JP2005176069A - Distributed parallel transcoder system and distributed parallel transcoding method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To keep high speed and prevent a deterioration of image quality in image reproduction in a distributed parallel transcoder system and a distributed parallel transcoding method for dividing moving image data per sequence and transcoding in parallel processes. <P>SOLUTION: The distributed parallel transcoder system, in which moving image data are separated per sequence and the transcoding operation in parallel is carried out, includes a segment dividing unit for dividing the motion image data into two or more segments in unit of sequence, two or more video image coding unit for transcoding each segment, and a coupling unit for carrying out coupling process of each segment which was encoded by two or more video imaging coding units at a scene change unit. In this case, detection process of scene change is carried out by each video coding unit. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a distributed parallel transcoder system and a distributed parallel transcoding method in which video is divided into sequence units and transcoding is performed in parallel processing.

Some conventional transcoders divide video in sequence units and perform parallel processing on the divided image data by a plurality of personal computers to increase the speed.
However, in the conventional transcoder described above, buffer underflow occurs at the time when image data processed in parallel is combined, and quality degradation with respect to reproduced video becomes a problem.
As an improvement measure for preventing this quality degradation, a method for preventing quality degradation by using a scene change portion as a division pointer has been proposed (see, for example, Patent Document 1).
JP 2002-199392 A

However, the video encoding method disclosed in Patent Document 1 has the following problems when video (moving image data) is divided in consideration of each scene using the scene change described above. Yes.
That is, in the conventional video encoding method, since the scene change determination method differs depending on the encoder, it is not possible to perform uniform division among the encoders.
In the conventional video encoding method, the encoding parameters are not always completely reset before and after the scene change, and the moving image data cannot be reproduced in the original state.
Further, in the conventional video encoding method, scene detection processing for determining a scene change becomes a bottleneck, and high-speed encoding cannot be performed.

In addition, in the conventional video encoding method, even if unified scene change detection is performed, the resulting division point is not necessarily a good division point for the encoding method.
Originally, in an encoding method in which information is compressed using redundancy in the time axis direction, optimal encoding can be realized by continuously encoding moving image data.

On the other hand, as in the conventional video encoding method, moving image data (video) is divided within a predetermined range and transcoded, so that optimal encoding cannot be performed, and the video quality of the reproduced moving image is reduced. This may cause deterioration.
Furthermore, in the conventional video encoding method, when the joint process is simply performed on the encoded streams that have been encoded separately by transcoding, the joint portion is not properly encoded, and the arithmetic process is performed during reproduction. Inconvenience occurs, and the situation that the memory buffer occupation amount of the decoder exceeds a specified value may occur, resulting in a defective stream in which the divided moving image data cannot be decoded.

  The present invention has been made in view of such circumstances. In a distributed parallel transcoder that divides moving image data in sequence units and performs transcoding in parallel processing, reproduction of moving image data is performed while maintaining high speed. An object of the present invention is to provide a distributed parallel transcoder system and a distributed parallel transcoding method that prevent deterioration of image quality in the system.

  The distributed parallel transcoder system of the present invention is a distributed parallel transcoder system that divides moving image data into sequence units and transcodes them in parallel, a segment dividing unit that divides moving image data into a plurality of segments in sequence units, A plurality of video encoding units that transcode each segment; and a combination unit that performs a process of combining each segment encoded by the plurality of video encoding units in a scene change portion, and Each unit performs a scene change detection process.

  The distributed parallel transcoder system of the present invention is characterized in that the video encoding unit detects a scene change in the segment in parallel with the transcoding process.

  In the distributed parallel transcoder system according to the present invention, the video encoding unit encodes the segment in combination with an overlapping portion obtained by dividing a part of a segment processed by another video encoding unit. .

In the distributed parallel transcoder system of the present invention, the video encoding unit processes another video encoding in which the segment immediately before the scene change is used as the overlapping unit in the segment. It transfers to a part.
In the distributed parallel transcoder system according to the present invention, the combining unit discards the overlapping portion from the segments transferred to the immediately preceding segment and combines the segments.

  The distributed parallel transcoder method of the present invention is a distributed parallel transcoder method that divides moving image data into sequence units and transcodes them in parallel. A segment dividing process in which the moving image data is divided into a plurality of segments in sequence units. And a transcoding process in which each segment is transcoded by a plurality of video encoding units, and a coupling process in which a combination process of each segment encoded by the plurality of video encoding units is performed in a scene change portion Each of the video encoding units performs scene change detection processing.

  The distributed parallel transcoder method of the present invention is characterized in that in the transcoding process, the video encoding unit detects a scene change in the segment in parallel with the transcoding process.

  In the distributed parallel transcoder method of the present invention, in the transcoding process, the video encoder encodes a combination of the segment and an overlapping part obtained by dividing a part of a segment processed by another video encoder. It is characterized by that.

  In the distributed parallel transcoder method of the present invention, the duplication unit transfers the data before the scene change from the head data in each segment to another video encoding unit that processes the immediately preceding segment. It is characterized by.

  The distributed parallel transcoder method of the present invention is characterized in that, in the combining process, the frame data of the overlapped portion is discarded from the segment transferred to the immediately preceding segment, and the segments are combined.

As described above, according to the present invention, since the scene change is performed by the video encoding unit that performs the transcoder, the scene change detection is performed in parallel with the transcoding, so that the time of the scene change that takes time is distributed. can do.
In addition, in the distributed parallel transcoding system of the present invention, in the transcoding process, the scene change is used as a reference position, frame data is duplicated in the preceding and succeeding segments, and each segment is encoded. Therefore, refer to the image information of the immediately preceding frame. However, since it is possible to perform code allocation of necessary frames, this code allocation can be optimized, and the combined server can be used for calculation processing during playback, that is, decoding and synthesis processing on the combined server side. It is possible to prevent the buffer overflow of the memory inside.
Further, according to the present invention, a time-series continuous segment input to the combining unit has compressed frame data that overlaps with the preceding and following segments with the scene change as a reference position, that is, a scene change. The frame data is duplicated so that the frame data from the scene change to the next scene change is performed by the same video encoding device so that the position of Thus, even if transcoding is performed, it is possible to obtain the same moving image reproduction quality as in the case of processing alone without distributed processing.
Accordingly, the present invention can increase the speed by performing distributed processing of the transcode and improve the reproduction quality of the moving image as compared with the conventional example.

Furthermore, according to the present invention, the encoded frame data is once decoded and converted into unencoded frame data, and then re-encoded with another format. Therefore, it is not necessary to carry out complicated control specialized for each encoder method. Therefore, by installing an application for encoding processing into a format format according to necessity for each video encoding unit, It is possible to easily cope with processing of a simple transcoder.
Furthermore, according to the present invention, as described above, since each video encoding unit is once decoded, it is possible to easily cope with hardware encoders that are difficult to customize and software encoders that are already installed. Can do.

Hereinafter, a distributed parallel transcoder system according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the embodiment.
In this figure, the source server 1 distributes, for example, compressed moving image data as content to the user's combined server 2 via the network I.
The source server 1 also includes a segment dividing unit 3 that divides the compressed moving image data into a plurality of segments in sequence units.
Here, the sequence indicates the time-series arrangement order of the image data, and the sequence unit is composed of a predetermined segment (a predetermined number of GOPs) divided in units of GOP (Group Of Picture) in the arrangement order of the image data. Is the length).
The video encoding unit 4 once decodes the above-mentioned segment obtained by dividing the moving image data input via the network I, and re-encodes the code corresponding to the encoding method used by the user. The re-encoded segment is output to the combined server 2 via the network I.

Here, the network I is a network composed of a public information network, a dedicated information network, the Internet, and the like.
A feature of the present invention resides in that the video encoding unit 4 re-encodes a predetermined portion in a segment preceding the segment together with the segment transmitted.
The predetermined part of the segment (hereinafter referred to as an overlapping part) indicates a part temporally preceding the scene change in the data of the preceding segment.
Further, when the merge server 2 performs the process of combining the input compressed segments, the merge server detects the position of the scene change in the continuous segment, that is, the overlapping portion of the subsequent segment and the previous segment, that is, The data before the scene change of the subsequent segment on the time axis is discarded, and the segments are sequentially connected to reproduce the moving image data.

Next, the configuration of the video encoding unit 4 in FIG. 1 will be described with reference to FIG. FIG. 2 is a block diagram illustrating a configuration example of the video encoding unit 4.
The data receiving unit 11 inputs the segment from the source server 1 via the network I, and stores the input segment in the segment file storage unit 12.
The overlapping part receiving unit 13 inputs the overlapping part from the other video encoding part 4 that processes the subsequent segment via the network I, and stores this overlapping part in the segment file storage part 12.

The decoder unit 14 performs decoding (decoding) in a state where the segments and overlapping portions stored in the segment file storage unit 12 are connected, and outputs the result to the encoder unit 15.
The encoder unit 16 re-encodes the decoded segments and sequentially stores them in the encoded file storage unit 19.
The scene change information acquisition unit 16 reads the uncompressed frame data before the decoded segment (consisting of uncompressed frame data) is encoded, and changes the scene according to a change between successive frame data. When a change part is detected and a scene change is detected, a detection signal indicating the detection is output to the division processing unit 17.

In addition, the scene change acquisition unit 16 uses the point (position) where the intra-coded frame is inserted as a result of encoding as a scene change, for example, when an intra-coded frame is inserted by a scene change depending on the encoding method. You can also
Further, for example, when the format used for re-encoding is WMV (registered trademark), the scene change information acquisition unit 16 similarly detects a scene change based on the decoding result, and the decoding result of the frame data is the key frame. If so, this data frame is determined to be a scene change.

The division processing unit 17 receives segment data in synchronization with the decoder unit 14, temporarily holds these data, and when the detection signal is input, the detection signal is output. The data stored before the transmission is transmitted as an overlapping portion to the other video encoding unit 4 that processes the previous segment via the data transmission unit 18 and the network I.
The encoded file transmission unit 20 sequentially transmits the segment data stored in the encoded file storage unit 19 to the combined server 2 via the network I.

Also, in the distributed parallel transcoder system of the present invention, a case where a scene change is detected with emphasis on image quality and transcoder processing using an overlapping portion is performed, and a case where a time change is made with emphasis on speed. The user may be able to select the case where the transcoder is performed only by the divided segment units without detecting the above.
For example, information on whether or not to perform scene change detection is added to each segment, and based on this information, the transcoder determines whether to perform processing focusing on image quality or speed. To do.

  In addition to the configuration described above, when the scene change detection signal is input from the scene change acquisition unit 16, the duplication unit receiving unit 13 transmits to the source server 1 the duplication unit of the part corresponding to this scene change. May be configured to receive the duplicate portion to be retransferred and store the duplicate portion in the segment file storage unit 12. Other configurations are the same as those already described.

Next, the operation of the embodiment of the present invention will be described with reference to the drawings. As an example, a transcoding process from MPEG-2 to an encoding method using a scene change as an intra frame (key frame), for example, WMV (registered trademark, Windous Media Video) will be described. (The moving image data is stored in the distribution source in MPEG-2, and the user requests the moving image data encoded in the WMV (registered trademark) format.)
The source server 1 reads out the moving image data requested by the user from the MPEG-2 master file database, and uses the moving image data as a partial file (segment) having a predetermined length in units of GOP. 3, it is divided into segments each composed of L [GOP] (consisting of L GOPs), and each segment is transmitted to the video encoding unit 4.

Here, the predetermined length L [GOP] indicates the length of the processing data, and the personal computer or server used as the video storage unit 4 from the scheduling (processing plan) set by the internal scheduler. Is calculated by the segment dividing unit 3 based on the processing capability (calculation speed and communication speed). Further, as described above, the segment is composed of a plurality of GOPs, and the GOP is never divided.
Further, the scheduler of the source server 1 assigns the segment data amount to each video encoding unit 4 based on the bit rate information of the MPEG-2 moving image data before decoding.
In addition, when the source server 1 divides the moving image data into segments in sequence units, a method of dividing the moving image data into segments having a predetermined length including at least one scene change in advance and transferring the segments may be considered. .

Next, the video encoding unit 4 decodes the input segment, newly re-encodes it as a different encoding format MWV, and transmits it to the combining server 2 as a segment compressed in the MWV format.
Here, the transcoding process in the video encoding unit 4 will be described in detail with reference to FIGS. 2 and 4. FIG. 2 is a flowchart showing an operation example of the video encoding unit 4.
For example, it is assumed that four video encoding units 4, 4 2, 4 3, 4 4 are used by the scheduler of the source server 1 based on the volume of moving image data to be distributed. Segments C1, C2, C3, and C4 are divided in this order, and are transmitted to video encoding units 41, 42, 43, and 44, respectively, as shown in FIG. In the following description, the video encoding unit 42 is taken as an example.

The data receiving unit 11 inputs a segment compressed in MPEG-2 (MPEG-2 format frame data) via the network I, and stores this segment in the segment file storage unit 12 (step S1).
Then, the data receiving unit 13 inputs an overlapping part (additional frame data in MPEG-2 format) F3 from another video encoding unit 4 that performs transcoding of the subsequent segment, that is, the video encoding unit 43, Store in the segment file storage unit 12 (step S2).

Here, in the video encoding unit 4 3, when the scene change information acquisition unit 16 detects a key frame when the encoder unit 15 re-encodes the decoded segment C 3, The division processing unit 17 divides the overlapping portion F3 (for example, A [GOP]) from the segment C3, and transmits this overlapping portion F3 to the video encoding unit 42. Each overlapping portion is configured by a plurality of GOP units similarly to the segment.
Then, the division in the division processing unit 17 of the video encoding unit 4 3 is performed with the overlapping unit F 3 from the first GOP of the segment C 3 processed by the video encoding unit 4 3 to the GOP including the scene change (A-th GOP). To do.

Next, returning to the description of the operation of the video encoding unit 42, the decoder unit 14 sequentially decodes the MPEG-2 segment C2 and the overlapping unit F3 from the encoded frame data in the MPEG-2 format, and performs normal encoding. The frame data is converted into uncompressed (uncompressed) frame data (step S3).
At this time, the decoder unit 14 performs a decoding process on the portion that can be decoded, that is, the segment C2, even if the overlapping unit F3 is not transmitted from the video encoding unit 43.
The encoder unit 15 performs encoding (compression processing) in units of frame data, and when the uncompressed frame data decoded by the decoder unit 14 is input, the input data is sequentially converted into MPEG- Encoding (encoding process) is performed in a format different from 2 (step S4).

Here, in the transcoding process, when it is specified to convert to a plurality of formats, after the decoding process is performed once, this data is temporarily stored in the segment file storage unit 12 or the like. Based on the above, encoding processing in a plurality of formats is performed in order.
At this time, the sea change information acquisition unit 16 inputs uncompressed frame data from the decoder unit 14 to the encoder unit 15, reads the re-encoded compressed frame data, detects the presence or absence of a key frame indicating a scene change, A detection signal indicating the presence or absence of a scene change is output (step S5).

Next, when the detection signal is input, the division processing unit 17 determines whether or not the detection signal has detected a scene change for each GOP unit in the segment C2 (step S7).
At this time, when the division processing unit 17 detects that there is no scene change from the detection signal, the division processing unit 17 proceeds to step S10, and when it detects that there is a scene change, the division processing unit 17 proceeds to step S8.

Then, the division processing unit 17 recognizes that the current GOP (Bth GOP) is a scene change part, and the frame data input before the scene change, that is, the first GOP to the Bth B [GOP] is extracted as the overlap portion F2, is divided from the segment C2 (step S8), and the overlap portion F2 and the scene change are detected by the video encoding portion 41 that transcodes the segment C1 in the previous stage. Position information indicating which frame of the B-th GOP is a scene change is transmitted (step S9).
Next, the decoder unit 14 detects whether or not it is the last GOP (Lth GOP) of the segment (there is no frame data thereafter). If it is determined that it is not the last GOP, the process proceeds to step S3. If the final GOP is detected, the process proceeds to step S11 (step S10).

Then, the encode file transmission unit 20 stores the re-encoded segment K2 (transcoded segment C2 and the overlapped part F3), which are stored in the encode file storage unit 19 and encoded in the MWV format after transcoding. The segment (corresponding to (L + A) [GOP] and re-encoded segment) is transmitted to the combination server 2 of the user of the distribution request destination via the network I (step S11).
Here, the encoding file transmission unit 20 sequentially distributes the encoded data input to the encoding file storage unit 19 to the combined server 2 after the encoding is completed.

That is, encoding by the encoder unit 15 and transmission of encoded data by the encoded file transmission unit 20 are performed in parallel.
Then, the video encoding unit 42 ends the transcoding process.
The other video encoding units 41, 43, and 44 also perform transcoding processing on the input segments C1, C3, and C4, respectively, and re-encode each, similarly to the video encoding unit 42 described above. The segments K1, K3, and K4 are transmitted to the combined server 2.

Next, with reference to FIG. 6, a description will be given of the process of combining recoded segments K1 to K2 in the combined server 2.
The combination server 2 performs a combination process of the re-encoded segments K1, K2, K3, and K4 input from the video encoding units 41, 42, 43, and 44 via the network I.
At this time, when a server that can perform a combining process from input segments, that is, a segment that is continuous in the time axis direction is input, the combining server performs a combining process of these segments.

For example, when the re-encoded segments K2 and K3 are input, the combining server 2 starts the combining process of the two re-encoded segments.
When the combined server 2 detects that the re-encoded segment K3 is a subsequent segment in the time axis direction, it reads the position information of the scene change data frame added to the re-encoded segment K2, and re-encodes it. The scene change position in the segment K3 is detected.

Next, the combining server 2 removes from the re-encoded segment K3 to the frame data immediately before the compressed frame data of the scene change from the first compressed frame data.
Then, the combining server 2 combines the re-encoded segment K3 from which the frame data corresponding to the overlapping portion F3 has been removed with the last frame data of the re-encoded segment K2, and re-encoded segment K2. And the process of combining K3 is terminated.
That is, in the re-encoded segment K3, the frame data included from the start frame data to immediately before the frame data of the scene change detected first is also included in the re-encoded segment K2.
Accordingly, in the re-encoded segment K3, the compressed frame data from the first frame data to the scene change is obtained by combining and reproducing the moving image data by removing the encoded portion from the halfway portion between the scene changes. It is possible to prevent a decrease in image quality.

Further, when the source server 1 divides the moving image data into segments, at this time, the source server 1 divides the moving image data into segments so as to have an overlapping portion of a predetermined size, and transmits the segment to each video encoding unit 4. It may be configured.
As a result, when the overlap portion has a scene change, when a scene change is detected in each video encoding unit 4, only the scene change information indicating the detection is processed in the immediately preceding segment. Therefore, it is not necessary to transfer the segment data, and the data transfer efficiency in the network can be improved. Even when there is no scene change in the overlapping portion, the data transfer amount can be reduced.

  A program for realizing the functions of the source server 1, the video encoding unit 4 and the combined server 2 in FIG. 1 is recorded on a computer-readable recording medium, and the program recorded on the recording medium is stored in a computer system. The functions of each unit may be realized by reading and executing. Here, the “computer system” includes an OS and hardware such as peripheral devices. The “computer system” includes a WWW system having a homepage providing environment (or display environment). The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Further, the “computer-readable recording medium” refers to a volatile memory (RAM) in a computer system that becomes a server or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. In addition, those holding programs for a certain period of time are also included.

  The program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, and what is called a difference file (difference program) may be sufficient.

It is a block diagram which shows the structural example of the distributed parallel transcoder system by one Embodiment of this invention. It is a block diagram which shows the structural example of the video encoding part 4 in FIG. It is a conceptual diagram explaining the process which divides | segments moving image data into a sequence segment. 4 is a flowchart illustrating an operation example of a video encoding unit 4. It is a conceptual diagram for demonstrating the division | segmentation process of the duplication part by the image | video encoding part 4. FIG. It is a conceptual diagram for demonstrating the joint process of the re-encoding segment by the combination part.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Source server 2 ... Combined server 3 ... Segment division part 4 ... Video encoding part 11 ... Data receiving part 12 ... Segment file storage part 13 ... Duplicate part receiving part 14 ... Decoder part 15 ... Encoder part 16 ... Scene change information acquisition Unit 17 ... Division processing unit 18 Data transmission unit 19 Encoding file storage unit 20 Encoding file transmission unit I Network

Claims (10)

In a distributed parallel transcoder system that divides moving image data into sequence units and transcodes them in parallel,
A segment dividing unit for dividing moving image data into a plurality of segments in sequence units;
A plurality of video encoding units that transcode each of the segments;
A combining unit that performs combining processing of each segment encoded by the plurality of video encoding units in a scene change portion;
Each of the video encoding units performs a scene change detection process.   The distributed parallel transcoding system according to claim 1, wherein the video encoding unit detects a scene change in the segment in parallel with the transcoding process.   3. The distribution according to claim 1, wherein the video encoding unit encodes the segment in combination with an overlapping portion obtained by dividing a part of a segment processed by another video encoding unit. Parallel transcoding system.   The video encoding unit transfers, in the segment, data before a scene change from the head data to another video encoding unit that processes the immediately preceding segment as the overlapping unit. Item 4. The distributed parallel transcoding system according to Item 3.   5. The distributed parallel transcoding system according to claim 4, wherein the combining unit discards the overlapping portion from the segment transferred to the immediately preceding segment and combines the segments. In a distributed parallel transcoder method for dividing moving image data into sequence units and transcoding them in parallel,
A segmentation process in which the moving image data is divided into a plurality of segments in sequence units;
A transcoding process in which each segment is transcoded by a plurality of video encoding units;
A process of combining the segments encoded by the plurality of video encoding units includes a combining process performed in a scene change portion;
A distributed parallel transcoder method, wherein scene change detection processing is performed in each of the video encoding units.   7. The distributed parallel transcoding method according to claim 6, wherein in the transcoding process, the video encoding unit detects a scene change in the segment in parallel with the transcoding process.   7. The transcoding process, wherein the video encoding unit encodes the segment and an overlapping part obtained by dividing a part of a segment processed by another video encoding unit in combination. The distributed parallel transcoding method according to claim 7.   9. The overlapped portion according to claim 8, wherein the overlapping portion is data before the scene change from the top data in each segment, and forwards it to another video encoding portion that processes the immediately preceding segment. Distributed parallel transcoding method. 5. The distributed parallel transcoding method according to claim 4, wherein, in the combining step, the frame data of the overlapping portion is discarded from the segment transferred to the immediately preceding segment, and the segments are combined.

Images