KR101396948B1 - Method and Equipment for hybrid multiview and scalable video coding - Google Patents

Abstract

The present invention proposes a hybrid video coding (HMSVC or Hybrid Multiview and Scalable Video Coding) method combining SVC or Scalable Video Coding and MVC or Multi-view Video Coding. In multi-view video coding, much resources are used for encoding, transmitting, and decoding by transmitting all video images obtained from various angles. The present invention proposes a method of providing sufficient quality for the displayed view and only limited quality for the other views. Thus, multiple views that are encoded with limited quality are encoded and decoded using MVC, and selected views can display enhanced layers using SVC. Here, the quality enhancement is made possible using spatial scalability, temporal scalability, quality scalability, or a combination of these methods. This layering method can use the layering method provided by the JVT SVC in one embodiment.

The present invention includes a method for enabling multi-view video service using few resources using a method and apparatus for simultaneously transmitting multiple views using MVC and displaying selected views at high quality.

The present invention is utilized for resource saving and application diversity in multi-view video services.

MPEG, JVT, scalable video coding (SVC), multi-view video coding (MVC), network adaptation layer (NAL)

Description

≪ Desc / Clms Page number 1 > METHOD AND APPARATUS FOR HYBRID DISTRIBUTION &

Figure 1. Example of prediction structure of multi-view video coding

Figure 2. HMSVC scheme in which 8 QCIFs and one CIF are decoded in parallel

Figure 3. HMSVC method with four QCIFs and one CIF

Figure 4. Real-time encoded / decoded 1: 1 interactive service

Figure 5. Random access without delay when streaming

Figure 6. Random access without degradation of image quality.

Figure 7. Random access using SI / SP

Figure 8. Concatenated MVC / SVC NAL header

Figure 9. Usage scenario of the present invention

Figure 10. Block diagram of streaming system using HMSVC encoding / demultiplexer, MANE

Figure 11. Example of streaming system using HMSVC

The present invention relates to a composite video coding method and apparatus that combines layered video coding and multi-view video coding.
Multi-view Video Coding (MVC), which is currently under standardization, improves coding performance based on MPEG-4 part 10 Advanced Video Coding (AVC), which is an existing video standard. And performs hierarchical B-pictures coding, which is a method performed in order to support temporal scalability in JSVC (Joint Scalable Video Coding) in the time direction, and also performs inter-view coding Prediction is performed to improve the coding.

FIG. 1 shows a prediction structure when a size of a group of pictures (GOP) in a time direction is 8, for example, in the case where eight views exist in multi-view video coding in which standardization is currently underway. S0, S1, S2, S3, ... , S7 each represent one view, and T0, T1, T2, T3, ... , And T100 represents an image in the time direction.

Normally, all views have a reference relationship with each other at the same time, and a view that can be decoded independently without reference to another view is called a base view. Frames belonging to each view have a reference relationship with respect to each other in terms of time, and periodically insert an anchor for random access and error propagation interruption. Here, an anchor refers to a frame that can be decoded independently of previous data with respect to time. In the figure, S0 is the default view, and the frames at T0, T8, and T96 are anchor frames.

In FIG. 1, in the coding in the temporal direction of each view, prediction is performed using a hierarchical B-picture structure, and images in the first time (T0) of each view are continuously (I.e., the images continuously separated by the GOP size in the temporal direction) (i.e., T8, T16, T24, ...) are only performing prediction from neighboring views. That is, S2 view is from S0 view, S1 view is from S0 view and S2 view, S4 view is from S2 view, S3 view is from S2 view and S4 view, S6 view is from S4 view, S5 view is S4 view and S6 view , Since the S7 view is the last view, the prediction is performed from the S6 view. And every other second view (i.e., S1, S3, S5, S7) performs prediction from the neighboring view along with temporal prediction. That is, the S1 view predicts from the S0 view and the S2 view, the S3 view from the S1 view and the S4 view, and the S5 view from the S3 view and the S6 view.

개의 매크로블럭을 복호화해야 하지만, 8개의 뷰는 QCIF 영상(176X144)로 하고, 하나의 뷰만 CIF 영상으로 한다면

개의 매크로블럭만 복호화하므로 계산량이 약 3/8으로 줄어든다.Since the eight views are decoded simultaneously, a very large amount of computation is required. It is an object of the present invention to reduce the amount of computation and transmission by reducing the quality of views other than the selected view. For example, if eight CIF images (352X288) are decoded at a frame rate of 30 Hz,

Macroblocks should be decoded, but the QCIF image (176X144) is used as the eight views, and if only one view is used as the CIF image

Since only macroblocks are decoded, the calculation amount is reduced to about 3/8.

개의 매크로블럭을 복호화해야 하지만, 8개의 뷰는 QCIF 영상(176X144)로 하고, 하나의 뷰만 CIF 영상으로 한다면

개의 매크로블럭만 복호화하므로 계산량이 약 3/8으로 줄어든다. 즉, 3개의 CIF급 비디오를 복호화하는 것과 계산량이 비슷하다. 만일 기본 계층을 QCIF로 하고, 향상계층을 4CIF로 한다면 계산량은 약 3/16으로 줄어들어서 4CIF 하나를 복호화 하는 것에서 50%만 계산량을 추가하면 된다.It is an object of the present invention to reduce the amount of calculation and the amount of transmission by reducing the quality of the view other than the view selected on the receiving side for the display. For example, if eight CIF images (352X288) are decoded at a frame rate of 30 Hz,

Macroblocks should be decoded, but the QCIF image (176X144) is used as the eight views, and if only one view is used as the CIF image

Since only macroblocks are decoded, the calculation amount is reduced to about 3/8. In other words, the computational complexity is similar to that of decoding three CIF videos. If the base layer is QCIF and the enhancement layer is 4CIF, the calculation amount is reduced to about 3/16, and only 50% of the calculation of 4CIF is added.

According to the present invention, the layer between the base layer and the enhancement layer can be a spatial scalability, a temporal scalability, a quality scalability, or a combination of these methods. However, in order to simplify the description, only one spatial layering layer is used as an embodiment of the present invention.

It is noted that there are many views being decoded simultaneously in the multi-view video service, but only one view is displayed. Depending on how much the random access (view switching) latency is allowed, there may be a way to save resources such as calculation amount and buffer amount.

For example, FIG. 2 shows a case where 8 QCIFs and one CIF are decoded simultaneously. At this time, 8 QCIFs are decoded by the MVC method, and one QCIF is a base layer of spatial layer decoding of CIF. The view from which the CIF image is restored is displayed and defined as display_view. This saves resources compared to MVC decoding of 8 CIFs at the same time, but there are problems such as random access delay and increased complexity. In the present invention, a coding and decoding method that minimizes such a problem is presented, and the necessary syntax is defined. Thus, the decoding method to be used again, the multi-view video decoding and layered video decoding at the same time HMSVC (Hybrid Multiview and Scalable Coding ) .

If only the minimum amount of data needed to decode the display_view is transmitted, it will save resources. Assume that random access is performed between MVC and QCIF when four QCIF streams and one CIF are decoded as shown in FIG. The reference relation is as shown in the figure, where S1 and S2 refer to S0, and S3 refers to S2. Assume that the CIF is decoded only in the displayed view, i.e., display_view. At this time, if view 2 is display_view, only QCIF data of S0 and S2 and CIF data of view 2 are transmitted and decoded. This saves resources such as computation volume, transfer rate, and decoding buffer memory. When the user requests switching from view 2 to view 1, decoding is not possible until the next anchor (until the start of the next GOP) because view 1 is not being decoded.

Application services can be classified into the following categories.

1. Is it encoded / decoded simultaneously in real time? (Example) Video call, video conference, sports relay

2. Can the server not listen to individual feedback requests because they are multicast? (Example) Sports broadcasting

3. Is it streaming already encoded content? (Example) VOD

Random access delay, image quality, and resource saving are closely related to each other, and this relationship can be different depending on the type of application service.

1. Real-time encoding / decoding

In real-time, they are simultaneously encoded / decoded. In the case of 1: 1 transmission, they can be implemented simply as one AVC or SVC. View switching can also be done simply by inputting the input image of the encoder in a different view. The inter-image reference mode can be applied without any discrimination even when switching. So in this case MVC is virtually unnecessary.

2. Streaming encoded video

And the already encoded multi-point sequence is streamed. The following scenarios also apply when multicasting in real time encoding / decoding. Only the enhancement layer is transmitted for the selected view. The base layer decodes all views, and only the enhancement layer for display_view is transmitted. In view switching, the base layer immediately switches to view, but in the enhancement layer, intra refresh (Intra or Enhanced Intra) must occur before view switching is possible.

If there is no degradation in quality and the view switching delay is allowed, the previous view is continuously displayed before the enhancement layer intra refresh as shown in FIG. Enhancement layer data of the previous frame should be transmitted during the view switching request and the actual switching intermediate period.

If the quality degradation is allowed and the view switching delay is not allowed, the base layer is immediately switched as shown in FIG. 5, and the enhancement layer upsamples and displays the base layer. The enhancement layer data of the previous view need not be transmitted during the view switching request and the actual switching intermediate period.

Since coding efficiency deteriorates frequently when intra refresh is frequently performed, it is desirable to reduce the view switching delay while maintaining the coding efficiency to some extent by using SI (Switching Intra) or SP (Switching Predictive) in the middle as shown in FIG.

HMSVC decoding process

1. MVC Decryption

In the base layer, there is no change in the existing MVC. However, information for recognizing the currently displayed view is necessary on the decoding side. Let us say display_view in one embodiment. Knowing the display_view in the side or the transmission end to be encoded by the encoder, only the data necessary for decoding the display_view can be selectively provided to the decoding side. For example, in FIG. 1, if view 4 is display_view, view 1 and view 3 do not need to be encoded. However, it is not allowed to perform random access on a GOP-by-GOP basis, and all views must be decoded simultaneously if completely random random access is desired.

reference Picture list management : When following the existing MVC inter-view prediction structure, the reference picture list management is the same as the existing method

2. SVC Decryption

Only the selected view among the multiple views performs enhancement layer decoding. The enhancement layer can select one or all. Among the enhancement layers, the view-to-view reference may be different from the view-to-view reference of the base layer. In one embodiment, the enhancement layer preferably does not have a cross-view reference.

View switching : When switching the view, the enhancement layer can be decoded from the nearest intra refresh. In order to reduce the view switching (random access) delay of the enhancement layer, it is preferable to use switching prediction (SP) or switching intra (SI).

reference Picture list management : The decoded image of the base layer of another view can be used as the reference image of the enhancement layer. To this end, a syntax for the reference picture structure is required.

Syntax for HMSVC

1. NAL header

In the case where only one view on the receiving side uses the SVC enhancement layer, it is possible to recognize that the display_view is automatically set to the SVC enhancement layer of the display_view even if the MVC header is not transmitted. However, when transmitting SVC NALU to multiple views, it is necessary to recognize the view and the hierarchy respectively.

■ Concatenated MVC / SVC NAL header

MVC and SVC 's NAL headers are attached to each other. SVC NALU is optional, so it is better to have MVC by default. In this case, whether the MVC base layer NALU (NAL Unit) or the SVC NALU should be distinguished is made in the MVC NAL header. For this purpose, the last bit of the MVC NAL header can be used in one embodiment. Let's call it SVCE (SVC Extension). If SVCE = 1, it is an HMSVC NAL header. If SVCE = 0, it is an SVC NAL header.

The H.264 / AVC NAL header in the first byte is preferably the same as the first byte in the SVC NAL header to be attached. In this case, the temporal_level (3) in the second byte is that of the base layer of the current SVC, and the temporal_level (3) in the fifth byte is that of the current SVC NALU.

MVC NAL  Limited by header SVC NAL  Header

If the number of SVC layers is very limited, information about the SVC can be expressed using View_id (10) or resv (6). In this case, the degree of freedom of the MVC NAL header is reduced, but the size of the NAL header can be reduced. In case of HMSVC, if one kind of SVC is allowed, if only 1 bit and 4 types of SVC are allowed, 2 bits can be used to display SVC. In one embodiment, view_id (10) can be used as view_id (8) + SVC_id (2). An example of using this is MVC with QCIF 15Hz and SVC with CIF 15Hz, 4CIF 15Hz, 4CIF 30Hz. At this time, the number of views can be up to 256. In one embodiment, this can be done as follows. Table 1 shows a layered information table for the restricted SVC NAL header.

QCIF 15Hz (base) CIF 15Hz CIF 30Hz 4CIF 30Hz SVC_id 00 01 10 11

In case of using SVC_id (1), the enhancement layer can be set in various ways. In one embodiment, SVC_id = 0 can be set to QCIF 15 Hz, and SVC_id = 1 can be set to 4CIF 30 Hz. At this time, the number of views can be up to 512. If View_id (10) is not available, SVC_id can be set in resv (6). That is, one bit or two bits can be used in resv (6) to recognize the SVC layer.

The allowable SVC can be set using SPS (Sequence Parameter Set) or PPS (Picture Parameter Set) message. If the SVC_id can be selectively made to be one bit or two bits, the choice of this should also be conveyed through the SPS or PPS message.

● Other methods

■ Using slice header: Overhead is too large to be realistic

■ Defining UDP ports by view

When the HMSVC streaming service of the server client structure is implemented, the MVC bitstream corresponding to the base layer is transmitted on one UDP port, and in the case of the SVC bitstream corresponding to the higher layer, the UDP port is allocated for each view and transmitted. In this case, the bitstreams of the base layer can be decoded using the existing MVC decoding method, and the view can be distinguished from the ports transmitted only to the higher layer. This does not require the syntax change in MVC, SVC NAL header.

■ How to use labels

All NAL headers shall be 2 bytes, the first byte shall be the same as that of H.264 / AVC, the second byte shall be the label and shall be made of a list of all kinds of NAL headers to be distinguished for one sequence, It is a method to give a label. The label and the corresponding NAL header are passed through the SPS message. There should also be a separate NAL header for SI / SP. If only the NAL header of H.264 / AVC is used, labeling is unnecessary for recognizable SPS or PPS.

The labeling method can define 256 different NAL headers by using a one-byte labeling method, and can recognize an infinite number of NAL headers by using the first 7 bits as a label and the last bit as an extension. . As shown in the following table, if there are three SVC layers including a base layer for one view, and three temporal levels, respectively, nine labels are required for one view. If there are 8 views, a total of 72 labels are required. If you use SP or SI, you need more than 9 labels for each view to distinguish them from each other. Table 2 shows a layered information table for the HMSVC NAL header using labels.

Label Temporal_ID (3) Anchor (1) View_id (8) SVC (2) 0 0 One 0 00 One One 0 0 00 2 2 0 0 00 3 0 One 0 01 4 One 0 0 01 5 2 0 0 01 6 0 One 0 10 7 One 0 0 10 8 2 0 0 10 9 0 One One 00 10 One 0 One 00 11 2 0 One 00 12 0 One One 01 13 One 0 One 01

Anchor (1) is not necessary if the anchor is redundant with temporal_id. Also, if view_level (3) is known from view_id by analogy, three bits are not needed. Also, if the contents of temporal_id (3) and SVC type are overlapped, the number of labels required for view is further reduced. However, a little computation is required for each NALU in order to specify redundant information in order to judge whether filtering is performed in the server, the MANE, and the decoder. In MANE, you only need to list the labels that need to be filtered by each user or access network. Specifically, it is not necessary to tell which label belongs to which layer or view.

2. Sequence Parameter Set (SPS)

The SPS contains information that both the encoder and the decoder need to have in common. In the present invention, the SPS of the existing MVC and SVC is required for both the encoder and the decoder. In addition, when the apparatus of the present invention is implemented using the restricted SVC NAL header using the MVC NAL header and the HMSVC NAL header using the label, the SPS needs to be modified.

■ Concatenated MVC / SVC NAL header

In case of using concatenated MVC / SVC structure, since the HMSVC NAL header contains layering information of the corresponding slice, both MVC and SVC SPS are needed to support MVC and SVC at the same time. .

■ MVC Limited SVC using NAL header NAL header

For a restricted SVC NAL header using the MVC NAL header, the SPS shall specify whether to represent the layering using the upper (or lower) bits of the 10 bits corresponding to the view_id of the MVC NAL header. In addition, the contents of the layered information taken from the view_id must be unified between the encoding and decoding units. Therefore, the SPS syntax needs to be changed.

3. PPS (Picture Parameter Set): No change

Services to which the invention applies : The present invention is useful in multicast services. The likelihood of delivery of the feedback request is irrelevant. Here, the feedback request tells the user what view the user wants to view. If there are only a small number of users, resources can be saved by checking the view that the user is viewing and not transmitting when no one is watching. In a case where the available bandwidth can be known for each user (for example, IEEE 802.16e), it is preferable that the user can selectively determine whether or not to transmit the enhancement layer according to the user. This determination is preferably made at the MANE (Media Aware Network Element) near the user group or at the server.

The quality of service may also be determined by the user's terminal or by the user's choice. If a fee is charged for each packet received, it is very important whether the user can select quality. When the number of users is so large that individual control is difficult, the present invention can be used for saving the terminal resources rather than reducing the transmission bandwidth.

In a service with four views, assume that views 1, 2, and 3 refer to view 0. The terminals 1 and 2 select and decode necessary streams in a network where all data of all views are broadcasted and display them. The terminal (1) selected for view 1 decodes only the base layer of view 0 and view 1 because the display screen size is small and the CPU performance is not sufficient. View 0 is required as a reference view of view 1. In the terminal 2 that has selected the view 2, the SVC enhancement layer is received and decoded because the display screen size is large and the CPU performance is sufficient. In view of the fact that the terminal 3 has selected the view 1 and the 1: 1 VOD service, the MANE filters and transmits unnecessary streams based on the feedback information from the terminal 3.

The syntax of the NAL header and SPS is required. Here, the NAL header specifies both concatenated and restricted SVCs.

■ NAL header structure of HMSVC packet

1. Concatenated NAL header (Concatenated MVC / SVC NAL header)

Tables 3a and 3b represent combined NAL headers, and Table 3b is the portion following Table 3a.

3. SPS, MVC NAL header structure for SVC restricted (restricted SVC) NAL header

The upper and lower bits of the view_id (u (10)) of the MVC NAL header are used to indicate restricted layering information. At this time, the number of bits of view_id to be used is set using the SPS.

Table 4 shows the SPS message processing method for the SVC-limited NAL header.

Tables 5a, 5b, and 5c show that the seq_parameter_set_mvc_extenstion () is performed according to the profile_idc value (XX) in the SPS processing routine.

Table 6a and Table 6b show the NAL MVC SPS processing routine.

Table 7 shows the nal_unit_header_svc_mvc_extension processing routine.

3. HMSVC NAL unit processing using Label

A. Read SPS and set decoders

i.NAL header to SPS processing routine (see Table 8)

ii. (XXX) seq_parameter_set_mvc_label_extenstion () according to the value of profile_idc in the SPS processing routine (see Tables 9a, 9b, and 9c)

iii. Set label via seq_parameter_set_mvc_label_extension () (see Table 10)

B. Hybrid MVC SVC NAL unit type code definition (see Tables 11a and 11b)

■ Semantics

● Concatenated MVC SVC NAL header

■ nal_unit_header_svc_mvc_extension ()

Hybrid_mvc_svc_extension_flag When the corresponding bit is set to 1 (or 0), the flag bits

● Restricted SVC NAL header

■ seq_parameter_set_mvc_extension ()

restricted_bits When the HMSVC NAL header is input, it indicates whether the upper (or lower) bits of the 10 bits corresponding to the view_id are used to determine the layering information.

dependency_id Indicates the degree of spatial layering that the HMSVC code / decoder can support. Using the corresponding value, the code / decoder creates the layered information table and determines the layered information when the HMSVC NAL header is input.

temporal_level Indicates the degree of temporal layering that an HMSVC code / decoder can support. Using the corresponding value, the code / decoder creates the layered information table and determines the layered information when the HMSVC NAL header is input.

quality_level indicates the degree of image quality layering that the HMSVC code / decoder can support. Using the corresponding value, the code / decoder creates the layered information table and determines the layered information when the HMSVC NAL header is input.

■ nal_unit_header_svc_mvc_extension ()

(svc_id) representing the layering information is extracted from the upper (or lower) bit of the view_id using the restricted_bits of view_id seq_parameter_set_mvc_extension (). The extracted svc_id expresses the corresponding layering information using the layering information table.

● HMSVC NAL header using Label

■ seq_parameter_set_mvc_label_extension ()

seq_parameter_set_mvc_label_extension () defines the kind of layering that can be supported by the code / decoder. Using this, the code / decoder creates and maintains a layered table (label).

num_views_minus_1 expresses the total number of views

temporal _ level HMSVC marks / indicates the extent to which the temporal layered decoder can support. Using the corresponding value, the code / decoder creates the layered information table and determines the layered information when the HMSVC NAL header is input.

dependency_id Indicates the degree of spatial layering that the HMSVC code / decoder can support. Using the corresponding value, the code / decoder creates the layered information table and determines the layered information when the HMSVC NAL header is input.

quality_level indicates the degree of image quality layering that the HMSVC code / decoder can support. Using the corresponding value, the code / decoder creates the layered information table and determines the layered information when the HMSVC NAL header is input.

■ nal_unit_header_HMSVC_label_extension ()

The routine to process when the HMSVC_label NAL header is entered according to the new nal_type definition (coded slice of an MVC-SVC picture).

by label number seq_parameter_set_mvc_label_extension _ () indicates the label number of the layered label Table set to a code / decoder.

the next _ label _ bit _ flag label number is not possible case, the value is 1 (or 0) as is, nal_unit_header_HMSVC_label_extension () 1 byte exists since if setting a flag bit which is used for extending expressed as a 7-bit .

4. Resource efficient multi- point streaming device using HMSVC

The following figure (FIG. 10) shows a block diagram of a resource efficient multi-point streaming device using HMSVC encoding, decoders, and MANE. For the multi-view video, one or more cameras are input, the input for the base layer is downsampled, and the input for the advanced layer is input to the encoder as it is. At this time, the MVC encoder corresponding to the base layer uses the prediction structure of the existing MVC encoder as it is. Each of the base layers is referenced to encode a high-quality advanced layer. In this case, the concatenated MVC / SVC NAL header or the Restricted SVC NAL header using the SPS and MVC NAL header is used for the bit stream of the advanced layer generated in this case. The bitstreams generated by the respective encoders are combined into one or a plurality of bitstreams using a multiplexer and transmitted to the MANE. Among them, the base layer transmits all of the bits, and the higher layer is transmitted according to a view change command , And selectively transmits a view that the user wants to view. The bit stream arriving at the decoder is divided into a base layer bitstream and an advanced layer bitstream through a demultiplexer and input to a corresponding decoder and decoded. At this time, the bit stream of the advanced layer corresponds to the view to be viewed at the current high image quality, and decodes it by referring to the same view (S2 in the figure) of the base layer. If there is a view change according to the user's request during viewing, the view change command is transmitted to MANE, and the filter transmits the advanced layer bit stream of the corresponding view. The switch located in the decoder plays the role of converting the reference direction of the advanced layer when there is a view switching. 11 shows an example of a streaming system using HMSVC.

The application range of MVC can be greatly expanded. At present, all views must be received and decoded, so bandwidth, CPU performance, and memory resources are required. However, it is now possible to selectively adjust the stream to be transmitted according to the view to be displayed and the required quality. The scope of application is expanded under the following conditions.

1. User can select the view and quality to display.

2. The range of bandwidth that can be transmitted is further expanded. (I.e., it can be serviced even when the available bandwidth is small)

3. The range of available terminals is expanded. (Applicable to lower CPU performance and smaller display sizes)

Claims (25)

The method comprising: receiving a bitstream including information about a plurality of views; And And decoding the bitstream based on information about the plurality of views, Wherein the information on the plurality of views includes at least one of identification information on the plurality of views and layering information on the plurality of views, Wherein the bitstream includes an extension flag for indicating layering information for the plurality of views and further includes layering information for the plurality of views according to a value of the extension flag. Way. The method according to claim 1, In the step of decoding the bitstream, Decodes an enhancement layer image for the display view based on information on a display view displayed among the plurality of views, The information on the display view may include, The identification information for identifying the display view, and the layering information for the display view. The method according to claim 1, The information about the plurality of views may include: (NAL) unit header. ≪ Desc / Clms Page number 13 > delete The method according to claim 1, Wherein the information for the plurality of views is derived from predetermined bits contained in a NAL unit header in the bitstream, Wherein the upper or lower bits of the predetermined bits indicate identification information for identifying the plurality of views and the remaining bits of the predetermined bits other than the upper or lower bits are layered information for the plurality of views The video decoding method comprising the steps of: 6. The method of claim 5, Wherein information on the number of bits of the upper or lower bits of the predetermined bits is set using a SPS (Sequence Parameter Set) or a PPS (Picture Parameter Set). The method according to claim 1, Wherein the bitstream includes label information indicating information on the plurality of views, Wherein the label information is a table predetermined by an encoder and a decoder and is information composed of identification information for the plurality of views and label for specifying layering information for the plurality of views. The method according to claim 1, If the received bitstream includes information indicating a view switching point for switching from a first view to a second view for a view to be decoded, Wherein the enhancement layer decodes the enhancement layer image for the second view after an intra refresh closest to the point indicated by the view switching point. 9. The method of claim 8, If you allow view switching latency, The image of the enhancement layer for the first view is decoded before the intra refresh, And decodes an enhancement layer image for the second view after the intra refresh. 9. The method of claim 8, If you do not allow view switching latency, The image of the base layer for the second view is up-sampled before the intra refresh, And decodes an enhancement layer image for the second view after the intra refresh. 3. The method of claim 2, The enhancement layer for the display view comprises: Wherein at least one of the decoded images of the base layer for the plurality of views is used as a reference image. 3. The method of claim 2, Receiving a bitstream including an image of a base layer for the plurality of views at a first UDP (User Datagram Protocol) port, And receiving a bitstream including an enhancement layer image for the display view by a second UDP port, wherein a different UDP port is used for each of the plurality of display views if the display views are a plurality of display views. Way. Identifying information about a plurality of views; And And encoding information about the plurality of views, Wherein the information on the plurality of views includes at least one of identification information on the plurality of views and layering information on the plurality of views, Further comprising the step of encoding extension flag information for indicating layering information for the plurality of views, wherein in the encoding of the extension flag information, And coding the layered information. 14. The method of claim 13, In the step of encoding information on the plurality of views, Determining a display view to be displayed among the plurality of views; And And encoding information about the display view, The information on the display view may include, The identification information for identifying the display view, and the layering information for the display view. 14. The method of claim 13, The information about the plurality of views may include: (NAL) unit header of the video signal. delete 14. The method of claim 13, In the step of specifying information on the plurality of views, Identifying information about the plurality of views using predetermined bits in a NAL unit header, Wherein the upper or lower bits of the predetermined bits specify identification information for identifying the plurality of views and the remaining bits of the predetermined bits other than the upper or lower bits are used for layering And the information is specified. 18. The method of claim 17, Wherein information on the number of bits of the upper or lower bits of the predetermined bits is set using a sequence parameter set (SPS) or a picture parameter set (PPS). 14. The method of claim 13, In the step of specifying information on the plurality of views, Information on the plurality of views is specified based on label information, Wherein the label information is a table predetermined by an encoder and a decoder and is information composed of identification information for the plurality of views and a label for specifying layering information for the plurality of views. 14. The method of claim 13, Further comprising the step of encoding a bitstream comprising information indicating a view switching point switching from a first view to a second view, Wherein the information indicating the view switching point comprises information about when an image of the enhancement layer for the second view should be decoded to enable switching from the first view to the second view, / RTI > 21. The method of claim 20, Wherein the information indicating the view switching point comprises: And instructs to decode an enhancement layer image for the second view after an intra refresh closest to a point indicated by the view switching point. 22. The method of claim 21, If the view switching delay is allowed, the information indicating the view switching point, Wherein the decoding unit decodes the enhancement layer image for the first view before the intra refresh and decodes the enhancement layer image for the second view after the intra refresh. 22. The method of claim 21, If the view switching delay is not allowed, the information indicating the view switching point, Wherein the controller instructs to decode and upsample an image of a base layer for the second view before the intra refresh and to decode an image of an enhancement layer for the second view after the intra refresh. 15. The method of claim 14, Further comprising: encoding an enhancement layer image for the display view based on information about the display view, The enhancement layer for the display view comprises: Wherein at least one of the encoded images of the base layer for the plurality of views is used as a reference image. 25. The method of claim 24, The base layer image for the plurality of views is transmitted to a first UDP (User Datagram Protocol) port, And transmitting an enhancement layer image for the display view to a second UDP port, wherein if the plurality of display views are a plurality of display views, the plurality of display views are transmitted using different UDP ports.

Images