JP4354962B2 - Video quality estimation apparatus and video quality estimation method - Google Patents

Description

  The present invention relates to a video quality estimation apparatus and a video quality estimation method for estimating video quality in a video communication service between communication terminals via a network.

  In recent years, video communication services (video communication services) such as video distribution services and video conferencing are becoming widespread as communication networks such as the Internet have become broadband. In these video communication services, a video signal is compression-encoded with high efficiency, converted into a packet sequence, and sent to a communication terminal such as a video distribution client or a bidirectional video communication service client.

  For example, an I (Intra Picture) frame that performs encoding processing within a frame without using inter-frame prediction, a P (Predictive Picture) frame that performs encoding processing by predicting the present from the past using motion compensation prediction, and bi-directional prediction Use forward and backward prediction and encode processing (Bidirectionally Predictive Picture) Video signal is highly efficient compression encoded by three types of frames called B frames, and this highly efficient compression encoded video signal is converted into a packet sequence And sent to the video distribution client and the interactive video communication service client via the network.

  In this case, for example, when packet loss occurs in the network, the frame information before and after is used for decoding at a communication terminal such as a video distribution client or a bidirectional video communication service client. There are cases where the video frame is not limited to one video frame but may extend to a plurality of video frames. In such a case, although there is no packet loss in the next video frame, quality degradation occurs in the decoded video, and the video quality (user experience quality) actually experienced by the user is greatly reduced. . This depends on the GOP (Group of pictures) configuration of the I, P, and B frames, that is, the configuration of the video frame in one unit time corresponding to a group of encoded video frames.

  The GOP configuration is expressed by a combination of an appearance period M of I or P and an appearance period N of I. An example of the GOP configuration is shown in FIG. This example shows the GOB configuration when M = 3 and N = 15. In the case of 30 frames per second, the number of frames of 1 GOP configuration is 15, so 1 GOP time (1 configuration unit time) ) Is 0.5 seconds.

  As described above, even if one packet is lost, the user experience quality varies depending on the type of the deteriorated video frame. Therefore, as a method for accurately estimating the user experience quality, the invalid packet (lost packet, reverse sequence packet, late arrival packet) in the network and communication terminal is detected, and the video frame to which the invalid packet belongs and the influence of the video frame are detected. The number of video frames that deteriorate due to reception is calculated as the number of invalid frames, and the ratio of the number of invalid frames to the total number of transmission frames in the measurement target section (for example, 10 seconds) is obtained as the invalid frame rate. A method for estimating the user experience quality has been considered (see Non-Patent Documents 1, 2, and 3). If this method is used, as reported in Non-Patent Document 3, the quality estimation accuracy is improved as compared with the case where the user experience quality is estimated based on the invalid packet rate.

  In this method, information of a video frame to which an invalid packet belongs is obtained from header information and payload information of the invalid packet. Of the lost packets, reverse order packets, and late arrival packets that make up invalid packets, the reverse order packets and late arrival packets exist, so the information of the video frame to which the packets belong is determined from the header information and payload information. It is possible to get.

Masuda, Tominaga: “Video quality estimation method based on invalid packet rate”, Shinso Univ., B-11-15, March 2004. Tominaga, Masuda, Hayashi: “Video quality estimation based on network quality scale considering video frame information”, Shinso Univ., BS-6-2, March 2005. Masuda, Tominaga, Hayashi: “In-service video quality control using invalid frame rate”, IEICE Technical Report, CQ2005-59, September 2005.

  However, in the above-described method, for a lost packet, since the packet is lost, information on the video frame to which the packet belongs cannot be obtained from header information or payload information. In this case, the information of the video frame to which the packet belongs is inferred from the packets before and after the lost packet. However, for example, when the preceding and following packets are also lost, it is difficult to estimate and the video quality (user experience) (Quality) estimation accuracy decreases.

  The present invention has been made to solve such a problem. The object of the present invention is to obtain information on a video frame to which an invalid packet belongs from header information and payload information, and before and after a lost packet. An object of the present invention is to provide a video quality estimation apparatus and a video quality estimation method capable of accurately estimating video quality even when packets are lost.

  In order to achieve such an object, the present invention provides an image quality estimation apparatus for estimating image quality in an image communication service between communication terminals via a network, in which one component unit corresponding to a group of encoded image frames is provided. Video coding information collecting means for collecting video coding information including video frame configuration information from the communication terminal from the communication terminal, and a section of one structural unit time width from the start of the video quality measurement target section, The first region is divided into n (n ≧ 2), and one structural unit time start position determining means for temporarily determining the starting positions of n structural unit times, and a video code for each of the starting positions of n structural unit times Based on the activation information, provisional setting means for temporarily determining the position and type of each video frame after the start position, and information on generation of invalid packets in the network and communication terminal are collected Based on the position and type of the video frame provisionally determined by the temporary setting means for each start position of the effective packet generation information collecting means and n one unit time, which invalid packet is generated after the start position. Invalid frame number calculating means for determining whether it belongs to a type of video frame and calculating the number of video frames to which the invalid packet belongs and the number of video frames deteriorated under the influence of the video frame as the number of invalid frames; Video quality estimation means for estimating the video quality in the measurement target section based on the number of invalid frames calculated for each start position of one structural unit time is provided.

  In this invention, when the communication terminal is a bidirectional video communication service client and one constituent unit when encoding a video signal is GOP, the video including the configuration information of the video frame during one GOP time from the bidirectional video communication service client. Encoding information (encoding method, bandwidth, I or P appearance period M, I appearance period N, number of packets constituting each type of video frame, etc.) is collected. Then, a section having a width of 1 GOP from the start of the video quality measurement target section (for example, 10 seconds) is used as a head area, and the head area is divided into n (for example, 10 sections) to start positions of n GOP times ( The starting position of one structural unit time) is provisionally determined. Note that the 1 GOP time is obtained from the video encoding information, for example, in the case of a video with a GOP configuration of M = 3, N = 15, and 30 frames per second, 0.5 seconds.

  Then, for each start position of the n GOP times, the position and type of each video frame after the start position are tentatively determined based on the video encoding information including the configuration information of the video frame during one GOP time. For example, the ratio of the total number of packets for each type of video frame to the total number of packets in one GOP time is obtained as the composition packet ratio by video frame type, and the total number of packets in one GOP time, the composition packet ratio by video frame type, and 1 GOP time Based on the arrangement of the video frames in the middle (the arrangement of I, P, and B frames), the position and type of each subsequent video frame are provisionally determined for each start position of n GOP times.

  In this case, the configuration packet ratio for each video frame type is obtained by obtaining the number of configuration packets for each video frame type obtained from video encoding information and the total number of packets for each video frame type from the GOP configuration. The total number of packets in the GOP may be obtained by totaling the total number of packets in the GOP, and the total number of packets for each video frame type may be divided by the total number of packets in the GOP. A table that defines the relationship between the number or bandwidth and the composition packet ratio for each video frame type may be prepared and obtained from this table. Also, an estimation expression that represents the relationship between the total number of packets or bandwidth in the GOP and the configuration packet ratio for each video frame type may be created in advance and obtained from this estimation expression.

  Then, for each start position of the n GOP times, it is determined which type of video frame the invalid packet belongs to based on the position and type of the video frame tentatively determined, and the video to which the invalid packet belongs The number of frames and the number of video frames that deteriorate due to the influence of the video frames is calculated as the number of invalid frames. As a result, the total number of invalid frames is calculated while shifting the start positions of the n GOP times one by one.

  Here, when calculating the number of invalid frames, for each start position of the GOP time, a method (method 1) in which a section from the start position of the GOP time to the end of the measurement target section is a calculation target section, and the GOP time For each start position, there can be considered a method (method 2) in which a time width of a measurement target section from the start position of the GOP time is used as a calculation target section.

  In the case of method 1, if the start position of the GOP time is shifted one by one, the invalid frame number calculation target section is gradually shortened. However, if the measurement target section is considerably longer than the 1 GOP time, such as 1 GOP time of 0.5 seconds and the measurement target section of 10 seconds, an error in the number of invalid frames calculated for each start position of the GOP time. Is small and has little effect on the estimation accuracy of video quality.

  In the case of method 2, even if the start position of the GOP time is shifted one by one, the calculation target section of the number of invalid frames is not shortened, and a constant time width (time width of the measurement target section) is always secured. Therefore, the use of method 2 can be expected to improve the estimation accuracy of the video quality as compared to the method 1. When the measurement target section is short, the method 2 is preferable.

  In the present invention, in principle, the measurement target section may be 1 GOP time or longer, but when the 1 GOP time is short, for example, 0.5 seconds, the measurement target period should be considerably longer than the 1 GOP time. Good. In addition, when 1 GOP time is long, for example, 8 seconds, it is desirable to set it to 2 GOP time or more.

  In the present invention, the estimation of the video quality in the measurement target section is performed based on the number of invalid frames calculated for each start position of n GOP times. For example, an average value of the number of invalid frames calculated for each start position of n GOP times is used as an invalid frame number expected value, and the ratio of the invalid frame number expected value to the total number of transmission frames in the measurement target section is used as an invalid frame. Obtained as an expected rate value, and substitutes this expected invalid frame rate value into a predetermined video quality estimation model indicating the relationship between invalid frame rate and video quality to obtain an estimated value of video quality within the measurement target section .

  Note that the expected number of invalid frames may be the maximum number of invalid frames calculated for each start position of n GOP times, or may be a 75% value (75% of the maximum value). By setting the maximum number of invalid frames as the expected number of invalid frames, quality management can be performed assuming the worst case. By setting the 75% value as the expected number of invalid frames, quality management can be performed on the safe side. Also, a video quality estimation model indicating the relationship between the invalid frame rate and the video quality may be obtained for each service band and content. Further, the video quality estimation model may be created as a model formula or may be prepared as a table.

  In the present invention, one structural unit for encoding is not limited to GOP. The GOP is sometimes called a GOV (Group of video). In addition, the present invention can be realized not as a video quality estimation device but as a video quality estimation method.

  According to the present invention, video encoding information including configuration information of video frames in one unit time corresponding to a group of encoded video frames is collected from a communication terminal, and from the beginning of a video quality measurement target section. A section of one structural unit time width is used as a head area, and the head area is divided into n to tentatively determine start positions of n one structural unit time, and for each start position of the n one structural unit time, one configuration Based on video encoding information including video frame configuration information during a unit time, the position and type of each video frame after the start position are provisionally determined, and the type of video frame to which the invalid packet belongs is determined. The number of video frames to which invalid packets belong and the number of video frames that deteriorate due to the influence of the video frames are calculated as the number of invalid frames. Since the video quality within the measurement target section is estimated based on the calculated number of invalid frames, the information about the video frame to which the invalid packet belongs is not obtained from the header information and payload information, and the packets before and after the lost packet Even when the image is lost, the video quality can be estimated accurately.

  Hereinafter, the present invention will be described in detail with reference to the drawings. FIG. 1 is a system configuration diagram showing an example of a video communication service system using a video quality estimation apparatus according to the present invention. In the present embodiment, the video quality represents the user experience quality and refers to a subjective evaluation value (for example, MOS: Mean Opinion Score) obtained by a subjective evaluation experiment.

  In FIG. 1, 10 is a video quality estimation apparatus according to the present invention, 20 (20-1, 20-2) is a bidirectional video communication service client (hereinafter referred to as a bidirectional user terminal), 30 is a video distribution server, 40 Is a video distribution client (hereinafter referred to as a video distribution terminal) that receives video distribution from the video distribution server 30, and these devices are connected to each other via a network 50.

  FIG. 2 is a block diagram showing the main part of the internal configuration of the video quality estimation apparatus 10. The video quality estimation apparatus 10 collects video encoding information (encoding method, bandwidth, video frame configuration information, number of configuration packets for each video frame type, etc.) from the interactive user terminal 20 and the video distribution server 30. An encoded information collection unit 11, an invalid packet generation information collection unit 12 that collects generation information of invalid packets (lost packets, reverse sequence packets, late arrival packets) from the interactive user terminal 20 and the video distribution terminal 40, and a video code A video packet type-specific configuration packet ratio calculation unit 13 that calculates a video frame type-specific configuration packet rate, which will be described later, based on the video encoding information collected by the video encoding information collection unit 11, and a video collected by the video encoding information collection unit 11 Encoding information, composition packet ratio by video frame type calculated by video packet type composition packet ratio calculation unit 13 and An invalid frame rate expected value calculation unit 14 that calculates an invalid frame rate expected value, which will be described later, based on invalid packet generation information collected by the valid packet generation information collection unit 12, and the relationship between the invalid frame rate and video quality Based on the video quality estimation model DB 15 stored as the estimation model and the expected invalid frame rate value calculated by the invalid frame rate expected value calculation unit 14, the estimated video quality value is determined according to the video quality estimation model stored in the video quality estimation model DB 15. And a desired video quality estimation unit 16.

  Note that the video quality estimation apparatus 10 is realized by hardware including a processor and a storage device, and a program that realizes the functions of the units illustrated in FIG. 2 in cooperation with the hardware. Further, in this example, a model formula indicating the relationship between the invalid frame rate and the video quality is created in the video quality estimation model DB 15 for each condition such as the encoding method, content, and bandwidth, and stored as a video quality estimation model. is doing.

FIG. 3 shows an example of a video quality estimation model. When an invalid frame rate is obtained, video quality (MOS) can be calculated. It has been found that the video quality estimation model for the invalid frame rate is represented by an exponential function such as the following equation, for example.
Y = y0 + A1 * exp (−x / t1) + A2 * exp (−x / t2) (1)
Y: user experience quality (MOS), x: invalid frame rate, y0, A1, A2, t1, t2: constants.

[Embodiment 1 (Method 1)]
In the following, according to the flowchart shown in FIG. 5, functions specific to the present embodiment of the video quality estimation apparatus 10 will be described by taking as an example the case of estimating the video quality in the interactive user terminal 20.

[Video coding information collection unit]
In the video quality estimation apparatus 10, the video encoding information collection unit 11 is configured to encode a video signal from the interactive user terminal 20, a band, video frame configuration information, and a configuration packet for each video frame type. Video encoding information such as the number is collected (step 101). In this example, it is assumed that the encoding method in the interactive user terminal 20 is high-efficiency compression encoding, and GOP configuration (M, N) is collected as video frame configuration information.

  In the case of a video distribution service, the number of configuration packets for each type of video frame may be collected for video content of a genre assumed in advance, or information on content actually used in the service may be data. May be collected. In the case of a bidirectional video communication service, the number of packet information for each video frame type is collected using sample video of a video content of a genre that is assumed in advance (for example, distance learning, free conversation, etc.). It may be left.

[Configuration packet ratio calculation unit by video frame type]
The video packet type composition packet ratio calculation unit 13 receives the video coding information collected by the video coding information collection unit 11 and calculates the video frame type composition packet ratio based on the video coding information (step). 102).

  In this example, the total number of packets for each type of video frame is obtained from the number of constituent packets for each type of video frame obtained from video coding information (the number of packets constituting each frame of I, P, and B) and the GOP configuration. The total number of packets for each type of video frame is totaled to obtain the total number of packets in the GOP, the total number of packets for each type of video frame is divided by the total number of packets in the GOP, and the video frame of the I frame The configuration packet ratio by type, the configuration packet ratio by video frame type of P frame, and the configuration packet ratio by video frame type of B frame are obtained.

  It should be noted that a table that defines the relationship between the total number of packets in the GOP or the bandwidth and the composition packet ratio for each video frame type (see FIG. 4) is prepared, and from this table, the video frame type for each of the I, P, and B frames. Another configuration packet ratio may be obtained. In addition, an estimation expression representing the relationship between the total number of packets or bandwidth in the GOP and the configuration packet ratio for each video frame type is created in advance, and the configuration packet for each frame of I, P, and B frames from this estimation formula The ratio may be obtained.

[Invalid packet occurrence information collection unit]
The invalid packet generation information collection unit 12 sets 10 seconds as a measurement section (measurement target section), and collects generation information of invalid packets in the measurement section from the interactive user terminal 20 (step 103). The invalid packet is a sum of lost packets and reverse order packets in the network 50 and late arrival packets due to buffer overflow in the bidirectional user terminal 20. In this embodiment, the interactive user terminal 20 collects invalid packet occurrence information and transfers the collected invalid packet occurrence information to the video quality estimation apparatus 10 in response to a request from the video quality estimation apparatus 10. Take the way. For this transfer, for example, RTCPXR described in “IETF RFC3611” is used as a control packet.

[Invalid frame rate expected value calculation unit]
The invalid frame rate expected value calculation unit 14 calculates the expected invalid frame rate based on the video encoding information, the configuration packet ratio for each video frame type, and the generation information of invalid packets as follows (step 104). FIG. 6 shows details of the processing for calculating the expected invalid frame rate in step 104.

  First, a section of 1 GOP time (TS shown in FIG. 7B) from the start of the measurement section (T (T1) shown in FIG. 7A) is set as the head area (step 201). The 1 GOP time is also called an intra refresh time, and is obtained from video coding information. In the case of MPEG-2, 1 GOP time is preferably 0.5 seconds or less. If the GOP configuration is M = 3 and N = 15, the appearance period M of I or P is 3 frames, and the appearance period of I. Since it can be seen that N is 15 frames, in the case of video of 30 frames per second, 1 GOP time is obtained as 0.5 seconds. In the case of an application that is severe in delay, such as a real-time and interactive video communication service, the 1 GOP time is often shortened. On the other hand, in the case of an application that is not so severe in delay, there may be a case where 1 GOP time is 3 seconds or more.

  Next, the head area TS determined in step 201 is divided into n by the number of packets (in this example, n = 10), and n GOP start packet positions (GOP time start positions) P1 to Pn are temporarily determined ( FIG. 7C: Step 202). That is, ten GOP start packet positions are temporarily determined from the total number of packets in one GOP time. For example, if the total number of packets in one GOP time is 1000, the positions of the 10 GOP start packets are “1”, “101”, “201”, “301”, “401”, “501”, “ The packets are “601”, “701”, “801”, and “901”.

  Then, the parameter indicating the order of the GOP start packet position is set to m = 1 (step 203), and from the first GOP start packet position P1, m = 1, the total number of packets within one GOP time, the packet ratio by video frame type, and the GOP configuration Based on (arrangement of video frames during 1 GOP time), the position and type of each video frame up to the end of the measurement section T are provisionally determined (step 204).

  For example, when the GOP configuration is M = 3 and N = 15, the arrangement of video frames during one GOP time is “IBBPBBPBBPBBPBB”. Here, if the total number of packets in one GOP time is 1000, and the packet ratio by video frame type of the I frame at that time is obtained as 10%, for example, the packet number for the I frame is from 1 to 100. It will be a turn. Similarly, packet numbers are obtained for P frames and B frames.

Then, based on the position and type of the video frame provisionally determined for the GOP start packet position P1, which type of video frame is the invalid packet generated in the section from the GOP start packet position P1 to the end of the measurement period T. (Step 205), and calculates the number of video frames to which the invalid packet belongs and the number of video frames that deteriorate due to the influence of the video frame as the number of invalid frames F _NG1 at the GOP start packet position P1. (Step 206).

  For example, if the generated invalid packet is the 90th packet, it is determined that it belongs to the I frame, and if it is the 101st packet, it is determined that it belongs to the B frame. The number of invalid frames is calculated based on the above, and the number of invalid frames for the invalid packets up to the end of the measurement section T is integrated.

Then, m = m + 1 = 2 is set, the process returns to step 204 (step 208), and the processes of steps 204 to 206 are repeated until m = n in step 207. As a result, the number of invalid frames F _NG2 at the second GOP start packet position P2 and the number of invalid frames F _NG3 at the third GOP start packet position P3 are shifted one by one while shifting the GOP start packet position one by one. th to GOP start packet position Pn, a total of n number of invalid frames (F _NG1 ~F _NGn) is calculated.

When the number of invalid frames in all GOP start packet position is calculated (YES in step 207), i.e. the GOP start packet position number of invalid frames F at P1 to Pn _NG1 to F _NGN is calculated, the invalid frame the number F _NG1 to F _NGN Request invalid frame number expected value than (step 209). In this example, the average value of the number of invalid frames F _NG1 to F _NGN as an invalid frame number expected value. Then, the total number of transmission frames in the measurement interval T is obtained, and the ratio of the expected number of invalid frames to the total number of transmission frames in the measurement interval T is calculated as the expected invalid frame rate (step 210).

In this example, it was to obtain the average value of the number of invalid frames F _NG1 to F _NGN as an invalid frame number expected value, to obtain the maximum value of the number of invalid frames F _NG1 to F _NGN as an invalid frame number expected value Alternatively, a 75% value (75% of the maximum value) may be obtained as the invalid frame number expected value. By setting the maximum number of invalid frames as the expected number of invalid frames, quality management can be performed assuming the worst case. By setting the 75% value as the expected number of invalid frames, quality management can be performed on the safe side.

[Video Quality Estimator]
The video quality estimation unit 16 receives the expected invalid frame rate calculated by the invalid frame rate expected value calculation unit 14 as input, and the video quality in the measurement section T according to the video quality estimation model stored in the video quality estimation model DB 15. Get an estimate.

  In this case, the video quality estimation unit 16 refers to the video encoding information from the video encoding information collection unit 11, and the video quality that meets the conditions such as the encoding method, content, and bandwidth included in the video encoding information An estimation model is selected from the video quality estimation model DB 15 (FIG. 5: step 105).

  Then, the estimated invalid frame rate value from the invalid frame rate expected value calculation unit 14 is substituted into the selected video quality estimation model to obtain an estimated value of the video quality in the measurement section T (T1) (step 106).

  In the same manner, the video quality is estimated for the next measurement section T (T2). In this case, the process returns to step 103 in response to YES in step 107, collection of invalid packet generation information in the measurement section T (step 103), calculation of expected invalid frame rate value in the measurement section T (step 104), video quality estimation The model selection (step 105) and calculation of the video quality estimation value (step 106) by substituting the expected invalid frame rate into the video quality estimation model are repeated. In this example, the video quality estimation model is a model formula, but the relationship between the invalid frame rate and the video quality may be prepared as a table.

  In this way, in the present embodiment, the position and type of each video frame up to the end of the measurement section T are temporarily determined for every n GOP start packet positions, and the position and type of each video frame determined temporarily are determined. The type of video frame to which the invalid packet belongs is judged and the video quality within the measurement section T is estimated, so the information about the video frame to which the invalid packet belongs is not obtained from the header information and payload information, and the loss is lost. Even when the packets before and after the packet are lost, the video quality can be accurately estimated.

[Embodiment 2 (Method 2)]
In the first embodiment (method 1), as can be seen from the explanatory diagram of FIG. 7, if the GOP start packet position is shifted by one, the calculation target section Tx of the number of invalid frames is gradually shortened. This is because the invalid frame count calculation target section Tx extends from the GOP start packet position to the end of the measurement section T. When the measurement interval T is considerably longer than the 1 GOP time, such as 1 GOP time of 0.5 seconds and measurement interval T of 10 seconds, the error in the number of invalid frames calculated for each GOP start packet position is small. There is little impact on the estimation accuracy of video quality. However, when the measurement section T is shortened, there is a concern about the influence on the estimation accuracy of the video quality.

  On the other hand, in the second embodiment (method 2), as shown in FIG. 10, for all GOP start packet positions, the number of invalid frames is calculated from the GOP start packet position to the time width of the measurement section T. Let it be the target section Tx. That is, for each GOP start packet position, a time span from the GOP start packet position to the measurement section T is set as a calculation target section Tx, and the number of invalid frames in the calculation target section Tx is calculated. In this case, a part of the next measurement section T is included in the calculation target section Tx of the number of invalid frames, but the calculation target section Tx of the number of invalid frames is not shortened and always has a constant time width (measurement A time width of the section T) is secured. Therefore, the use of method 2 can be expected to improve the estimation accuracy of the video quality as compared to the method 1. When the measurement section T is short, the method 2 is preferable.

  FIG. 8 shows a flowchart of the second embodiment corresponding to FIG. FIG. 9 shows a flowchart of the second embodiment corresponding to FIG. In the second embodiment, invalid packet occurrence information of “measurement interval + 1 GOP time” is collected in step 103 ′. Further, in step 204 ', the position and type of each video frame in the time width section of the measurement section T are temporarily determined for each GOP start packet position. In step 205 ′, for each GOP start packet position, it is determined to which type of video frame the invalid packet in the time width of the measurement section T belongs. In step 206 ′, the number of invalid frames in the time width of the measurement section T is calculated for each GOP start packet position.

In the above-described embodiment, the measurement interval T is 10 seconds. However, in principle, it may be 1 GOP time or more. When the 1 GOP time is as short as 0.5 seconds, for example, it is better to make the measurement target period considerably longer than the 1 GOP time. In addition, when 1 GOP time is long, for example, 8 seconds, it is desirable to set it to 2 GOP time or more.
In the above-described embodiment, the head region TS is divided into ten parts, but is not limited to ten parts, and may be two or more parts.
In the above-described embodiment, the case where the video quality at the interactive user terminal 20 is estimated has been described as an example. However, the video quality at the video distribution terminal 40 is similarly estimated.
Further, the flowcharts shown in FIGS. 5, 6, 8, and 9 are merely examples, and it goes without saying that various modifications are possible.

1 is a system configuration diagram showing an example of a video communication service system including a video quality estimation apparatus according to the present invention. It is a block diagram which shows the principal part of the internal structure of the video quality estimation apparatus in this video communication service system. It is a figure which shows an example of a video quality estimation model (relationship between an invalid frame rate and video quality). It is a figure which shows the relationship between the total packet number or band in GOP, and the structure packet ratio according to video frame kind. It is a flowchart which shows the process of the estimation process (Embodiment 1 (method 1)) of the video quality in a video quality estimation apparatus. 12 is a flowchart illustrating details of an invalid frame rate expected value calculation process in a video quality estimation process (Embodiment 1 (method 1)). 10 is a time chart for explaining calculation processing of expected invalid frame rate values in video quality estimation processing (Embodiment 1 (method 1)). It is a flowchart which shows the process of the estimation process (Embodiment 2 (method 2)) of the video quality in a video quality estimation apparatus. 10 is a flowchart showing details of an invalid frame rate expected value calculation process in a video quality estimation process (Embodiment 2 (Method 2)). 10 is a time chart for explaining calculation processing of an expected invalid frame rate value in video quality estimation processing (Embodiment 2 (Method 2)). It is a figure which shows an example (in the case of M = 3, N = 15) of GOP structure.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 ... Video quality estimation apparatus, 11 ... Video coding information collection part, 12 ... Invalid packet generation | occurrence | production information collection part, 13 ... Composition packet ratio calculation part according to video frame type, 14 ... Invalid frame rate expected value calculation part, 15 ... Video Quality estimation model DB, 16 ... video quality estimation unit, T ... measurement section (measurement target section), TS ... head region, Tx ... calculation target section, P1 to Pn ... GOP start packet position.

Claims (7)

In a video quality estimation apparatus for estimating video quality in a video communication service between communication terminals via a network,
Video coding information collecting means for collecting video coding information including configuration information of video frames in one unit time corresponding to a group of encoded video frames from the communication terminal;
One configuration in which the section of one unit time duration from the beginning of the video quality measurement target section is set as a head region, and the head region is divided into n (n ≧ 2) to tentatively determine start positions of n one unit time. A unit time start position determining means;
Temporary setting means for temporarily determining the position and type of each video frame after the start position based on the video encoding information for each start position of the n unit time units,
Invalid packet occurrence information collecting means for collecting occurrence information of invalid packets in the network and the communication terminal;
For each start position of the n unit time units, based on the position and type of the video frame provisionally determined by the temporary setting means, which type of video frame the invalid packet generated after the start position is Invalid frame number calculating means for determining whether it belongs, and calculating the number of video frames to which the invalid packet belongs and the number of video frames deteriorated under the influence of the video frame,
A video quality estimation device comprising: video quality estimation means for estimating the video quality in the measurement target section based on the number of invalid frames calculated for each of the n start positions of one unit time. The video quality estimation apparatus according to claim 1,
The invalid frame number calculating means includes:
For each start position of the one unit time, a section from the start position to the end of the measurement target section is used as a calculation target section, and the number of invalid frames in the calculation target section is calculated. apparatus. The video quality estimation apparatus according to claim 1,
The invalid frame number calculating means includes:
For each start position of the n unit time units, a time width of the measurement target section from the start position is set as a calculation target section, and the number of invalid frames in the calculation target section is calculated. Video quality estimation device. In the video quality estimation device according to any one of claims 1 to 3,
The temporary setting means includes
The ratio of the total number of packets for each video frame type to the total number of packets in one structural unit time is obtained as a constituent packet ratio for each video frame type, and the total number of packets in one structural unit time, the constituent packet ratio for each frame type And, based on the arrangement of the video frames during the one unit time, the position and type of each subsequent video frame are provisionally determined for each start position of the n one unit time. apparatus. In the video quality estimation device according to any one of claims 1 to 3,
The video quality estimation means includes
An average value of the number of invalid frames calculated for each start position of the n unit time units is defined as an invalid frame number expected value, and a ratio of the invalid frame number expected value to the total number of transmission frames in the measurement target section is expressed as follows. Estimating the invalid frame rate, and substituting the expected invalid frame rate into a video quality estimation model indicating the relationship between the predetermined invalid frame rate and video quality, and estimating the video quality within the measurement target section A video quality estimation apparatus characterized by obtaining a value. In the video quality estimation device according to any one of claims 1 to 3,
The video quality estimation means includes
The maximum value of the number of invalid frames calculated for each start position of the n unit time units is set as an invalid frame number expected value, and the ratio of the invalid frame number expected value to the total number of transmission frames in the measurement target section is expressed as follows. Estimating the invalid frame rate, and substituting the expected invalid frame rate into a video quality estimation model indicating the relationship between the predetermined invalid frame rate and video quality, and estimating the video quality within the measurement target section A video quality estimation apparatus characterized by obtaining a value. In a video quality estimation method for estimating video quality in a video communication service between communication terminals via a network,
A first step of collecting, from the communication terminal, video encoding information including configuration information of video frames in one unit time corresponding to a group of encoded video frames;
A section having the one unit unit time width from the beginning of the video quality measurement target section is set as a head area, and the head area is divided into n (n ≧ 2) to temporarily determine the start positions of n number of one unit time. Steps,
A third step of tentatively determining the position and type of each video frame after the start position based on the video encoding information for each start position of the n unit time units;
A fourth step of collecting occurrence information of invalid packets in the network and the communication terminal;
For each start position of the n unit time units, based on the position and type of the video frame provisionally determined by the temporary setting means, which type of video frame the invalid packet generated after the start position is A fifth step of determining whether it belongs and calculating the number of video frames to which the invalid packet belongs and the number of video frames deteriorated under the influence of the video frame as the number of invalid frames;
A video quality estimation method comprising: a sixth step of estimating video quality in the measurement target section based on the number of invalid frames calculated for each start position of the n unit time units.

Images