to select the second subset of images n; of the current period Ti at the end of the current period Ti. This type of selection makes it possible, as illustrated in FIGS. 5a and 5b, to avoid breaks in the boundaries of the intervals Ti in the transmission of the images of the video sequence.

This embodiment is particularly well suited to the transmission of IPPP type image sequences, that is to say in the case where the video consists of a sequence of images each having about the same importance in the bit stream of data to be transmitted and decoded. According to another embodiment, the images of the second subset of images n ;, for which higher hierarchical level coding units are transmitted, are reference images used in the prediction of subsequent images of the sequence of pre-encoded images. This embodiment is particularly well suited when the sequence of pre-encoded images comprises both predictive images and hierarchical images. This is particularly the case when the video signal is organized in the form of a GOP (acronym for the English term Group Of Pictures). In the case of a GOP sequence, the sequence of pre-encoded images comprises both predictive images P and hierarchical images B. The images which are used as reference images in the prediction of the following images are then preferred. that is to say here preferably the predictive images P. For example, for a video configuration IP B3 B2 B3 B1 B3 B2 B3 P configuration, the P-type images will be preferred over the B1 images, which themselves will be privileged compared to B2 images, themselves privileged compared to B3 images.

This selection mode thus makes it possible to send to higher hierarchical coding levels, and thus to higher spatial and / or temporal resolutions, the important images of the data stream used as reference images during decoding by the client.

It will thus be noted that the transmission method used makes it possible to optimize the use of the available bandwidth on the communication network. The present invention thus makes it possible to broadcast a stream of compressed video data making maximum use of the average flow setpoint calculated by a congestion control algorithm. This solution is simple and relies on the analysis of possible flows over a period of time. Moreover, this solution has the advantage of not requiring a change of a syntax element of a data stream encoded according to the SVC standard. Furthermore, it is important to maintain the coherence of the estimated average rate, in order to avoid untimely fluctuations in the transmission of video data. In this respect, in the step of estimating the average bit rate, the loss of the coding units selected in the selection step, that is to say the coding units associated with the second subset of images. n ;, and whose associated rate is greater than the estimated average rate DE ;, is ignored in a calculation of the loss rate implemented by a bandwidth estimation algorithm. In fact, the fact of not temporarily respecting the estimated bandwidth over a period 25 Ti could have consequences on the behavior of the network and cause congestion, thus leading to an estimate of the available flow rate that is erroneous and evaluated downwards. In order to avoid this phenomenon, the loss rate of the evaluated network does not take into account the losses affecting the transmitted higher layer coding units, their purpose being only to improve the quality of the transmitted data stream.

In practice, it will be possible to report the loss of data packets containing higher level coding units by means of a signal NACK (Non Acknowledgment) recognized as such during a calculation of the probability of loss for the evaluation of the bandwidth available.

The transmission method thus described makes it possible to transmit encoded video data streams without strong constraint at the level of the hierarchical coding while fine-tuning the bandwidth constraints of the network. Of course, many modifications can be made to the examples described above without departing from the scope of the invention.

Claims (17)

REVENDICATIONS1. A method of transmitting on a communication network (201) a sequence of pre-encoded images according to a hierarchical coding, each image being encoded according to several coding units associated respectively with different transmission rates, comprising the following steps: estimating (E31) an average rate (DE;) corresponding to the bandwidth of the communication network over a current period (Ti); identification (E32) of a set of images (Ni) associated with said current period (Ti); characterized in that it further comprises the following steps: selecting (E33) a set of coding units of said set of images (Ni) so that the sum of the bit rates associated with said selected coding units is substantially equal to said estimated average flow rate (DE;); and transmitting (E34) said set of selected coding units during the current period (Ti). 2. Transmission method according to claim 1, characterized in that in said selection step (E33), the set of coding units comprises a first subset of coding units associated with a first subset. set of images (m;), the associated bit rate (DCo) respectively to a coding unit of said first subset being strictly less than said average bit rate (DE;), and a second subset of coding units associated with a second subset of images (n;), the associated rate (DC1) respectively to a coding unit of said second subset being strictly greater than said average rate (DE;). Transmission method according to claim 2, characterized in that in said selection step (E33), the selected coding unit associated with an image of said first subset (m;) has a bit rate (DCo) as close as possible to said average flow rate (DE;) of the current period (Ti). 4. Transmission method according to one of claims 1 to 3, characterized in that said selecting step (E33) comprises the following sub-steps: - determination (E331) of the transmission position of a second sub-step; set of images (n; _I) during a previous period (T; _I); and selecting (E332, E333) said second subset of images (m;) of the current period (Ti) at the beginning of said current period (Ti) when said second subset of images (n; _I) is transmitted at the end of the preceding period (T; _I) and at the end of said current period (Ti) when said second subset of images (n; _I) is transmitted at the beginning of the preceding period (T; _1). 5. Transmission method according to claim 4, characterized in that the image sequence is of the IPPP type. 6. Transmission method according to one of claims 1 to 3, characterized in that the images of said second subset of images (n;) are reference images used in the prediction of subsequent images of said sequence pre-encoded images. 20 7. Transmission method according to claim 6, said sequence of pre-encoded images comprising predictive images and hierarchical images, characterized in that said second subset of images preferably comprises predictive images. 8. Transmission method according to one of claims 1 to 7, characterized in that in the estimation step (E31) of an average rate (DE;), the estimated bandwidth is performed from a reception report (RR) including information on the loss rate, a round trip time and the variance of transit time. 9. Transmission method according to claim 8, characterized in that in the estimation step (E31) of the average rate (DE;), the loss of the coding units selected in said selection step (E33) and whose associated rate (DCA) is greater than the estimated average rate (DE;) is ignored in a calculation of the loss rate. 10. Transmission method according to one of claims 1 to 9, characterized in that said sequence of pre-encoded images is transmitted in real time between a server (200) and a client (300) of the communication network ( 201). 11. Transmission method according to one of claims 1 to 10, characterized in that said sequence of pre-encoded images is encoded according to coding units having different spatial resolutions and / or temporal resolutions. 12. Device for transmitting on a communication network (201) a sequence of pre-encoded images according to a hierarchical coding, each image being encoded according to several coding units associated respectively with different transmission rates, comprising: estimation means (202, 203, 204) of an average bit rate (DE;) corresponding to the bandwidth of the communication network over a current period (Ti); means for identifying (202, 203, 204) a set of images (Ni) associated with said current period (Ti); characterized in that it further comprises: - selection means (202, 203, 204) of a set of coding units of said set of pictures (Ni) such that the sum of the bit rates associated with said units selected coding is substantially equal to said estimated average rate; and - transmission means (202, 203, 204) of said set of selected coding units during the current period (Ti). Transmission device according to claim 12, characterized in that said selection means (202, 203, 204) comprise: means for determining the transmission position of a second subset of images (n I) during a previous period (T; _i); and- means for selecting said second subset of images (n;) at the current period (Ti) at the beginning of said current period (Ti) when said second subset of images (n; _I) is transmitted at the end of the preceding period (T; _I) and at the end of said current period (Ti) when said second subset of images (n; _I) is transmitted at the beginning of the preceding period (T; _1). A server of a communication network adapted to transmit a sequence of pre-encoded images according to a hierarchical coding, characterized in that it comprises means (202, 203, 204) adapted to implement the transmission method. according to one of claims 1 to 11. 15. Server according to claim 14, characterized in that said sequence of pre-encoded images is transmitted in real time between said server (200) and a client (300) of the communication network (201). 16. Server according to one of claims 14 or 15, characterized in that said pre-encoded image sequence is encoded according to coding units having different spatial resolutions and / or temporal resolutions. 17. A computer program loadable into a computer system, said program comprising instructions for implementing the transmission method according to one of claims 1 to 11, when the program is loaded and executed by a computer system.