KR20090045882A - Image complexity computation in packet based video broadcast systems - Google Patents

Abstract

A method is disclosed for determining real time video complexity in video streaming, IPTV and broadcast applications using a statistical model indicative of video complexity and channel bandwidth variation taking into account scene content changes. The available channel bandwidth is unequally distributed among multiple video streams, in proportion to the video complexity and bandwidth change of the broadcast video stream. The distribution of available channel bandwidths is determined based on the image complexity factor of each video stream as determined from the probability matrices taking into account bandwidth variations and image complexity.

Description

Calculation of Image Complexity in Packet-Based Video Broadcast Systems {IMAGE COMPLEXITY COMPUTATION IN PACKET BASED VIDEO BROADCAST SYSTEMS}

The present invention relates generally to broadcast systems. More specifically, the present invention relates to methods for estimating the complexity of a series of pictures in compressed video programs using MPEG compatible encoding.

In typical broadcast systems, in conventional broadcast systems such as IPTV (Internet Protocol Television) and direct broadcast satellite (DBS) applications, multiple video programs are encoded in parallel, and the digitally compressed bitstreams are single. Multiplexed into one constant or variable bit rate channel. The available channel bandwidth may be distributed unevenly among the programs in proportion to the complexity / information content of the respective video sources. A monitoring system that calculates video quality by measuring impairments may consider the video complexity factor of the video stream to calculate different effects of impairments in fewer or more complex images.

MPEG encoded variable bit rate (VBR) video traffic is expected to control the bandwidth of broadband networks. This may be sent in on-demand streaming, IPTV or DBS types of environments. Accurate models of VBR or CBR video complexity need to enable monitoring systems to predict the performance of any proposed network during its operation. 1 illustrates the components involved in transmitting video content in a typical IPTV environment. Video sources originating as analog signals are encoded using encoders and packetized and transmitted using IP networks. It can be sent to the network as multicast or unicast. The core contains various elements for the facility and manages subscribers and traffic flows. The content is stored on content servers and sent on demand upon user request. At various points in the network, measurements may be performed for repairs by service assurance management systems.

MPEG coding standards define three picture types (I, B and P) and encode pictures in a fixed arrangement. Picture type changes may occur due to scene transitions. In the event of a sudden switchover, the first frame of the new scene is inner-coded (I-frame) to prevent serious coding errors. During gradual scene transition, the distance between two reference frames I or P can be changed to improve the picture quality. During most of these gradual transitions, temporary correlations tend to be reduced. This situation requires more frequent placement of reference frames (P-frames) that are predicted to maintain the required picture quality. If the video sequence includes fast motions, it may require frequent P-frames to improve picture quality. This increases the bit rate. On the other hand, if the scene does not contain any fast motions or progressive scene transitions, the inter-frame (I-frame) reference distance can be increased without affecting the picture quality. This is due to the strong correlation between the frames.

Thus, without limitation, bits in the video stream may be analyzed by analyzing VCL parameters including slices, macroblocks, quantization, INTER / INTRA coded reference, and non-reference macro / slice / picture types. A process is needed to analyze speed changes and video coding layer (VCL) complexity indication changes and arrive at a statistical model that dynamically calculates image complexity, and damage monitors use these values to influence the sequence of complex images. Can be determined.

The present invention provides a method for estimating image complexity in real time by statistical analysis of bandwidth variation and VCL parameters in a video program stream. This value can be used by other applications and monitoring for estimating video quality of lost states, allowing a good estimation of the quality perceived by the human visual system.

The process for broadcasting multiple video streams on a single channel begins with analyzing bit rate changes and complexity indication changes in the video coding layer of each of the multiple video streams. A statistical model is then created to dynamically calculate the picture complexity of each of the multiple video streams. The impact of the picture complexity of each of the multiple video streams in the broadcast is then determined. The available channel bandwidth is distributed among the multiple video streams based on the determined impact of the picture complexity of each of the multiple video streams.

The process further includes estimating video quality of certain lost states.

Analyzing the complexity indication changes includes analyzing changes in parameters of individual sections of the video streams. The separate sections of the video streams are sliced, macros, quantized, inter-coded references, intra-coded references, and non-reference macros / slices / pictures. Contains types.

Generating the statistical model includes generating a first statistical model of video coding layer complexity indication changes for individual sections of each video stream. In addition, a second statistical model of video coding layer bit rate changes or bandwidth changes is generated for the same individual sections of each video stream. The first and second statistical models from the individual sections of each video stream are then combined. The picture complexity of the individual sections of each video stream is calculated based on the combined first and second statistical models.

High quantization changes, slice / macroparts, and inter / intra prediction types for picture / slice / macropart types are counted by determining quantization changes in each video stream. The bandwidth change is counted by determining the bandwidth of the video coding layer data in each video stream. Counting includes a first counter for each quantization change, a second counter for each macro portion, a third counter for each slice, and a fourth counter for each low bandwidth, average bandwidth, and high bandwidth state transition. Is achieved by increment.

The complexity probability of the video coding layer complexity for the individual sections of each video stream is calculated using the first, second, third and fourth counters. In addition, the probability for low bandwidth, average bandwidth, and high bandwidth state for each video stream is calculated using the first, second, third, and fourth counters. The first transform probability matrix is configured for video coding layer complexity transformation of the individual sections of each video stream, and the second transform probability matrix is configured for bandwidth state transformation of the individual sections of each video stream. The picture complexity values of the individual sections of each video stream are calculated using the constraint state probabilities obtained from each transform probability matrix.

The method obtains image complexity values from distributed remote probes; Using image complexity as a more accurately acquired variable for sensed video quality to facilitate calculation of impairments in a packetized video stream; Provide picture complexity at regular intervals for packetized video applications; Provide an estimate of video complexity as sensed by the human visual system; Provide image complexity measurements for typical industrial video quality assessment models including, but not limited to, peak signal-to-noise ratio (PSNR), MPQM, MQUANT, and mean square root error (RMSE); Provide offline and real-time image complexity measurements that can be used or included by video encoders, multiplexers, routers, VOD servers (on-demand), broadcast servers and video quality measurement facilities; Provide a statistical model for bandwidth variation that contributes to image complexity; Provide a statistical model for video coding layer complexity that contributes to scene transitions; It can be used by collectors to determine the statistical distribution of a series of images at low and high complexity states.

Other features and advantages of the invention will be apparent from the following more detailed description, taken in conjunction with the accompanying drawings, which illustrate, by way of example, the principles of the invention.

The accompanying drawings illustrate the invention.

1 shows an example of an IPTV (IP TV) distribution network with potential points at which measurements for image complexity may be performed.

2 illustrates a typical protocol stack in which MPEG frames are encapsulated in IP (Internet Protocol) and measurement values are extracted at the VCL level.

3 shows a statistical model for calculating the image complexity through the final curve fit equation.

4 shows a Markov transformation process for a bandwidth model.

5 shows a Markov transformation process for a video coding layer complexity model.

6 shows the counters and transformation matrix relationship for the bandwidth model.

7 illustrates counters and transform matrix relationships for a video coding layer complexity model.

8 shows a transform probability matrix for a bandwidth change model.

9 illustrates a transform probability matrix for a video coding layer complexity model.

10 illustrates curve fit equation equations and probability values for calculating image complexity.

11 shows a flowchart for calculating bandwidth and video coding layer model.

Preferred embodiments of the invention are shown in FIGS. 2-10. One embodiment of the present invention can be used in an IPTV transmission system as shown in FIG.

The present invention relates to a method for estimating the image complexity of a series of pictures in a video stream supporting MPEG type picture encoding. The method provides a probability that, during the flow of the encoded video stream, quantization, macro portion / slice counts, macro portion sizes 16 × 16, 16 × 8, 8 × 8, 4 × 4, 8 × 16, scene transitions. Generating a statistical model indicative of the VCL parameters as the picture type change as the inter, intra, I / B / P frame / macropart types to determine. During the same flow of encoded video stream, a statistical model is also generated that represents the bandwidth change that determines the probability of high and low bandwidth states. The picture complexity is then determined from two statistical models generated from the encoded video stream of the same flow. The method can be used to provide a distribution system to estimate the sensed video complexity.

The method also includes: high quantization transforms, slice / macropart counts for the monitoring interval, inter / intra prediction types for picture / slice / macropart types (I, B, P) Determining quantization changes to determine the bandwidth of the VCL data to count the bandwidth change; Incrementing a counter for quantization changes, incrementing counters for macro portion and slice types and sizes, and incrementing a counter for low, average, and high bandwidth state transitions; Calculating a probability from the counters for state transitions for video coding layer complexity and calculating a probability from the counters for state transitions for low, average, and high bandwidth states; And calculating a transform probability matrix for the video coding layer complexity transform and calculating a transform probability matrix for the bandwidth state transform.

As summarized above, FIG. 1 shows a typical IPTV distribution network 21 comprising a video content acquisition unit 12, an IPTV management system 14, an IPTV content distribution unit 16, and an IPTV consumer 18. . Video source 20 is generally obtained in analog form, encoded by video encoder 22 in MPEG 1/2/4 format, and sent to video on demand (VOD) server 24 or broadcast server 26. do. VOD server 24 encapsulates the content into a program stream for delivery to network core 28. Network core 28 is a relatively higher bandwidth pipe. The IPTV network 21 also consists of various management, supply and service assurance components. Typically includes an operating support system (OSS) 30, a subscriber management system 32 and application servers 34 for creating new value added services. After management, supply and service assurance, the content may be stored on the VOD server 36 or the broadcast server 38 accessible by the consumer. It is typically located at the edge 40 of the net 21. The consumer has access to broadband access line 42, which may be a cable / DSL line 44. Typically, the television is connected to a set top box 46 which decodes the video stream into component outputs.

The protocol stack for the packetized video stream is shown in FIG. The media dependent accessory 48 may be an Ethernet, Sonnet, DS3, cable, or DSL interface. PHY 50 performs media dependent packet processing. IP (Internet Protocol) 52 is a network layer part that mainly provides addressing for packet routing in IPTV network 21. UDP / RTP 54 is a transport layer that provides application level addressing for ports. The video stream may be encapsulated in UDP / RTP or just UDP layer 54. The encoded video may be compressed to MPEG 1/2/4 and may be sent in RTP encapsulation for video 56 or as a transport stream. It may be an optional network abstraction layer as in the case for H.264 / AVC. The video coding layer packet input 60 is decoded and the necessary parameters are extracted to obtain the measurement values 62 for the image complexity model, as described below.

3 provides high level logic for statistical models in one embodiment of the invention. The MPEG VCL input 64 is provided to the VCL complexity (I-frame) model 66 and the bandwidth model 68 to calculate the counters needed for the statistical models. Curve fit equation 70 has model output parameters and calculates image complexity 72.

4 shows individual Markov process state transitions for the bandwidth model 68. Bandwidth changes of the video sequence are modeled with a three state Markov process to determine the probability of low and high bandwidth state transitions. State 1 (S1) 74, state 2 (S2) 76, and state 3 (S3) 78 represent states of model 68 of low, constant and high bandwidth states, respectively.

5 shows individual Markov process state transitions for the VCL layer complexity quantization model 66. Quantization transforms retrieved from the macro layer are modeled in a two state Markov process. K1 80 and K2 82 show the states of the VCL layer complexity model 66-high quantization and low quantization states.

6 shows counters 86 used to calculate transform probabilities 90 of bandwidth model 68. The VCL bandwidth monitor 84 monitors the bandwidth changes of the VCL stream and updates the counters cXY 86, where X represents the initial state and Y represents the resulting state. The initial and resultant states may be low, constant or high bandwidth states, designated 1, 2 or 3, respectively. For example, C11 represents a state transition event from low bandwidth state 74 to low bandwidth state 74, and C23 represents a state transition event from constant bandwidth state 76 to high bandwidth state 78.

State transformation probabilities 90 are calculated to obtain a transformation matrix 88. State transition probabilities 90 are represented by pXY, where X represents the initial state and Y represents the resulting state. The initial and resultant states may be low, constant or high bandwidth states, designated 1, 2 or 3, respectively. For example, p12 is the conversion probability going from the low bandwidth state (S1) 74 to the constant bandwidth state (S2) 76. From the transform probabilities 90, a transform matrix 88 is formed. From the transformation matrix 88, the limited state probabilities are calculated without the initial conditions for obtaining BP101 92 and BP103 94. These values represent the probability of staying in a low bandwidth state and a high bandwidth state, respectively.

Counters 98 used to calculate transform probabilities for the VCL layer complexity quantization model are shown in FIG. 7. VCL slice and macro portion monitor 96 monitors the quantization parameter of the macro portion and updates the counters dXY, where X represents the initial state and Y represents the resulting state. The initial and resultant states may be high or low quantization reception states, designated as 1 or 2, respectively. For example, d12 represents state transition event counts from a high quantization reception state to a low quantization reception state. State transformation probabilities are calculated to obtain the transformation matrix 100 for the VCL layer quantization model 66. From the transform matrix 100, a high quantization occurrence probability in the picture sequence is calculated and set to the variable IP100 102.

The transform probability matrix 88 for the bandwidth model 68 is shown in FIG. 8. States S1 74, S2 76, and S3 78 represent low, average, and high bandwidth states, as summarized above, with each cell of the matrix 88 moving from one state to another. Represents a state transition probability.

9 shows a transform probability matrix 100 for the VCL layer quantization model 66. States K1 104 and K2 106 represent high quantization and low quantization occurrence states, and each cell of matrix 100 represents a state transition probability from one state to another.

FIG. 10 shows parameters BP101 92, BP103 94 and IP100 used in the curve fit equation 73 of FIG. 3 to obtain an image complexity (Γ) 72 ranging from 2 to 3; VCL layer complexity model 66 and bandwidth model 68 from 102 are shown.

11 shows a flow diagram for the main functional units of the progressive process. Bandwidth model initialization 108 is the first step that needs to be performed to execute the bandwidth 68 and VCL layer complexity 66 models. Variables for calculating the average bandwidth are initialized (110). The VCL input is read from NAL (Network Abstraction Layer) / Transport Stream 112. The average bandwidth for the VCL packets is calculated 114 and set 116. During this operation, bandwidth model 68 and VCL layer complexity model 66 are executed in parallel (118). Bandwidth model 68 is initialized for conversion counters 120. The VCL packet size is read from the NAL / Transport Layer Stream 122. The bandwidth for the VCL is calculated 124. The transform counters are updated 126, and the transform probability matrix is updated 128. The next step is to calculate the high and state limited state probabilities 130 using equations (1) and (2), as detailed below. Variables BP101 and BP103 are set (132). For all macro portions, the VCL complexity model 66 is executed 118 at the same time. The counters are initialized 136 and the macro portion and slice quantization parameters are read from the NAL / Transport Stream 138 by decoding slice data from the VCL. The VCL complexity quantization transmission probability matrix is calculated (140) and the limited state probabilities are calculated (142). Then, the IP 100 variable is set (144). The final curve fit equation is calculated using the variables BP101, BP103, and IP100 (146).

The operation of one embodiment will now be described in more detail. Bandwidth model 68 is constructed using the Markov model of FIG. States S1 74, S2 76, and S3 78 relate to the state of the VCL packet rate at any instance of time after processing a particular number of VCL packets or individual sections. The bandwidth model 68 is initialized after generating the MPEG video stream. In this step, the bandwidth model 68 determines the average bandwidth of the video stream in each individual section, i.e., at each sampling moment. The procedure for determining the average bandwidth is as follows:

Initializing counters A100, A101, A102, A103, A104 to zero;

O Read VCL packet size for all NAL / Transport layer packets received from the MPEG layer, and set A100 for the received cumulative size;

A103 increments for all INTRA macroparts / picture types;

A104 increments for all slice types;

O set A101 to the first VCL reception time in milliseconds;

O setting A102 for all VCL reception times in milliseconds; And

O Average bandwidth calculation at each sampling instant.

The calculation follows the following procedure:

A100 = A100 + VCL_size_rcvd from MPEG layer

If (A101 = 0), then A101 = current time

A102 = current time

C100 = A100 * 8 / (A102-A101) / 1000 (in kbps)

The average bandwidth (C100) range will be C100 ± 10 kbps.

The model is executed only when the minimum pre-defined count of A103 is received. This counter represents scene transitions, and multiple scene transitions are needed to effectively calculate the model. The model is currently in a low bandwidth state S1 if the video stream bandwidth is lower than C100-10 kbps; For bandwidths higher than C100 + 10 kbps the model is in a high bandwidth state S3. If the bandwidth is within the average bandwidth value, the model is in constant bandwidth state S2.

The average bandwidth C100 is determined continuously for the VCL packets, and the bandwidth change can be modeled using the discrete transform Markov process shown in FIG. The transitions of the three states S1, S2 and S3 are calculated by monitoring the video stream bandwidth change. Transformation matrix 88 (FIG. 8) is obtained, where each cell represents the probability of state transition from one state to the next. Since the Markov model for this process has no periodic states and no recurrent states from a single chain, the limit-state probabilities are independent of the initialization conditions. This condition may be applied to obtain P1 (probability in the S1 state), P2 (probability in the S2 state), and P3 (probability in the S3 state). For the limit-state probabilities, the following equations are maintained:

방정식 (1)

Equation (1)

방정식 (2)

Equation (2)

Since there are three variables P1, P2, P3 to be solved, three simultaneous equations are needed, each of which can be generated from the transformation matrix 88 (FIG. 8). Transform matrix 88 is constructed from MPEG video stream bandwidth change statistics. Transformation matrix 88 is obtained by calculating the probability of transformation from a particular state to any other possible state, as shown in FIG. For example, the probability of staying in state S1 is represented by p11.

These transform probabilities are written in equations (1) and (2) to obtain three simultaneous equations that can be solved to obtain P1, P2 and P3, where they represent: P1 (model in low bandwidth state) Probability of; P2 (probability of the model in average / constant bandwidth state); And P3 (probability of the model in high bandwidth state).

The probabilities of the low and high transforms proceed to the final curve fit equation 70 to obtain the image complexity value 72. The algorithm for obtaining P1, P2 and P3 is described as follows:

O Initialize counters c11, c12, c13, c21, c22, c23, c31, c32 and c33 to zero;

O = S2; And

O For some VCL packets of the MPEG video elementary stream (configurable count),

For each sampling instant (eg 10 seconds), the transformation matrix 88 is calculated from above. Transform probabilities are calculated from the relative frequencies of the state transforms.

From the transformation matrix, the probabilities P1 (low speed probability), P2 (constant / average speed probability) and P3 (high speed probability) are calculated using three simultaneous equations formed using equations (1) and (2) do.

Putting the transform probabilities into equation (1), the following is achieved:

방정식 (3)

Equation (3)

방정식 (4)

Equation (4)

From equation (2), the following is achieved:

방정식 (5)

Equation (5)

After solving the equations, the probabilities are assigned to these variables:

After solving the three equations, P1 and P3 are calculated for use in the curve fit equation 70 to obtain the final image complexity 72.

For all VCL inputs, the VCL layer complexity model 66 needs to be executed in parallel. VCL parameters are monitored for scene transitions and picture quality. INTER / INTRA macropart types are analyzed to determine scene transitions, and quantization parameters inside the macro part are read to determine picture quality as a contribution to image complexity. After the VCL complexity model 66 is executed, the curve fit equation 70 for image complexity may be solved to obtain the final image complexity value 72.

VCL complexity probability calculation follows a similar process as described above, but the Markov states are limited to two states. 10 shows a state transition process of the VCL complexity model. The states represent the following:

K1 (a state in which a high quantization macro portion is received); And

K2 (a state in which the low quantization macro portion is received).

Transformation matrix 100 (FIG. 9) is calculated that includes the transformation probabilities. Each cell represents a state transition, for example p12 represents the probability of having a low quantization K2 in the high quantization reception state K1.

The procedure for calculating transform probabilities is as follows:

For all VCL inputs, in the MPEG video elementary stream,

O Reset all counters d11, d12, d21, d22 to zero. Set state = K1.

O To determine the quantization threshold to set high / low quantization states, read the initial quantization value from the picture parameter setting (as in MPEG4) or from the preconfigured value if it is not available. Set C101 to this value.

O Set C102 to zero for macro part counts

○ Set C103 to zero for INTRA macropart types

O Set C105 to zero for slice types

C102 increments for all macro sections processed

C103 increment for all INTRA macropart types

C105 increments for all slice types

O Read the quantization value of C104 for all macro parts where quantization is available

Every sampling instant (eg, 10 seconds) from the counters, transform probabilities are calculated to obtain the transform matrix 100.

From the transform probabilities, the probability of high quantization occurrence P1 and low quantization occurrence P2 can be calculated. The high probability of quantization will be used in the curve goodness-of-fit function 70 to obtain the final image complexity value 72.

Since the constraint state probabilities are independent of the initial conditions, the simultaneous equations for the constraint-state probabilities can be solved as follows:

방정식 (6)

Equation (6)

방정식 (7)

Equation (7)

Substituting and extending the transform probabilities in equations (6) and (7),

방정식 (8)

Equation (8)

방정식 (9)

Equation (9)

The two equations are solved to obtain P1 and P2. Assign IP100 = P1 (high probability of macroblock VCL layer complexity) to be used in the image complexity equation.

BP101, BP103 and IP100 are used in the curve fit equation 70 (FIG. 10) to obtain image complexity 72 in the range of 2-3.

The image complexity (Γ) is for (Γ> 3) Γ = 3

방정식 (10)

Equation (10)

While some embodiments have been described in detail for purposes of illustration, various modifications may be made to each without departing from the spirit and scope of the invention.

Claims (25)

A process for broadcasting multiple video streams on a single channel, Analyzing complexity indication changes and bit rate changes in a video coding layer of each of the multiple video streams; Generating a statistical model to dynamically calculate image complexity of each of the multiple video streams; Determining the effect of picture complexity of the multiple video streams in the broadcasting; And Distributing the available channel bandwidth between the multiple video streams based on the determined impact of the picture complexity of each of the multiple video streams. Broadcasting process of multiple video streams comprising a. The method of claim 1, Said analyzing step comprises analyzing video coding layer complexity indication changes in parameters of individual sections of said video streams. The method of claim 2, The separate sections of the video streams are sliced, macroblocks, quantized, inter-coded references, intra-coded references, and non- A broadcasting process for multiple video streams, characterized by comprising non-reference macroparts / slices / picture types. The method of claim 1, Estimating video quality of lost states. The method of claim 1, The generating step, Generating a first statistical model of video coding layer complexity indication changes for individual sections of each video stream; Generating a second statistical model of bandwidth variation or video coding layer bit rate variations for the same individual sections of each video stream; Combining the first and second statistical models from separate sections of each video stream; And Computing image complexity for individual sections of each video stream based on the combined first and second statistical models. The method of claim 5, wherein Counting high quantization transitions, slice / macroparts, and inter / intra prediction types for picture / slice / macropart types by determining quantization changes in separate sections of each video stream; And Counting the bandwidth change by determining the bandwidth of the video coding layer data in the individual sections of each video stream. The method of claim 6, The count steps increment the first counter for each quantization change, increment the second counter for each macro portion, increment the third counter for each slice, and each low bandwidth, average bandwidth and Incrementing a fourth counter for high bandwidth state transitions. The method of claim 7, wherein Calculating probabilities for video coding layer complexity for individual sections of each video stream using the first, second, third and fourth counters; And Calculating probabilities for low bandwidth, average bandwidth, and high bandwidth states for individual sections of each video stream using the first, second, third, and fourth counters. Broadcasting process of multiple video streams. The method of claim 8, Constructing a first transform probability matrix for the video coding layer complexity transform of the individual sections of each video stream; And And constructing a second transform probability matrix for the bandwidth state transitions of the individual sections of each video stream. The method of claim 9, Calculating the image complexity of the individual sections of each video stream using the limiting state probabilities obtained from each transformation probability matrix. A process for broadcasting multiple video streams on a single channel, Analyzing bit rate changes and complexity indication changes in a video coding layer in parameters of individual sections of each of the multiple video streams; Generating a statistical model to dynamically calculate image complexity of each of the multiple video streams; Determining the effect of picture complexity of the multiple video streams in the broadcasting; Distributing available channel bandwidth between the multiple video streams based on the determined effect of the image complexity of each of the multiple video streams; And Estimating video quality of lost states Broadcasting process of multiple video streams comprising a. The method of claim 11, The separate sections of the video streams comprise slices, macro portions, quantization, inter-coded references, intra-coded references, and non-reference macro portion / slice / picture types. Broadcasting process. The method of claim 11, The generating step, Generating a first statistical model of video coding layer complexity indication changes for individual sections of each video stream; Generating a second statistical model of bandwidth variation or video coding layer bit rate variations for the same individual sections of each video stream; Combining the first and second statistical models from separate sections of each video stream; And Based on the combined first and second statistical models, calculating image complexity for individual sections of each video stream. The method of claim 13, Counting high quantization transforms, slice / macroparts, and inter / intra prediction types for picture / slice / macropart types by determining quantization changes in separate sections of each video stream; And Counting the bandwidth change by determining the bandwidth of the video coding layer data in the individual sections of each video stream. The method of claim 14, The counting steps increment the first counter for each quantization change, increment the second counter for each macro portion, increment the third counter for each slice, and each low bandwidth, average bandwidth. And incrementing a fourth counter for high bandwidth state transitions. The method of claim 15, Calculating probabilities for video coding layer complexity for individual sections of each video stream using the first, second, third and fourth counters; And Using the first, second, third and fourth counters to calculate probabilities for low bandwidth, average bandwidth and high bandwidth states for individual sections of each video stream. Broadcasting process of multiple video streams. The method of claim 16, Constructing a first transform probability matrix for video coding layer complexity transform of individual sections of each video stream; And And constructing a second transform probability matrix for the bandwidth state transformation of individual sections of each video stream. The method of claim 17, Using the limited state probabilities obtained from each transform probability matrix, calculating image complexity of individual sections of each video stream. A process for broadcasting multiple video streams on a single channel, Analyzing bit rate changes and complexity indication changes in the video coding layer of each of the multiple video streams; Generating a first statistical model of video coding layer complexity indication changes for individual sections of each video stream; Generating a second statistical model of bandwidth variation or video coding layer bit rate variations for the same individual sections of each video stream; Combining the first and second statistical models from separate sections of each video stream to dynamically calculate the picture complexity of each of the multiple video streams; Determining the effect of picture complexity of the multiple video streams in the broadcasting; Distributing available channel bandwidth between the multiple video streams based on the determined effect of the image complexity of each of the multiple video streams; Calculating image complexity for individual sections of each video stream based on the combined first and second statistical models; And Estimating video quality of lost states Broadcasting process of multiple video streams comprising a. The method of claim 19, Counting high quantization transforms, slice / macroparts, and inter / intra prediction types for picture / slice / macropart types by determining quantization changes in each video stream; And Counting the bandwidth change by determining the bandwidth of the video coding layer data in each video stream. The method of claim 20, The count steps increment the first counter for each quantization change, increment the second counter for each macro portion, increment the third counter for each slice, and each low bandwidth, average bandwidth and Incrementing a fourth counter for high bandwidth state transitions. The method of claim 21, Calculating probabilities for video coding layer complexity for individual sections of each video stream using the first, second, third and fourth counters; And Using the first, second, third and fourth counters to calculate probabilities for low bandwidth, average bandwidth and high bandwidth states for individual sections of each video stream. Broadcasting process of multiple video streams. The method of claim 22, Constructing a first transform probability matrix for video coding layer complexity transform of individual sections of each video stream; And And constructing a second transform probability matrix for the bandwidth state transformation of individual sections of each video stream. The method of claim 23, Using the limited state probabilities obtained from each transform probability matrix, calculating image complexity of individual sections of each video stream. The method of claim 19, The analyzing step includes analyzing video coding layer complexity indication changes in parameters of individual sections of the video streams, wherein the individual sections of the video streams are sliced, macro portions, quantized, inter-coded references, intra. A process of broadcasting multiple video streams comprising coded references and non-reference macros / slices / picture types.

Images