JP5706144B2 - Method and apparatus for transmitting scalable video according to priority - Google Patents

Description

  The present invention relates to a scalable video coding (SVC) technical field, and more particularly, to a method and apparatus for transmitting scalable video according to priority.

  SVC is an efficient video encoding method having a scalability function. SVC can encode a video sequence into a bitstream consisting of one base layer and multiple enhancement layers with different scalability. Here, the base layer can restore an image sequence having the lowest resolution and quality, while the enhancement layer can restore an image sequence having a higher resolution and quality. According to the three scalability schemes provided in SVC: time domain, spatial domain, and quality domain, a terminal can generate a bit stream satisfying a certain target bit rate, frame rate, and spatial resolution from a received SVC bit stream. Extraction is guaranteed to be able to restore images of video with different bit rates, frame rates or resolutions. Therefore, SVC is an algorithm suitable for network video transmission and is expected to be applied more and more widely. With the development of broadband networks and the increasing demand for video content from people, in the SVC based video system, different terminals require faster frame rate or higher resolution or better quality for the same video content there is a possibility. Due to the scalability of SVC, it is determined that the requirements of different terminals can be satisfied with only one encoding on the source.

  In a network having a constant packet loss rate or being congested, different terminals may not be able to receive all bitstreams within a certain time. However, since a video having a lower resolution, frame rate, or quality is obtained by reconstructing from a lower-layer bitstream, SVC causes a certain amount of distortion when receiving a partial bitstream. It is possible to restore the image of the video it has. In this case, how to guarantee the video service quality of different terminal users is the biggest challenge in the near future for video transmission. Here, the order of the transmission bit stream and the format of the data packet of the bit stream are two key issues in the video transmission method.

  When transmitting an SVC layered bit stream in a network, the source information is encoded with the maximum resolution and bit rate. In the transmission process, when there is no packet loss caused by bandwidth change, different scalable video receivers can extract bit streams of different resolutions according to their reception capabilities. However, when packet loss occurs, the scalable video receiver may not be able to receive data packets sufficient to reconstruct a video image that satisfies the resolution requirement. In this case, the quality of the video sequence restored on the receiving side of the scalable video is degraded. That is, the quality of the video image reconstructed on the scalable video receiving side is affected by the packet loss rate of the channel in the network. As the packet loss rate increases, the quality of the reconstructed image clearly decreases. Therefore, an effective method for processing the SVC layered bit stream is required so as to reduce the influence on the quality of the reconstructed image from the channel packet loss rate.

  In order to solve the above technical problem, in the present invention, a method and apparatus for transmitting scalable video according to priority so as to improve the quality of a video image reconstructed from a bitstream of scalable video. Is provided.

According to an embodiment of the present invention, a method of transmitting scalable video according to priority includes a priority for a hierarchical bitstream after scalable video coding (SVC) of each picture group (GOP) of the scalable video. Reordering is performed, and the priority ordering means rearranging the priority order in accordance with the importance of the SVC layered bitstream after SVC encoding in units of GOP. for arranging after sorting SVC hierarchical bit stream by performing a package and Raptor coding, via a transmission channel, the bit stream of a scalable video after encoding, to the receiving side of a scalable video, it And performing the priority reordering on the SVC layered bitstream is a minimum resolution feature. All possible nodes on the reordering path from the bitstream having the highest resolution characteristic to the bitstream having the highest resolution characteristic, the constant (s, t, q) characteristic (s, t and q are integers, respectively State nodes corresponding to the resolution characteristics of the spatial domain, temporal domain, and quality domain), and state nodes having the same increased step size are placed on the same stage due to the difference in the increased step sizes of s, t, and q. In some cases, all state nodes are partitioned into different stages, and the increment step size sum is the sum of the increment step sizes of s, t and q, and the bit rate of each state node and its corresponding peak signal pair. Depending on the noise ratio (PSNR),
To calculate the reordering factor for each state node, where g _j represents the j th stage, t _i (g _j ) represents the i th state node in the j th stage, and t (g ₀ ) Represents the initial state node of the initial stage, t (g _K-1 ) represents the destination state node of the destination stage, the destination stage is the stage where the destination state node is located, and the destination state node is the highest resolution characteristic. Where PSNR _{ti (gj)} represents the PSNR of state node t _i (g _j ), R _{ti (gj)} represents the bit rate of state node t _i (g _j ), and PSNR _{t (g0 )} Represents the PSNR of the initial state node t (g ₀ ), R _{t (g0)} represents the bit rate of the initial state node t (g ₀ ), and PSNR _{t (gK−1)} represents the destination state node t (g _K represents PSNR of _-1), R _{t (gK- )} Represents the bit rate of the destination state node _{t (g K-1),} O i (g j) represents a sort factor state node t _i (g _j), K represents the total number of stages, the initial stage The optimal path that maximizes the weighted sum of the reordering factors of each state node is determined from all the paths that can be configured from the state node to the destination stage, and each path from the initial stage on the optimal path to the destination stage is determined. Outputting the SVC layered bitstream according to the order of the state nodes .

Determining the optimal path from all paths that can be configured with state nodes from the initial stage to the destination stage
Calculates the weighted sum of the reordering factors of each state node on all the paths that can be configured with the state nodes from the initial stage to the destination stage, and determines the path with the largest weighting sum of the reordering factors as the optimal path. And including that. Here, S _X represents the weighted sum of the reordering factors of each state node on the x-th path among all the paths that can be configured by the state nodes from the initial stage to the destination stage, and O (g _j ) is The reordering factor of the state node in the j-th stage of the x-th path is represented as t _i (g _j ), where O (g _j ) = O. _i (g _j ), and ω _j represents the weighting factor of the j-th stage (0 <ω _j <1, 0 ≦ j ≦ K−1).

To determine the optimum path from all paths that can be configured with state nodes from the initial stage to the destination stage, the initial state node of the initial stage is selected as a starting point, and the node identifier j is set to 1. A step A0 for setting the weighted sum S ₀ (t) of the rearrangement factors in the initial stage to 0 and initializing the state node set of the optimal path of the initial node to the initial state node;
and
Is used to calculate the weighted sum S _j (t _i ) of the reordering factors of each state node in the jth stage, and the state node set of the optimal path from the initial stage to each state node in the jth stage Step A1 to be recorded and whether j is equal to K-1 or not, if they are equal, the path consisting of each state node in the state node set of the optimal path of the destination state node is the optimal path, and if not equal, And “j = j + 1”, and the process returns to step A1 and step A2.

Here, S _j (t _i ) represents the weighted sum of the reordering factors of the i-th state node t _i in the j-th stage, and S _j-1 (t _k ) represents the state node in the j-th stage. represents the weighted sum of the reordering factors of the _kth state node t _k connected to t _i , ω _j represents the weighting factor of the j th stage, and O _i (g _j ) represents the i th state in the j th stage Represents a reordering factor of node t _i , M _j (t _i ) represents a state node set of an optimal path of i-th state node t _i in the j-th stage, and M _j−1 (t _k ) represents _j- th The state node set of the optimal path of the state node corresponding to the maximum value of the weighted sum of the rearrangement factors among the state nodes connected to the state node t _i in the stage −1.

To determine the optimum path from all the paths that can be configured with the state nodes from the initial stage to the destination stage, the destination state node of the destination stage is selected as the starting point, and the node identifier j is set to K-2. A step a0 for setting the weighted sum T _K-1 (t) of the reordering factors of the destination stage to 0 and initializing the state node set of the optimal path of the destination node to the destination state node;
and
Is used to calculate the weighted sum T _j (t _i ) of the reordering factors from each state node to the destination stage in the jth stage, and the optimal path from each state node to the destination stage in the jth stage Step a1 for recording each state node set and whether j is equal to 0 or not, if they are equal, the path consisting of each state node in the state node set of the optimal path from the initial state node to the destination stage is If not equal, “j = j−1” is set, and step a2 is returned to step a1.

Here, T _j (t _i ) represents the weighted sum of the reordering factors from the i-th state node t _i to the destination stage in the j-th stage, and T _{j + 1} (t _k ) is in the j + 1-th stage. represents a weighted sum of the sort factor k th state node t _k leading to the state node t _i, omega _j represents a weighting factor of a stage of the j, O _i (g _j) is the i-th in the stage of the j of it represents sort factor state node t _i, M _j (t _i) represents the state node set of optimal path to a destination stage from the i-th state node t _i at stage of the _j, M j _{+ 1} ( t _k) denotes the state node set of optimal path from state node corresponding to the maximum value of the weighted sum of the sort factors to the destination stage from each state node leading to state nodes t _i in the j + 1 stage to the destination stage .

The j-th stage weighting factor ω _j is ω ^{j + 1} (0 <ω <1, 0 ≦ j ≦ K−1).

  When there is more than one optimal path, the method according to an embodiment of the present invention increases the value of the weighting factor of each stage, and from among all paths that can be configured with state nodes from the initial stage to the destination stage. Re-determining the optimal path.

  In the Raptor coding, RS coding and interleaving are performed on each source matrix block of each layer after rearrangement and packaging to obtain m data packets, and the order of LT coding is determined. Performing LT coding on the m data packets according to the LT coding order obtained to obtain n data packets.

Determining the order of the LT encoding is
Selecting the value of the order d within the range of. Here, θ is a parameter of an order distribution function, and when the order distribution function is a Robust-Soliton function,
(Constant c> 0, δ is an acceptable decoding failure rate , k is for controlling the form of the order distribution function, and k = 1 if the order distribution function is a Robust-Soliton function ).

Alternatively, determining the degree of the LT coding establishes a correspondence between the optimum range of values in the range and degree d value of m, the values and the correspondence relation of m, the optimum value of degree d A range is determined, and an integer is selected at random within the optimum range of the value of the order d to obtain the order d of the LT encoding, including the range of the value of m and the value of the order d. Establishing the correspondence with the optimal range classifies the range of the value of m, obtains a decoding failure rate δ when the m value is in a different order range by simulation,
To determine, including a call.

In the embodiment of the present invention, an apparatus for transmitting scalable video according to priority is disclosed. The apparatus reorders the priority of SVC coding means for performing scalable video coding (SVC) on each picture group (GOP) of scalable video and the SVC layered bitstream after SVC coding. gastric row, said the priority sorting units of GOP, with respect to SVC hierarchical bit stream after SVC coding, depending on their importance is that sort order of priority priority Reordering means, packaging means for packaging the SVC layered bitstream after rearrangement, and raptor encoding for the SVC layered bitstream after packaging, via the transmission channel , seen containing a Raptor coding means for transmitting to the receiving of a scalable video, wherein the priority sorting means, minimum solutions All possible nodes on the reordering path from the bitstream having the degree characteristic to the bitstream having the highest resolution characteristic, the constant (s, t, q) characteristic (s, t and q are integers; State nodes corresponding to the resolution characteristics of the spatial domain, the temporal domain, and the quality domain, respectively, and state nodes having the same increased step size sum are the same due to the difference in the increased step sizes of s, t, and q. As in the stage, all state nodes are partitioned into different stages, and the increased step size sum is a stage partition module that is the sum of the increased step sizes of s, t, and q, and the bit rate of each state node and By its corresponding peak signal to noise ratio (PSNR), the formula
To calculate the reordering factor for each state node, where g _j represents the j th stage, t _i (g _j ) represents the i th state node in the j th stage, and t (g ₀ ) Represents the initial state node of the initial stage, t (g _K-1 ) represents the destination state node of the destination stage, the destination stage is the stage where the destination state node is located, and the destination state node is the highest resolution characteristic. Where PSNR _{ti (gj)} represents the PSNR of state node t _i (g _j ), R _{ti (gj)} represents the bit rate of state node t _i (g _j ), and PSNR _{t (g0 )} Represents the PSNR of the initial state node t (g ₀ ), R _{t (g0)} represents the bit rate of the initial state node t (g ₀ ), and PSNR _{t (gK−1)} represents the destination state node t (g _K represents PSNR of _-1), R _{t (gK- )} Represents the bit rate of the destination state node _{t (g K-1),} O i (g j) represents a sort factor state node t _i (g _j), K Sorting factor represents the total number of stages A decision module, an optimum path decision module that decides an optimum path that maximizes the weighted sum of the reordering factors of each state node from all the paths that can be configured by the state nodes from the initial stage to the destination stage, and an optimum A rearrangement result output module that outputs an SVC layered bitstream according to the order of each state node from the initial stage to the destination stage on the path.

  The Raptor encoding means performs RS encoding on the source matrix block of each layer after packaging and obtains m data packets, and LT encoding order for determining the order of LT encoding A determination module; and an LT encoding module that performs LT encoding on the m data packets according to the determined LT encoding order to obtain n data packets.

  The method and apparatus for transmitting scalable video according to priority according to the present invention first performs priority reordering on the SVC layered bitstream according to the layering characteristics of the SVC layered bitstream, Next, package and Raptor encoding are performed on the rearranged SVC layered bitstream. Thus, it is possible to solve the problem of optimization selection and transmission of a bit stream of scalable video, and finally improve the quality of an image reconstructed from the bit stream of scalable video.

6 is a flowchart of a method for transmitting scalable video according to priority according to an embodiment of the present invention; 4 is a flowchart of a method for performing priority reordering on an SVC layered bitstream according to an embodiment of the present invention. It is a figure which shows that all the state nodes were divided into the different stages. It is a flowchart of the method of searching all the state nodes according to the order from an initial stage to a destination stage, and obtaining an optimal path. It is a flowchart of the method of searching all the state nodes according to the order from a destination stage to an initial stage, and obtaining an optimal path. It is a figure which shows the SVC layered bit stream after a package. FIG. 6 is a diagram illustrating a raptor encoding process according to an embodiment of the present invention. It is a figure which shows the internal structure of the apparatus based on the Example of this invention which transmits a scalable image | video according to a priority. The figure which shows the rate distortion performance at the time of employ | adopting the default rearrangement method, the priority rearrangement method based on a quality layer, and the priority rearrangement method provided in the Example of this invention with respect to scalable video sequence Foreman. It is. The figure which shows the rate distortion performance at the time of employ | adopting the default rearrangement method, the priority rearrangement method based on a quality layer, and the priority rearrangement method provided in the Example of this invention with respect to scalable video sequence Football. It is. The figure which shows the rate distortion performance at the time of employ | adopting the default rearrangement method, the priority rearrangement method based on a quality layer, and the priority rearrangement method provided in the Example of this invention with respect to the scalable video sequence City. It is. The figure which shows the rate distortion performance at the time of employ | adopting the default rearrangement method, the priority rearrangement method based on a quality layer, and the priority rearrangement method provided in the Example of this invention with respect to scalable video sequence Harbour. It is. Decoding on the receiving side of scalable video when adopting the standard LT coding scheme and the LT coding scheme adaptively selecting the order provided in the embodiment of the present invention with different packet loss rates and the number of source data packets N It is a figure which shows a success rate.

  Conventional H.264. In order to improve the service quality of a video stream by directly applying to the H.264 / SVC video coding standard, an embodiment of the present invention provides a method for transmitting scalable video according to priority. As shown in FIG. 1, the method includes the following steps.

  In step 101, priority ordering is performed on the SVC layered bitstream after each picture group (GOP: Group of Pictures) of the scalable video is SVC-encoded.

  In step 102, the rearranged SVC layered bitstream is packaged and Raptor encoded, and the encoded scalable video bitstream is transmitted to the scalable video receiver via the transmission channel. To do.

  After receiving the encoded bitstream, the scalable video receiving side can obtain an SVC layered bitstream by Raptor decoding and bitstream combination, and then performs SVC decoding to regenerate the video image. To construct.

  In the embodiment of the present invention, according to the layering characteristics of the SVC layered bitstream, first, the priority is rearranged for the SVC layered bitstream, and then the SVC layered bitstream after the rearrangement is performed. Then, package and Raptor encoding are performed. Thus, it is possible to solve the problem of optimization selection and transmission of a bit stream of scalable video, and finally improve the quality of an image reconstructed from the bit stream of scalable video.

  Hereinafter, a method for performing the priority rearrangement on the SVC layered bitstream in step 101 will be described in detail with reference to the drawings.

  In the embodiment of the present invention, the priority rearrangement means that the priority order is rearranged according to the importance of the SVC layered bitstream after SVC encoding in units of GOP. is there. Its purpose is to optimize the quality of the restored image when the receiving side receives a certain number of data packets correctly.

It is assumed that the size of each GOP in the video sequence is N, and after each GOP image is SVC encoded, an L-layer SVC layered bitstream having a bit rate of R can be generated. Here, the bit rate of the i-th layer bit stream is represented by R _i , and the peak signal-to-noise ratio parameter (reflecting the degree of distortion of the reconstructed image corresponding to the bit stream from the first layer to the i-th layer) PSNR: Peak Signal to Noise Ratio) is represented by PSNR _i (1 ≦ i ≦ L). After the SVC layered bit stream of each GOP is rearranged in priority, the bit rate R _i and the corresponding PSNR _i value of the obtained L layer bit stream satisfy the following Equation 1.
[Formula 1]

  It is known that priority reordering can be performed on an SVC layered bitstream by employing a default SVC layered bitstream rearrangement method. Specifically, in the default SVC layered bitstream rearrangement method, rearrangement is performed based on the time domain layer in units of GOPs. Each time domain layer includes a number of spatial domain layers. These spatial domain layers may include one quality domain (eg, SNR) layer that is a base layer and several quality domain layers that are enhancement layers.

  In addition, priority sorting can be performed on the SVC layered bitstream by employing a method for rearranging the SVC layered bitstream based on the quality layer. Specifically, in the SVC layered bitstream rearrangement method based on the quality layer, the image of the i-th frame is encoded into the d layer (for example, the QCIF layer, the CIF layer, and the 4CIF layer). Layer and quality layer, where the resolution characteristic of the quality layer is assumed to be q, and the image bit rate function R (d, i, q) and distortion function D (d, i, q) and a rate-distortion RD curve is created to obtain an optimal permutation order of the quality layer of the image of each frame.

  In the embodiment of the present invention, a new priority reordering method is provided. FIG. 2 illustrates a method for performing priority reordering on an SVC layered bitstream according to an embodiment of the present invention. As shown in FIG. 2, the method includes the following steps.

  In step 201, all possible nodes on the reordering path are defined as state nodes corresponding to certain (s, t, q) characteristics, and due to the difference in the increasing step sizes of s, t and q All state nodes are partitioned into different stages so that state nodes with the same incremental step size are in the same stage.

In this embodiment, for convenience of explanation, the characteristics of the SVC layered bit stream are represented by triples (s, t, q), where s, t, and q are the resolutions of the spatial domain, temporal domain, and quality domain, respectively. Represents a characteristic. Here, the bit stream having the lowest resolution characteristic is represented by (0, 0, 0), and the receiving side of the scalable video can restore the video with the most basic quality after receiving the bit stream. The bit stream with the highest resolution is represented by (s ^* , t ^* , q ^* ), and after receiving the bit stream, the receiver of the scalable video completely restores the video of various resolution quality including the highest resolution. can do. A bit stream having an intermediate resolution is represented by (s, t, q), and after receiving the bit stream, a scalable video receiving side can restore videos of various resolution qualities including the intermediate resolution. .

The goal of the priority rearrangement method according to the present embodiment is from the bit stream having the lowest resolution characteristic to the bit stream having the highest resolution characteristic, that is, the state node (0, 0, 0) (also called the initial state node). To the state node (s ^* , t ^* , q ^* ) (also called the destination state node) looking for the path with the optimal increase in resolution characteristics, and the bitstream recovered at the receiver side has the optimal rate distortion characteristics It is to ensure that it has.

  FIG. 3 is a diagram illustrating that all state nodes are divided into different stages. As shown in FIG. 3, twelve state nodes (state (0, 0, 0) to state (1, 2, 1)) having different resolution characteristics are divided into five stages (stage 0 to stage 4), respectively. The In this case, the task of selecting a path having an increase in the optimal resolution characteristic is to select an appropriate state node from the initial state node state (0, 0, 0) to the destination state node state (1, 2, 1). It was converted to the question of how to construct the optimal path.

  In step 202, a reorder factor for each state node is calculated according to the bit rate of each state node and its corresponding PSNR.

In the above step 202, the rearrangement factor of each node can be calculated by the following formula 2.
[Formula 2]

Here, g _j represents the j-th stage, t _i (g _j ) represents the i-th state node in the j-th stage, t (g ₀ ) represents the initial state node in the initial stage, and t ( g _K-1 ) represents the destination state node of the destination stage,
Represents the PSNR of state node t _i (g _j ),
_Represents the bit rate of the state node t _i (g _j ),
Represents the PSNR of the initial state node t (g ₀ ),
Represents the bit rate of the initial state node t (g ₀ ),
Represents the PSNR of the destination state node t (g _K-1 ),
Represents the bit rate of the destination state node t (g _K-1 ), O _i (g _j ) represents the reordering factor of the state node t _i (g _j ), and K represents the total number of stages.

In the embodiment of the present invention, O _i (g _j ) is used to describe a camber of a path from the j−1 stage to the i th state node of the j th stage. The camber may be considered an increase over the reference quality.

  In step 203, an optimal path that maximizes the weighted sum of the rearrangement factors of each state node is determined from all paths that can be configured by the state nodes from the initial stage to the destination stage.

In this step, the weighted sum of the rearrangement factors of the respective state nodes on all paths that can be constituted by the state nodes from the initial stage to the destination stage can be calculated by the following Equation 3. Then, the path having the maximum weighting sum of the rearrangement factors is set as the optimum path.
[Formula 3]

Here, S _x represents the weighted sum of the reordering factors of each state node on the x-th path among all the paths that can be configured by the state nodes from the initial stage to the destination stage, and O (g _j ) is The state node rearrangement factor in the j-th stage of the x-th path is represented, and ω _j represents the weighting factor of the j-th stage (0 <ω _j <1, 0 ≦ j ≦ K−1).

Preferably, the weighting factor of the g _jth stage may be ω ^{j + 1} (0 <ω <1, 0 ≦ j ≦ K−1). As can be seen from the experiment, a relatively good priority rearrangement result can be obtained for a general video sequence when ω = 0.2.

  In step 204, the SVC layered bit stream is output according to the order of each state node from the initial stage to the destination stage on the optimal path.

  In this step, a state node having the same quality region characteristic in each node on the optimum path can be output as a one-layer SVC layered bit stream.

  Specifically, in the embodiment of the present invention, various methods may be adopted to determine the optimum path from all possible paths composed of state nodes from the initial stage to the destination stage. For example, in order to obtain the optimum path, all state nodes may be searched in the order from the initial stage to the destination stage, or all state nodes may be searched in the order from the destination stage to the initial stage. Good.

  FIG. 4 is a flowchart of a method for obtaining the optimum path by searching all the state nodes in the order from the initial stage to the destination stage. The method mainly includes the following steps.

In step 301, the initial state node of the initial stage is selected as a starting point to initialize the iteration parameters. In this initialization, the node identifier j is set to 1, the weighted sum S ₀ (t) of the rearrangement factor in the initial stage is set to 0, and the state node set of the optimal path of the initial node is initialized to the initial state node. Including doing.

In step 302, the weighted sum S _j (t _i ) of the reordering factors of each state node in the jth stage is calculated, and the state node set of the optimal path from the initial stage to each state node in the jth stage is calculated. Record.

In this step, the weighted sum S _j (t _i ) of the rearrangement factors of the respective state nodes in the j-th stage can be calculated by the following Equation 4.
[Formula 4]

Here, S _j (t _i ) represents the weighted sum of the reordering factors of the i-th state node t _i in the j-th stage, and S _j-1 (t _k ) represents the state node in the j-th stage. represents the weighted sum of the reordering factors of the _kth state node t _k connected to t _i , ω _j represents the weighting factor of the j th stage, and O _i (g _j ) represents the i th state in the j th stage Represents the sort factor of node t _i . As can be seen from Equation 4 above, the weighting sum S _j (t _i ) of the rearrangement factors of the i-th state node t _i in the j-th stage is one in the weighting value of the rearrangement factors of the state node. This is a value obtained by adding the maximum value of the weighted sums of the rearrangement factors of the respective state nodes connected to the state node in the previous stage (that is, the j−1th stage).

Further, the state node set of the optimal path of the i-th state node t _i in the j-th stage can be calculated by the following Equation 5.
[Formula 5]

Here, M _j (t _i ) represents the state node set of the optimal path of the i-th state node t _i in the j-th stage, and M _j−1 (t _k ) represents the state node in the j-th stage. Among the state nodes connected to t _i , the state node set of the optimal path of the state node corresponding to the maximum weighted sum of the reordering factors is represented.

  Step 303 determines whether the jth stage is the destination stage, i.e., determines whether j is equal to K-1, and if so, executes Step 304; Set “j = j + 1” and return to Step 302.

  In step 304, the path consisting of each state node in the state node set of the optimal path of the destination state node is set as the optimal path.

  FIG. 5 is a flowchart of a method for obtaining the optimum path by searching all the state nodes in the order from the destination stage to the initial stage. The method mainly includes the following steps.

In step 401, the destination state node of the destination stage is selected as a starting point to initialize the iteration parameters. In this initialization, the node identifier j is set to K-2, the weighted sum T _K-1 (t) of the destination stage reordering factors is set to 0, and the state node set of the optimal path of the destination node is set to the destination state. Includes initializing the node.

In step 402, the weighted sum T _j (t _i ) of the reordering factors from each state node to the destination stage in the jth stage is calculated, and the state of the optimal path from each state node to the destination stage in the jth stage Record each node set.

In this step, the weighted sum T _j (t _i ) of the rearrangement factors of the respective state nodes in the j-th stage can be calculated by the following Equation 6.
[Formula 6]

Here, T _j (t _i ) represents the weighted sum of the reordering factors from the i-th state node t _i to the destination stage in the j-th stage, and T _{j + 1} (t _k ) is in the j + 1-th stage. Represents the weighted sum of the reordering factors of the _kth state node t _k connected to the state node t _i , ω _j represents the weighting factor of the j th stage, and O _i (g _j ) represents the i th in the j th stage. Represents the reordering factor of the state node t _i of. As can be seen from Equation 6 above, the weighting sum T _j (t _i ) of the rearrangement factors from the i-th state node t _i to the destination stage in the j-th stage is the weight value of the rearrangement factor of the state node. And the maximum value of the weighted sum of the rearrangement factors from each state node connected to the state node to the destination stage in the next stage (ie, the (j + 1) th stage).

Further, the state node set of the optimal path from the i-th state node t _i to the destination stage in the j-th stage can be calculated by the following Equation 7.
[Formula 7]

Here, M _j (t _i ) represents the state node set of the optimal path from the i-th state node t _i to the destination stage in the j-th stage, and M _{j + 1} (t _k ) is in the j + 1-th stage. This represents the state node set of the optimal path from the state node to the destination stage corresponding to the maximum value of the weighted sum of the reordering factors from each state node to the destination stage connected to the state node t _i .

  In step 403, it is determined whether the jth stage is an initial stage, that is, whether j is equal to 0, and if so, execute step 404; = J-1 "and the process returns to step 402.

  In step 404, the path composed of each state node in the state node set of the optimum path from the initial state node to the destination stage is set as the optimum path.

  As should be explained, what has been provided by the method according to FIGS. 4 and 5 is merely a representative example showing the search for the optimal path among all the state nodes. Embodiments of the present invention are not limited to adopting the above two methods.

  Also, it should be explained that if there are more than one optimum paths determined in step 203, all paths that can be configured with state nodes from the initial stage to the destination stage by increasing the value of the weighting factor of each stage. The optimum path can be re-determined. For example, in step 302 of the search process shown in FIG. 4, if it is found in the previous stage that more than one state node corresponds to the maximum weighting sum of the reordering factors, or as shown in FIG. In step 402 of the search process, in the next stage, if it is found that more than one state node corresponds to the maximum weighting sum of the reordering factors, the weighting factor for each stage is increased and step 301 or 401 is performed. Returning to FIG. 4, the search process shown in FIG. 4 or 5 can be re-executed.

  As can be seen from the above, in the method for performing priority reordering on the SVC layered bitstream according to the embodiment of the present invention, layering of different spatial regions, time regions and quality regions in a single GOP is performed. Since the rate distortion performance of the bitstream is fully considered, the priority reordering result becomes more rational, and further the rate distortion performance of the SVC layered bitstream is improved.

  Hereinafter, the package and Raptor encoding method in step 102 will be described in detail with reference to the drawings.

FIG. 6 is a diagram showing the SVC layered bitstream after packaging. As shown in FIG. 6, in the packaging process, first, the SVC hierarchical bit stream of L layers, into a source matrix block of w _i × h _i, respectively. Here, w _i is the width of the source matrix block, h _i is the height of the source matrix block, and satisfies the constraint condition shown in Equation 8 below. The constraint guarantees that after channel decoding, the previous source code is always restored preferentially over the subsequent source code, i.e., the more important information is always restored more preferentially than the less important information. To do. Next, in each source matrix block, Raptor encoding is performed to obtain n data packets.
[Formula 8]

The n data packets are transmitted through a channel having a constant packet loss rate, and reach the scalable video receiving side. Assuming that α (0 ≦ α ≦ n) data packets are received on the receiving side of the scalable video and the number of code elements obtained by the bitstream combination is r, the relationship between α and r Satisfies the constraint condition shown in Equation 9 below.
[Formula 9]

Further, the probability Pi that the scalable video receiving side receives the i-th layer bit stream is a function of α, and it is assumed that the constraint condition shown in Equation 10 below is satisfied.
[Formula 10]

For a given bandwidth and packet loss rate, as can be seen from Equation 10, transmission of a scalable video stream should meet the following goals: That is, when 0 <α ≦ n, the scalable video receiving side maximizes the PSNR of the video image restored from the r code elements. The goal can be achieved by optimizing the w _i in the package parameter. That is, the optimization target function is as shown in the following Expression 11. Currently, in order to maximize the PSNR of the restored video image, the package width w _i of the L-layer SVC layered bitstream can be determined by various search algorithms. Here, the description is omitted.
[Formula 11]

  In addition, in the embodiment of the present invention, Raptor encoding adopts the format of RS + LT, and the encoding process mainly includes the following steps as shown in FIG.

In step 501, RS coding and interleaving are performed on each of the source matrix blocks of each layer after being rearranged and packaged to obtain m data packets. The length of the m data packets is fixed to P _length codes.

That is, in this step, a matrix block (for example, a source matrix having a size w _i × h _i ) having a size (mw _i ) × h _i and having redundant information attached after each source matrix block. the block, forward error correction (FEC: forward error correction) by coding, by padding magnitude of generating a (mw _i) × h _i a is the matrix block) sizes, respectively obtaining L number of matrix blocks is m × h _i. As shown in FIG. 6, since the sum of the heights of the L matrix blocks is P _length , m lengths of which the lengths are fixed to P _length codes by combining the L matrix blocks. Data packets can be obtained.

  In step 502, the LT encoding degree (Degree) d is determined.

In step 503, the m data packets are subjected to LT encoding according to the determined LT encoding order d to obtain n data packets. The length of the n data packets is also fixed to P _length codes.

That is, in this step, β (1 ≦ β ≦ d) data packets are randomly selected from the m data packets, and the corresponding information in the β data packets is selected for each bit. An exclusive OR operation is performed to obtain one new data packet. This is repeated until n data packets whose _length is similarly fixed to P _length codes are obtained.

As can be seen from the Raptor encoding process shown in FIG. 7, the size of the LT encoding order d is a key point for determining the complexity of LT encoding. Also, the decoding success rate on the receiving side of the scalable video is determined by selecting the size of the order d in LT coding. In general LT coding, the value of m is relatively large (10 ³ to 10 ⁴ ). In this case, the value of the order d satisfies dε [1, m]. In the video system according to the present invention, the basic unit of bit stream processing and transmission is GOP, and the size of the GOP bit stream directly affects the range of values of m. In order to guarantee the video playback quality and delay of the terminal, the terminal needs to decode the bit stream of the GOP within a certain time to restore the video image having the minimum distortion. However, if the GOP bitstream is relatively small (m <10 ³ ), the number of data packets whose degree d is 1 decreases as the m value decreases. As a result, the decoding success rate of the LT code gradually decreases, and the video restored at the terminal is severely distorted and cannot be decoded. Therefore, the selection range of dε [1, m] is not suitable for transmission of a scalable video bit stream. For this reason, in the embodiment of the present invention, a method for adaptively selecting the range of the value of the order d is also provided.

In the LT (n, m) algorithm, it is known that the distribution of the order d is a Robust-Soliton distribution. The distribution function is as shown in Equation 12 below.
[Formula 12]

here,
, Constant c> 0,
, Δ is an acceptable decoding failure rate.

As can be seen from the distribution function of order d shown in Equation 11 above, the order d is
The probability of being over 95%, but the order d is
Is less than 5%. Therefore, when adaptively selecting the magnitude of the order d,
Only the range may be considered. That is,
Within this range, the order d may be selected. Here, θ is a parameter of the Robust-Soliton distribution function.

In addition, it is necessary to consider the decoding capability of RS encoding for the selection range of the order d. During LT decoding, it is necessary to ensure that packets that cannot be successfully decoded are within the range of erasure correction of RS encoding. That is, it is necessary to satisfy δ ≦ α. Here, α is the redundancy of RS encoding corresponding to the m value. According to the above criteria, first, the range of the value of m is classified, and then the decoding failure rate when the m value is in a different order range is obtained by simulation. Thus, the optimum range of the value of the order d within the range of different values of m is determined, that is, as shown in Table 1, the correspondence between the range of the value of m and the optimum range of the value of the order d is expressed as follows. Establish.

  In this way, in the above step 502, first, the optimum range of the value of the order d can be determined from the value of m. Next, one integer is randomly selected within the optimum range of the value of the order d, and is set as the order d of LT coding.

  As can be seen from the method of adaptively selecting the range of values of the order d, the selection range of the order d is greatly reduced in the LT encoding process of this embodiment. This greatly reduces the complexity of LT encoding. Further, by selecting the order value within the optimum range of the order d value, the decoding success rate on the receiving side of the scalable video can be further improved.

  In addition to the above-described method for transmitting scalable video according to priority, an embodiment of the present invention also provides an apparatus for transmitting scalable video according to priority. As shown in FIG. 8, the internal configuration mainly includes priority for SVC encoding means for performing SVC encoding for each GOP of scalable video, and for SVC layered bitstream after SVC encoding. Priority encoding means for performing rearrangement, packaging means for performing packaging on the SVC layered bitstream after rearrangement, and Raptor encoding for the SVC layered bitstream after packaging, Raptor encoding means for transmitting to the receiving side of the scalable video via the transmission channel.

  In this embodiment, the priority sorting means includes all possible nodes on the sorting path, and constant (s, t, q) characteristics (s, t, and q are the space region, time, All of the state nodes with the same increased step size in the same stage due to the difference in the increased step sizes of s, t, and q. A state partition module, a reorder factor determination module for calculating a reorder factor of each state node according to a bit rate of each state node and its corresponding PSNR, and a destination from the initial stage Among all paths that can be configured with state nodes up to the stage, the weighted sum of the reordering factors of each state node is maximized An optimal path determination module that determines an optimal path, and a rearrangement result output module that outputs an SVC layered bitstream according to the order of each state node from the initial stage to the destination stage on the optimal path.

  As described above, the rearrangement factor determination module can calculate the rearrangement factor of each node according to Equation 2.

  As described above, the optimal path determination module can determine the optimal path by adopting the method shown in FIGS.

  Further, the Raptor encoding means includes an RS encoding module that performs RS encoding on each of the source matrix blocks of each layer after packaging to obtain m data packets, and LT encoding. LT coding order determination module for determining order d, and LT coding for performing LT encoding on the m data packets according to the determined LT encoding order d to obtain n data packets Module.

  As described above, according to the embodiment of the present invention, an apparatus for transmitting a scalable video according to a priority first has a priority for an SVC layered bitstream according to a layering characteristic of the SVC layered bitstream. Rearrangement is performed, and then package and Raptor encoding are performed on the rearranged SVC layered bitstream. Thus, it is possible to solve the problem of optimization selection and transmission of a bit stream of scalable video, and finally improve the quality of an image reconstructed from the bit stream of scalable video.

  In an apparatus for transmitting scalable video according to priority according to an embodiment of the present invention, when performing priority reordering on an SVC layered bitstream, different spatial regions and time in a single GOP Since the rate distortion performance of the layered bitstream in the region and the quality region is sufficiently considered, the priority reordering result becomes more rational, and further, the rate distortion performance of the SVC layered bitstream is improved. Also, an apparatus for transmitting scalable video according to priority according to an embodiment of the present invention can adaptively select a range of order values of LT encoding when performing LT encoding. As a result, the complexity of LT encoding can be greatly reduced, while the decoding success rate on the receiving side of scalable video can be greatly improved.

Hereinafter, the technical effect of the method according to the embodiment of the present invention will be described in detail by simulation. Table 2 shows the parameters of the simulation model.

  FIGS. 9 to 12 are provided in the embodiment of the present invention for the default reordering method, the priority reordering method based on the quality layer, and four scalable video sequences of Foreman, Football, City, and Harbour. The rate distortion performance when the priority ordering method is adopted is shown respectively. Here, the square curve represents the rate distortion performance of the default sorting method, the rhombus curve represents the rate distortion performance of the priority sorting method based on the quality layer, and the circle curve represents this 4 represents the rate distortion performance of the priority reordering method provided in an embodiment of the invention.

  As can be seen from FIGS. 9 to 12, when the bit rate is relatively low, the rate distortion performance of the priority reordering method provided in the embodiment of the present invention is clearly different from the simulation results of all sequences. It is better than these two methods. On the other hand, when the bit rate is relatively high, the rate distortion performance of the reordering method provided in the embodiments of the present invention is still better than the other two algorithms.

Table 3 shows average PSNRs obtained by performing priority reordering on video sequences having different resolutions and bit rates by employing the above three reordering methods. As can be seen from Table 3, the priority reordering methods provided in the embodiments of the present invention have the best performance at different bit rates. Here, in the CIF sequence Foreman having a relatively low bit rate, the average PSNR of the priority reordering method provided in the embodiment of the present invention is equal to the default reordering method and the priority reordering method based on the quality layer. Thus, the total improvement is 8.21 dB. In the CIF sequence Football having a relatively high bit rate, the average PSNR performance of the priority reordering method provided in the embodiment of the present invention is improved by 6.41 db. In the low bit rate 4CIF sequence City, the average PSNR of the priority reordering method provided in the embodiment of the present invention is 3 .3 for the default reordering method and the priority reordering method based on the quality layer, respectively. Increased by 28 dB and 2.83 dB. In the high bit rate 4CIF sequence Harbor, the average PSNR of the priority reordering method provided in the embodiments of the present invention is 1. for the default reordering method and the priority reordering method based on the quality layer, respectively. 94 dB and 1.64 dB improvement.

  FIG. 13 illustrates a standard LT coding scheme (referred to as SLT), and an LT coding scheme (ALT and LT coding scheme) that adaptively selects the order provided in the embodiment of the present invention, with different packet loss rates and source data packet numbers N. It is a figure which shows the decoding success rate (SDR) at the receiving side of a scalable image | video in the case of employ | adopting. In FIG. 13, the redundancy of the transmission data packet is 25%, and the square and round curves represent the SDR on the receiving side of the scalable video when N is 576 and SLT and ALT are adopted, respectively. Curves with triangle marks and inverted triangle marks indicate SDRs on the receiving side of scalable video when N is 1278 and SLT and ALT are employed. As can be seen from FIG. 13, when the value of N is different, the SDR value of ALT is clearly better than the result of SLT.

  The above are only preferred embodiments of the present invention and do not limit the protection scope of the present invention. Various modifications, equivalent replacements, improvements and the like made within the spirit and principle of the present invention should all be included in the protection scope of the present invention.

Claims (11)

A method of transmitting scalable video according to priority,
Prioritization is performed on the hierarchical bit stream after scalable video coding (SVC) of each picture group (GOP) of scalable video, and the priority rearrangement is SVC coding in units of GOP. For the later SVC layered bitstream, the order of priority is rearranged according to its importance.
The rearranged SVC layered bitstream is subjected to package and Raptor encoding, and the encoded scalable video bitstream is transmitted to the scalable video receiver via the transmission channel.
Look at including it,
Performing the priority reordering on the SVC layered bitstream is:
All possible nodes on the reordering path from the bitstream with the lowest resolution characteristic to the bitstream with the highest resolution characteristic, the constant (s, t, q) characteristic (s, t and q are integers) , Representing the resolution characteristics of the spatial domain, temporal domain and quality domain, respectively)
All state nodes are partitioned into different stages so that the state nodes having the same increased step size sum are in the same stage due to the difference in the increased step sizes of s, t and q, and the increased step size sum is s, is the sum of the increasing step sizes of t and q,
Depending on the bit rate of each state node and its corresponding peak signal-to-noise ratio (PSNR), the formula
To calculate the reordering factor for each state node, where g _j represents the j th stage, t _i (g _j ) represents the i th state node in the j th stage, and t (g ₀ ) Represents the initial state node of the initial stage, t (g _K-1 ) represents the destination state node of the destination stage, the destination stage is the stage where the destination state node is located, and the destination state node is the highest resolution characteristic. Where PSNR _{ti (gj)} represents the PSNR of state node t _i (g _j ), R _{ti (gj)} represents the bit rate of state node t _i (g _j ), and PSNR _{t (g0 )} Represents the PSNR of the initial state node t (g ₀ ), R _{t (g0)} represents the bit rate of the initial state node t (g ₀ ), and PSNR _{t (gK−1)} represents the destination state node t (g _K represents PSNR of _-1), R _{t (gK- )} Represents the bit rate of the destination state node _{t (g K-1),} O i (g j) represents a sort factor state node t _i (g _j), K represents the total number of stages,
From all the paths that can be configured with the state nodes from the initial stage to the destination stage, determine the optimal path that maximizes the weighted sum of the reordering factors of each state node,
Outputting an SVC layered bitstream according to the order of each state node from the initial stage to the destination stage on the optimal path;
A method comprising: Determining the optimum path from all paths configurable by the state nodes from the initial stage to the destination stage is as follows:
Formula
To calculate the weighted sum of the reordering factors of each state node on all paths that can be configured with the state nodes from the initial stage to the destination stage,
The path having the largest weighted sum of the reordering factors is taken as the optimal path.
Including
Here, S _X represents the weighted sum of the reordering factors of each state node on the x-th path among all the paths that can be configured by the state nodes from the initial stage to the destination stage, and O (g _j ) is The reordering factor of the state node in the j-th stage of the x-th path is represented as t _i (g _j ), where O (g _j ) = O. _i (g _j ), and ω _j represents the weighting factor of the j-th stage (0 <ω _j <1, 0 ≦ j ≦ K−1),
The method according to claim 1 . Determining the optimum path from all paths configurable by the state nodes from the initial stage to the destination stage is as follows:
The initial state node of the initial stage is selected as the starting point, the node identifier j is set to 1, the weighted sum S ₀ (t) of the reordering factors of the initial stage is set to 0, and the optimal path state of the initial node Step A0 for initializing the node set to an initial state node;
Formula
To calculate the weighted sum S _j (t _i ) of the reordering factors of each state node in the j-th stage, and obtain the state node set of the optimal path from the initial stage to each state node in the j-th stage Step A1 to record each,
It is determined whether or not j is equal to K−1. If it is equal, the path consisting of each state node in the state node set of the optimal path of the destination state node is the optimal path, and if not equal, “j = j + 1” is set. Step A2 returning to Step A1,
Including
Here, S _j (t _i ) represents the weighted sum of the rearrangement factors of the i-th state node t _i in the j-th stage, and S _j−1 (t _k ) represents the state node in the j-th stage. represents a weighted sum of the sort factor k th state node t _k leading to t _i, omega _j represents a weighting factor of a stage of the j, O _i (g _j) is the i-th state at the stage of the j Represents a reordering factor of node t _i , M _j (t _i ) represents a state node set of an optimal path of i-th state node t _i in the j-th stage, and M _j−1 (t _k ) represents _j- th among the state node leading to the state node t _i at -1 stage, representative of the state node set of optimal path state node corresponding to the maximum value of the weighted sum of the sorting factor,
The method according to claim 1 . Determining the optimum path from all paths configurable by the state nodes from the initial stage to the destination stage is as follows:
Select the destination state node of the destination stage as the starting point, set the node identifier j to K-2, set the destination stage reorder factor weighting sum T _K-1 (t) to 0, and optimize the destination node Initializing the state node set of the path to the destination state node; and
Formula
Is used to calculate the weighted sum T _j (t _i ) of the reordering factors from each state node to the destination stage in the jth stage, and the optimal path from each state node to the destination stage in the jth stage Recording step a1 for each set of state nodes;
It is determined whether or not j is equal to 0. If they are equal, the path consisting of each state node in the state node set of the optimal path from the initial state node to the destination stage is set as the optimal path. 1 ”and returning to step a1, step a2;
Including
Here, T _j (t _i ) represents the weighted sum of the reordering factors from the i-th state node t _i in the j-th stage to the destination stage, and T _{j + 1} (t _k ) represents the state node in the j + 1-th stage. represents a weighted sum of the sort factor k th state node t _k leading to t _i, omega _j represents a weighting factor of a stage of the j, O _i (g _j) is the i-th state at the stage of the j Represents the reordering factor of node t _i , M _j (t _i ) represents the state node set of the optimal path from the i-th state node t _i to the destination stage in the j-th stage, and M _{j + 1} (t _k ) is Jo best path from state node corresponding to the maximum value of the weighted sum of the sort factors from each state node to the destination stage leading to state nodes t _i in the j + 1 stage to the destination stage It represents a set of nodes,
The method according to claim 1 . The j-th stage weighting factor ω _j is ω ^{j + 1} (0 <ω <1, 0 ≦ J ≦ K−1).
The method according to claim 2 , 3 or 4 . If there are more than one optimal path,
Increase the value of the weighting factor for each stage, and re-determine the optimal path from all paths that can be configured with state nodes from the initial stage to the destination stage.
The method according to claim 1, characterized by further comprising. The Raptor encoding is
RS encoding and interleaving for each of the source matrix blocks of each layer after reordering and packaging to obtain m data packets,
Determine the order of LT coding;
According to the determined LT coding order, LT data is subjected to LT coding to obtain n data packets.
The method of claim 1, comprising: Determining the order of the LT encoding is
Select a value of the order d within the range of
Including
Here, θ is a parameter of an order distribution function, and when the order distribution function is a Robust-Soliton function,
(Constant c> 0, δ is an acceptable decoding failure rate , k is to control the form of the order distribution function, and k = 1 if the order distribution function is a Robust-Soliton function ),
The method according to claim 7 . Determining the order of the LT encoding is
establishing a correspondence between the range of values of m and the optimal range of values of order d;
The optimal range of the value of the order d is determined by the value of m and the correspondence relationship ,
In the optimum range of the value of the order d, one integer is selected at random to obtain the LT coding order d.
Including
Establishing a correspondence between the range of values of m and the optimal range of values of order d is
classify the range of values of m,
By simulation, the decoding failure rate δ when the m value is in a different order range is obtained,
Formula
To decide,
9. The method of claim 8 , comprising: A device that transmits scalable video according to priority,
SVC encoding means for performing scalable video encoding (SVC) for each picture group (GOP) of the scalable video;
Against SVC hierarchical bit stream after SVC coding, have rows priority sort, the A priority sorting units of GOP, with respect to SVC hierarchical bit stream after SVC coding, the Priority sorting means, which means sorting the priority ranking according to importance ,
Packaging means for packaging the rearranged SVC layered bitstream;
Raptor encoding means for performing Raptor encoding on the SVC layered bitstream after packaging and transmitting the SVC layered bitstream to a scalable video receiving side via a transmission channel;
Only including,
The priority sorting means includes:
All possible nodes on the reordering path from the bitstream with the lowest resolution characteristic to the bitstream with the highest resolution characteristic, the constant (s, t, q) characteristic (s, t and q are integers) , Which represent the resolution characteristics of the space domain, the time domain, and the quality domain, respectively, and the state nodes having the same increased step size sum due to the difference of the increased step sizes of s, t, and q. Partitioning all state nodes into different stages so that they are in the same stage, wherein the increment step size sum is a sum of increment step sizes of s, t and q;
Depending on the bit rate of each state node and its corresponding peak signal-to-noise ratio (PSNR), the formula
To calculate the reordering factor for each state node, where g _j represents the j th stage, t _i (g _j ) represents the i th state node in the j th stage, and t (g ₀ ) Represents the initial state node of the initial stage, t (g _K-1 ) represents the destination state node of the destination stage, the destination stage is the stage where the destination state node is located, and the destination state node is the highest resolution characteristic. Where PSNR _{ti (gj)} represents the PSNR of state node t _i (g _j ), R _{ti (gj)} represents the bit rate of state node t _i (g _j ), and PSNR _{t (g0 )} Represents the PSNR of the initial state node t (g ₀ ), R _{t (g0)} represents the bit rate of the initial state node t (g ₀ ), and PSNR _{t (gK−1)} represents the destination state node t (g _K represents PSNR of _-1), R _{t (gK- )} Represents the bit rate of the destination state node _{t (g K-1),} O i (g j) represents a sort factor state node t _i (g _j), K Sorting factor represents the total number of stages A decision module;
An optimal path determination module that determines an optimal path that maximizes the weighted sum of the reordering factors of each state node from all paths that can be configured by the state nodes from the initial stage to the destination stage;
A reordering result output module that outputs an SVC layered bitstream according to the order of each state node from the initial stage to the destination stage on the optimal path;
The apparatus characterized by including. The Raptor encoding means is
An RS encoding module that performs RS encoding on the source matrix block of each layer after packaging to obtain m data packets;
An LT encoding order determination module that determines the order of LT encoding;
An LT encoding module that performs LT encoding on the m data packets according to the determined LT encoding order to obtain n data packets;
The device of claim 10 , comprising: