KR20050082102A - Selective fine granular scalable coding apparatus and method thereof - Google Patents

Abstract

1. TECHNICAL FIELD OF THE INVENTION

An apparatus and method for FGS coding, optionally using temporal prediction information.

2. Technical problem to be solved by the invention

An algorithm and method for selectively applying network state adaptive, high coding efficiency and minimizing drift by introducing an algorithm for selectively applying temporal prediction information are provided.

3. Summary of Solution to Invention

Base layer coding means for coding an input image using a temporal correlation with an adjacent image, and enhancement layer coding means for generating enhancement information for an error image D1 between the input image and an image coded by the base layer coding means. Wherein the enhancement layer coding means generates temporal prediction information D2 for the input image and selectively uses the generated prediction information D2 for generating the enhancement information.

4. Important uses of the present invention

Can be used to send and receive video

Description

Selective Fine Granular Scalable Coding Apparatus and Method

The present invention relates to a video coding apparatus and a method thereof, and more particularly, to a microparticle scalable coding apparatus and method using selectively temporal prediction information.

The current field of information technology is rapidly integrating. Accordingly, there is a demand for a scalable video coding method capable of transmitting a video to correspond to various communication networks having different bandwidths such as broadcasting networks and wired and wireless internets, and various terminals having different computing capabilities such as digital TVs, PDAs, and notebook computers.

MPEG-4 based fine granular scalable (FGS) coding method, proposed by the Moving Picture Expert Group (MPEG) under ISO / IEC, is designed to achieve optimal image quality at a given bandwidth. It is a network-adaptive scalable coding method.

In the MPEG-4 based FGS coding method, when scalable coding is performed, when the state of the transmission line is not stable, for example, the state of the transmission line changes rapidly, such as wired or wireless Internet, the bandwidth that can be allocated to each user. This coding method is designed to be applied to the case of severe fluctuations of the video signal. The video information of the enhancement layer is coded and transmitted in units of bit planes.

That is, in the enhancement layer, the enhancement information is divided into bit planes, and the most significant bit (MSB) is transmitted first, and then important bits are continuously transmitted.

Therefore, even when the bandwidth of the transmission line is suddenly changed at the receiving end and thus all the bits necessary for image restoration are not received, only the partial image bitstream received until then can restore the image having the best image quality.

1 is a block diagram of an MPEG-4 based FGS encoder. As shown, the base layer is a complex motion estimation (ME), motion compensation (MC) and discrete cosine transform, which is a coding method of MPEG-4 part 2 visual optimized in a single layer. (discrete cosine transform: DCT) is used.

The enhancement layer calculates a residual image between the original image and the reconstructed image in the base layer, and the block unit (8 * 8 units) DCT 110 calculates the error image. The error image values are converted into DCT coefficients. If a block having a good image quality is selectively selected, the bitplane shift unit 120 may selectively perform the process of firstly transmitting the image information of the corresponding block. This is defined as selective enhancement (SE).

Thereafter, the maximum value calculator 130 calculates the maximum value of the absolute values of all the values of the DCT. The calculated maximum value is used to obtain the maximum number of bitplanes for transmitting the corresponding video frame.

The bitplane VLC (bitplane variable length coding) 140 is a row of 64 DCT coefficients (bits of the corresponding bitplane of the DCT coefficient: 0 or 1) obtained in units of blocks in a zigzag scan order. Each matrix is coded in a run-length according to a variable-length-length code table.

2 is a block diagram of an MPEG-4 based FGS decoder. As shown, a bitstream received divided into a base layer and an enhancement layer through a transmission line is decoded in the reverse order of the coding process described with reference to FIG. 1.

Referring to the drawings, the base layer uses a reverse hybrid MC / DCT scheme, which is a decoding method of MPEG-4 video.

The bitplane variable length decoding (VLD) 210 performs VLD for each bitplane with respect to the enhanced information input to the decoder, and the bitplane shift unit 220 selectively performs bitplane shift.

Thereafter, the IDCT 230 performs an inverse discrete cosine transform on a block basis to reconstruct the image transmitted from the enhancement layer, and the adder 240 combines the enhancement layer reconstructed image with the image reconstructed in the base layer.

Afterwards, the clipping unit 250 restores the finally improved image quality by clipping the combined image value to a value between 0 and 255.

Although the MPEG-4 based FGS coding method has a network state adaptive feature, coding performance is significantly lower than a single layer coding method optimized in terms of coding efficiency.

Therefore, there is an urgent need for a coding method having excellent coding performance while maintaining network state adaptive features such as MPEG-4 based FGS coding method.

The present invention was developed in response to the above-described request, and by adopting an algorithm for selectively applying temporal prediction information to an H.264-based FGS coding apparatus, it is network state adaptive, has high coding efficiency, and minimizes drift. It is an object of the present invention to provide an optional FGS coding apparatus and method thereof.

Other objects and advantages of the invention will be described below and will be appreciated by the practice of the invention. In addition, the objects and advantages of the present invention can be realized by means and combinations indicated in the claims.

An optional FGS coding apparatus of the present invention for achieving the above object comprises a base layer coding means for coding an input image using a temporal correlation with an adjacent image and coded by the input image and the base layer coding means. And enhancement layer coding means for generating enhancement information on an error image D1 between images, wherein the enhancement layer coding means generates temporal prediction information D2 for the input image, and generates the generated prediction information D2 to generate the enhancement information. It is used selectively.

In addition, the optional FGS decoding apparatus of the present invention uses a base layer decoding means for decoding base layer bitstream using a temporal correlation with an adjacent picture, and reproduction of an enhancement layer with respect to input enhancement bitstream. And enhancement layer decoding means for generating an image D1, wherein the enhancement layer decoding means generates temporal prediction information D2 for the enhancement information, and selectively uses the generated prediction information D2 for generating the reproduced image D1.

In addition, the selective FGS coding method of the present invention includes a base layer coding step of coding an input picture using a temporal correlation with an adjacent picture, and an error image D1 between the input picture and the picture coded at the base layer coding step. An enhancement layer coding step of generating enhancement information, wherein the enhancement layer coding step includes generating temporal prediction information D2 for the input image and selectively using the generated prediction information D2 for generating the enhancement information. do.

In addition, the selective FGS decoding method of the present invention uses a base layer decoding step of decoding base layer bitstream using a temporal correlation with neighboring images, and reproduction of an enhancement layer with respect to input enhancement bitstream. An enhancement layer decoding step of generating an image D1, wherein the enhancement layer decoding step includes generating temporal prediction information D2 with respect to the enhancement information and selectively using the generated prediction information D2 for generating the reproduced image D1. do.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Prior to this, terms or words used in the specification and claims should not be construed as having a conventional or dictionary meaning, and the inventors should properly explain the concept of terms in order to best explain their own invention. Based on the principle that can be defined, it should be interpreted as meaning and concept corresponding to the technical idea of the present invention.

Therefore, the embodiments described in the specification and the drawings shown in the drawings are only the most preferred embodiment of the present invention and do not represent all of the technical idea of the present invention, various modifications that can be replaced at the time of the present application It should be understood that there may be equivalents and variations.

H.264 coding method, which is attracting attention as the next generation video compression technology, is a coding method of MPEG-4 Part 10 Advanced video coding (AVC), which is a video compression standard proposed by MPEG and video coding expert group (VCEG) under ITU-T. .

The H.264 coding method was developed with a focus on improving the compression efficiency, and aims to improve the compression efficiency by 50% or more when compared to the H.263 + coding method.

3 is a block diagram of an H.264 encoder. The H.264 encoder compresses the image in units of 16 * 16 macro blocks, and each unit is coded in an intra mode or an inter mode according to correlation with an adjacent image. .

Referring to the drawings, the intra mode is a mode in which a temporal correlation with an adjacent temporal image is small or intentionally coded without motion compensation MC.

Intra mode performs intra prediction on a 16 * 16 block or 4 * 4 block basis and subtracts the prediction value from the input macroblock. Thereafter, the difference value is transmitted to the receiver by 4 * 4 integer DCT, quantization (Q), and entropy coding.

On the other hand, the inter mode extracts a motion vector of each macro block from a reference frame image stored in a basic buffer after being coded and transmitted with the current image input in order to use a temporal correlation with an adjacent image. Calculate to perform motion compensation.

Inter mode also predicts motion with up to 1/8 pixel precision using various block sizes (16 * 16, 16 * 8, 8 * 16, 8 * 8, 8 * 4, 4 * 8, 4 * 4) (ME) and motion compensation (MC). In addition, the inter mode may refer to a plurality of temporally adjacent images in order to maximize coding efficiency (multiple reference frames).

4 is an overall configuration diagram of an H.264 based FGS encoder to which a temporal prediction algorithm is applied. As shown, the above-described H.264 encoder is used for base layer coding, and a structure in which a temporal prediction algorithm is added to an MPEG-4 based FGS encoder is used for enhancement layer coding.

Referring to the drawings will be described the coding process of the enhancement layer.

<Step 1: Building Reinforcement Information of First Video>

1-1: The original image is input to the enhancement layer by a residual image between the base layer image and the original image reconstructed by the coding method described with reference to FIG. 3.

1-2: The DCT unit 402 performs a discrete cosine transform on the input error image, and the bitplane coding unit 404 performs bitplane coding to generate enhancement bitstreams. The generated enhanced information is transmitted to the receiving end.

1-3: The bitplane extraction unit 406 extracts the most significant N bitplane information in the order of the significant bits among the generated enhancement information.

1-4: The inverse bitplane coding unit 408 and the IDCT unit 410 inversely code the extracted most significant N bitplane information in the enhancement buffer 412 for subsequent information processing.

<Step 2: Reinforcement Information Processing Step of Subsequent Video>

2-1: Subsequent original image The error image between the base layer image reconstructed by the coding method described with reference to FIG. 3 and the subsequent original image is input to the enhancement layer.

2-2: The motion compensator 416 performs motion compensation on the reference image information stored in the enhancement buffer 412 using the motion vector 414 derived from the base layer, and provides temporal prediction information on the current input image. Create

2-3: The subtractor 418 calculates a difference between the error image of 2-1 and the temporal prediction information of 2-2, and the DCT unit 402 and the bitplane coding unit 404 code the subtracted information. To generate enhanced information. The generated enhanced information is transmitted to the receiving end.

2-4: The bitplane extraction unit 406 extracts the most significant N bitplane information in the order of the significant bits among the generated enhancement information.

2-5: The inverse bitplane coding unit 408 and the IDCT unit 410 inversely code the extracted most significant N bitplane information, and the adder 420 performs 2-2 current input image on the inverse coded information. The temporal prediction information for s is summed and stored in the enhancement buffer 412. The image information stored in the enhancement buffer 412 is used as reference image information of the previous image for the subsequent image, and the above-described process is repeated until the input image sequence of the encoder is finished.

The H.264-based FGS coding method to which the above-described temporal prediction algorithm is applied can significantly reduce the amount of enhancement information generated by reducing the temporal redundancy between the image information input to the enhancement layer.

However, in the coding method, if the reference picture information of the enhancement layer is not transmitted to the receiving end safely, a drift phenomenon may seriously occur.

If all of the reference picture information of the enhancement layer transmitted from the encoder is decodable by the decoder, the drift caused by the mismatch of the reference picture information can be removed.

However, the FGS coding method is designed to obtain the best image quality by using only the partial image information received up to that point even when the bandwidth of the transmission line changes suddenly and thus does not receive all the information necessary for image restoration. It is unreasonable to use a coding method to which the above-described temporal prediction algorithm is applied to a transmission environment that changes rapidly with time.

Experimentally, it is possible to improve the image quality of transmission video only when about 3 bitplane information is used as the reference image information of the enhancement layer. Therefore, in order to apply the temporal prediction algorithm, the decoder can receive all the at least 3 bitplane information. Should be provided.

FIG. 5 is a diagram comparing coding performance of an H.264 based FGS coding method (N = 3) to which a H.264 based FGS coding method and a temporal prediction algorithm are applied to a Foreman QCIF reference picture.

The horizontal axis of the figure is a received bit rate (bitrate: second), and the vertical axis of the figure is an average peak signal to noise ratio (PSNR) for the Y component of the image reconstructed at the corresponding bit rate (unit dB).

The base layer was decoded at 56 kbps, and the enhancement layer was experimented by incrementally increasing the received bit rate from 32 kbps to 1024 kbps.

As shown, in the low bit rate band on the left of the A origin, the H.264 based FGS coding method 520 to which the temporal prediction algorithm is applied is up to 4 dB lower than the H.264 based FGS coding method 510. Has a PSNR. It can be estimated that the drift phenomenon occurs because the H.264 based FGS decoder to which the temporal prediction algorithm is applied does not receive all the bitplane information.

On the other hand, in the mid / high bitrate band on the right side of the A origin, the H.264 based FGS coding method 520 to which the temporal prediction algorithm is applied is up to 1.5 dB higher than the H.264 based FGS coding method 510. Has a high PSNR. It can be interpreted that the H.264 based FGS decoder to which the temporal prediction algorithm is applied receives all three bitplane information and effectively uses the temporal prediction information.

As a result, in order to improve coding performance using a temporal prediction algorithm, the transmission environment must have a predetermined transmission bit rate or more.

6 is a schematic diagram of an optional FGS encoder in accordance with the present invention. The FGS encoder of the present invention generates temporal prediction information from the image information in the enhancement layer and performs coding by selectively using the prediction information as necessary.

Referring to the drawings, reference numeral D1 denotes an error image in macroblock units between an original image and an image coded in a base layer.

Reference numeral D2 denotes temporal prediction information in units of macroblocks in which motion compensation is performed on the N bitplane information among the coding results of the previous image in the enhancement layer by using the motion vector of the base layer.

The present invention selectively uses the temporal prediction information D2 within a range that minimizes the occurrence of drift. That is, when temporal prediction information D2 is used in the enhancement layer, D1-D2 is coded as reinforcement information. When temporal prediction information D2 is not used, D1 is coded as reinforcement information.

7 is a diagram for describing a process of determining whether to use temporal prediction information D2 according to the present invention. As shown, first, the sum of absolute difference (SAD) of D1 and D1-D2 is calculated (S710). D1 and D1-D2 in macroblock units are all differentials between images. SAD values can be calculated by taking absolute values of each difference and adding them together.

 When the calculated SAD value of D1 is called SAD_D1 and the SAD value of D1-D2 is called SAD_DD, temporal prediction information D2 is used for coding when SAD_DD is smaller than SAD_D1 (S720).

The weight w uses a value greater than 1 as a belief factor for using temporal prediction information. In this embodiment, 1.8 is used as the weight w value.

When temporal prediction information D2 is used, enhancement information (enhancement bitstream) is generated by coding D1-D2 (S740 and S750).

8A is a diagram illustrating a signal transmission process when temporal prediction information is used for coding. As shown, if SAD_DD is smaller than SAD_D1, the reinforcement information is generated by coding D1-D2.

On the other hand, when temporal prediction information is not used for coding, D1 is coded as it is to generate reinforcement information (S730 and S750).

8B is a diagram illustrating a signal transmission process when temporal prediction information is not used for coding. As shown in the figure, when SAD_DD is not smaller than SAD_D1, the reinforcement information is generated by coding D1 as it is.

As such, by using the temporal prediction information selectively as necessary, the drift phenomenon can be minimized.

Meanwhile, as illustrated in FIG. 9, the enhancement information has a selection flag bit ('S': selelction flag bit) that informs the decoder whether to use temporal prediction information in macroblock units.

Hereinafter, an optional FGS coding process according to the present invention will be described in detail with reference to FIG. 6.

<Step 1: Building Reinforcement Information of First Video>

1-1: Received initial image information D1 between the base layer image and the original image reconstructed by the coding method described with reference to FIG. 3 is input to the enhancement layer in macroblock units.

1-2: The DCT unit 602 performs discrete cosine transform on the input error image information D1, and the bitplane coding unit 604 performs bitplane coding to generate enhancement birstream.

1-3: For subsequent information processing, the bitplane extraction unit 606 extracts the most significant N bitplane information in the order of the important bits of the generated enhancement information.

1-4: The inverse bitplane coding unit 608 and the IDCT unit 610 inversely code the extracted highest N bits information and store the result in the enhancement buffer 612.

<Step 2: Processing reinforcement information of subsequent video>

2-1: Subsequent original image Error image information D1 between the base layer image reconstructed by the coding method described with reference to FIG. 3 and the subsequent original image is input to the enhancement layer in macroblock units.

2-2: The motion compensator 616 performs motion compensation on the reference image information stored in the enhancement buffer 612 using the motion vector 614 derived from the base layer, and temporal prediction information D2 for the current input image. Create in macro block units.

<Step 3: Determine whether to use temporal prediction information>

3-1: As described with reference to FIG. 7, the switching unit 620 determines whether to use the temporal prediction information D2 for coding.

3-2: When using D2 A value of '1' is assigned to the selection flag 'S' and D1-D2 is used as input information of the enhancement layer.

3-3: When it is determined not to apply D2, a value of '0' is assigned to the selection flag 'S' and D1 is used as input information of the enhancement layer.

<Step 4: Repetition of reinforcement image information processing of subsequent image>

4-1: The DCT unit 602 performs discrete cosine transform on the enhancement layer input information determined in step 3, and the bitplane coding unit 604 performs bitplane coding to generate reinforced information.

4-2: For subsequent information processing, the bitplane extraction unit 606 extracts the most significant N bitplane information in the order of the significant bits of the generated enhancement information.

4-3: The inverse bitplane coding unit 608 and the IDCT unit 610 inversely code the extracted most significant N bitplane information. If the input information determined in step 3 is D1, the result obtained by inversely coding the N bitplanes is stored in the enhancement buffer 612 as it is.

4-4: If the input information determined in step 3 is D1-D2, D2 is added to the result obtained by inversely coding the N bitplanes and then stored in the enhancement buffer 612.

The information stored in the enhancement buffer 612 is used as reference image information of the previous image for the subsequent image, and the above-described processes 2 to 4 are repeated until the input image sequence of the encoder is finished.

10 is a block diagram of an optional FGS decoder in accordance with the present invention. As shown, the bitstream received divided into the base layer and the enhancement layer through the transmission line is decoded in the reverse order to the above-described coding process.

Referring to Fig. 10, the selective FGS decoding process according to the present invention will be described in detail.

<Step 1: Restoration of reinforcement information of the first image>

1-1: Base layer bitstream of the first image is input to the base layer and decoded, and enhancement bitstream D0 is input to the enhancement layer in macroblock units.

1-2: The inverse bitplane coding unit 1010 performs inverse bitplane coding on the input enhancement information D0, and the IDCT unit 1020 performs inverse discrete cosine transform to generate a reproduced image D1 of the enhancement layer. In this process, the switch 1030 is operated based on the selection flag 'S' included in the enhanced information D0. However, since the selection flag 'S' for the first video is 0, D1 is restored to the playback video of the enhancement layer as it is.

1-3: The image D1 reproduced in the enhancement layer is displayed in combination with the image reproduced in the base layer.

1-4: For subsequent information processing, the bitplane extractor 1040 extracts N bitplane information sequentially from the upper bitplane among the input enhanced information D0.

1-5: The inverse bitplane coding unit 1050 and the IDCT unit 1060 inversely code the extracted N bitplane information and store the result in the enhancement buffer 1070. If the number of bitplanes of the enhancement information D0 is less than N, the obtained number of bitplanes is reverse coded and stored in the enhancement buffer 1070.

<Step 2: Processing reinforcement information of subsequent video>

2-1: Basic information about a subsequent image is input to the base layer and decoded, and enhancement information D0 is input to the enhancement layer in units of macro blocks.

2-2: The inverse bitplane coding unit 1010 performs inverse bitplane coding on the input enhancement information D0, and then the IDCT unit 1020 performs inverse discrete cosine transform to generate a reproduced image D1 of the enhancement layer. .

<Step 3: Determine whether to use temporal prediction information>

3-1: Whether to use temporal prediction information for decoding is based on the decoded selection flag 'S'. That is, when a value of 1 is input to the switch 1030, the motion compensator 1090 performs motion compensation on the reference image information stored in the enhancement buffer 1070 using the motion vector 1080 decoded in the base layer. . Subsequently, temporal prediction information D2 in macroblock units in which temporal prediction is performed on D1 is generated, and D1 + D2 is output as a reproduced image of the enhancement layer.

3-2: When a value of 0 is input to the switch 1030, D1 is output as it is as a reproduced video of the enhancement layer.

3-3: Displays the reproduced image in macroblock units with the image reproduced in the base layer.

<Step 4: Repetition of reinforcement information processing of subsequent video>

4-1: For subsequent information processing, the bitplane extraction unit 1040 extracts N bitplane information sequentially from the upper bitplane among the input D0.

4-2: The inverse bitplane coding unit 1050 and the IDCT unit 1060 inversely code the extracted N bitplane information, and if the reproduced video determined in step 3 is D1, the inverse coded information is intactly buffered ( 1070).

4-3: If the reproduced video determined in step 3 is D1 + D2, D2 is added to the decoded information and stored in the enhancement buffer 1070.

The image information stored in the enhancement buffer 1070 is used as reference image information of the previous image for the subsequent image, and the above-described processes 2 to 4 are repeated until the input image bitstream of the decoder is finished.

As described above, the selective FGS coding method according to the present invention can improve coding performance by selectively using temporal prediction information and can minimize occurrence of drift.

11A to 11E are diagrams comparing coding performances of an FGS coding method and another FGS coding method according to the present invention with respect to various reference pictures.

11A shows an Akiyo QCIF 30Hz video sequence, FIG. 11B shows a Foreman QCIF 30Hz video sequence, FIG. 11C shows a Container CIF 30Hz video sequence, FIG. 11D shows a Mobile CIF 30Hz video sequence, and FIG. 11E shows a Foreman CIF 30Hz video sequence. One result.

Reference numerals 1110, 1120, 1130, and 1140 in the drawings denote MPEG-4 based FGS coding method, H.264 based FGS coding method, FGS coding method using temporal prediction algorithm, and coding performance of FGS coding method according to the present invention. Indicates.

As shown, the coding performance of the optional FGS coding method 1140 in accordance with the present invention has up to 5dB improvement in performance over other FGS coding methods 1110, 1120, 1130 in PSNR.

In particular, although the FGS coding method 1130 using the temporal prediction algorithm has poor coding performance in the low bit rate band, the coding performance of the selective FGS coding method 1140 using the temporal prediction algorithm selectively according to the present invention is full band. It can be seen that it has excellent coding performance over.

 As mentioned above, although this invention was demonstrated by the limited embodiment and drawing, this invention is not limited by this, The person of ordinary skill in the art to which this invention belongs, Of course, various modifications and variations are possible within the scope of equivalent claims.

As described above, according to the present invention, by suggesting an FGS coding method that selectively uses a temporal prediction algorithm, it is possible to maintain network state adaptive features, minimize drift generation, and increase coding efficiency.

1 is a block diagram of an MPEG-4 based fine particle scalable (FGS) encoder (encoder),

2 is a block diagram of an MPEG-4 based FGS decoder;

3 is a configuration diagram of an H.264 based encoder;

4 is a configuration diagram of an H.264 based FGS encoder to which a temporal prediction algorithm is applied;

5 is a diagram comparing coding performance of an H.264-based FGS coding method (N = 3) to which a H.264-based FGS coding method and a temporal prediction algorithm are applied to a Foreman QCIF reference picture;

6 is a schematic diagram of an optional FGS encoder in accordance with the present invention;

7 is a diagram illustrating a process of determining whether to use temporal prediction information according to the present invention;

8A is a diagram illustrating a signal transmission process when temporal prediction information is used for coding;

8B is a diagram illustrating a signal transmission process when temporal prediction information is not used for coding;

9 is a diagram illustrating enhanced information with a selection flag bit inserted;

10 is a block diagram of an optional FGS decoder according to the present invention;

11A to 11E are diagrams comparing coding performances of an FGS coding method and another FGS coding method according to the present invention with respect to various reference pictures.

<Description of main reference numerals in the drawings>

602: DCT unit 604: bitplane coding unit

606: bitplane extraction unit 608: inverse bitplane coding unit

610: IDCT unit 612: enhanced buffer

614: motion vector 616: motion compensation unit

620: switching unit

1010: inverse bitplane coding unit 1020: IDCT unit

1030: switching unit 1040: bit plane extraction unit

1050: reverse bit plane coding unit 1060: IDCT unit

1090: motion compensation unit

Claims (20)

Base layer coding means for coding an input image using a temporal correlation with an adjacent image; And Enhancement layer coding means for generating enhancement information on the error image D1 between the input image and the image coded by the base layer coding means Including, The enhancement layer coding means Generate temporal prediction information D2 for the input image, and selectively use the generated prediction information D2 for generating the reinforcement information. Optional FGS Coding Device. The method of claim 1, The enhancement layer coding means A buffer that stores a reference picture of an enhancement layer already coded, A motion compensator for generating the temporal prediction information D2 by performing motion compensation on the reference image; Switching unit for selectively switching whether to use the temporal prediction information D2 Containing Optional FGS Coding Device. The method of claim 2, The switching unit If SAD (D1) is greater than SAD (D1-D2) * w, switch to D1, Switch to D1-D2 when SAD (D1) is less than SAD (D1-D2) * w Optional FGS Coding Device. Here, SAD (D1) is a value obtained by taking the absolute value of D1 and summing all of them, SAD (D1-D2) is the sum of all the absolute values of D1-D2, w is a weight greater than 1. The method of claim 3, wherein The switching unit Inserting a selection flag bit into the enhancement information according to a switching operation; Optional FGS Coding Device. The method of claim 3, wherein The weight w is about 1.8 Optional FGS Coding Device. The method according to claim 1, wherein The base layer coding means Coding the input image with an H.264 algorithm Optional FGS Coding Device. Base layer decoding means for decoding base layer bitstream using a temporal correlation with a neighboring image; And Enhancement layer decoding means for generating a reproduction image D1 of the enhancement layer with respect to the input enhancement bitstream Including, Enhancement layer decoding means The temporal prediction information D2 is generated with respect to the enhancement information, and the generated prediction information D2 is selectively used to generate the reproduced image D1. Optional FGS decoding device. The method of claim 7, wherein The enhancement layer decoding means A buffer that stores a reference picture of an enhancement layer that is already decoded, A motion compensator for generating the temporal prediction information D2 by performing motion compensation on the reference image; Switching unit for selectively switching whether to use the temporal prediction information D2 Containing Optional FGS decoding device. The method of claim 8, The switching unit Based on the selection flag bit inserted in the enhancement information. Selectively switching D1 + D2 to which temporal prediction information D2 has been added to the enhancement layer reproduction image D1 or the enhancement layer reproduction image D1. Optional FGS decoding device. The method according to claim 7 to 9, The base layer decoding means Decode the basic information with H.264 algorithm Optional FGS decoding device. A base layer coding step of coding an input image using a temporal correlation with an adjacent image; And An enhancement layer coding step of generating enhancement information on an error image D1 between the input image and the image coded in the base layer coding step. Including, The enhancement layer coding step Generating temporal prediction information D2 for the input image; Selectively using the generated prediction information D2 to generate the enhanced information Containing Optional FGS Coding Method. The method of claim 11, The enhancement layer coding step Storing a reference picture of an enhancement layer that is already coded, Generating the temporal prediction information D2 by performing motion compensation on the reference image, and Determining whether to use the temporal prediction information D2 and selectively switching Containing Optional FGS Coding Method. The method of claim 12, The switching step If SAD (D1) is greater than SAD (D1-D2) * w, switch to D1, Switch to D1-D2 when SAD (D1) is less than SAD (D1-D2) * w Optional FGS Coding Method. Here, SAD (D1) is a value obtained by taking the absolute value of D1 and summing all of them, SAD (D1-D2) is the sum of all the absolute values of D1-D2, w is the weight. The method of claim 13, The switching step Inserting a selection flag bit into the enhanced information according to the switching operation. More containing Optional FGS Coding Method. The method of claim 13, The weight w is about 1.8 Optional FGS Coding Method. The method according to claim 11, wherein The base layer coding step Coding the input image with an H.264 algorithm Optional FGS Coding Method. A base layer decoding step of decoding base layer bitstream using a temporal correlation with a neighboring image; And An enhancement layer decoding step of generating a reproduction image D1 of an enhancement layer with respect to input enhancement bitstream. Including, Enhancement layer decoding step Generating temporal prediction information D2 with respect to the enhanced information; Selectively using the generated prediction information D2 to generate the reproduced image D1. Containing Optional FGS decoding method. The method of claim 17, The enhancement layer decoding step Storing a reference picture of the enhancement layer that is already decoded, Generating the temporal prediction information D2 by performing motion compensation on the reference image, and Determining whether to use the temporal prediction information D2 and selectively switching Containing Optional FGS decoding method. The method of claim 18, The switching step Based on the selection flag bit inserted in the enhancement information. Selectively switching D1 + D2 to which temporal prediction information D2 has been added to the enhancement layer reproduction image D1 or the enhancement layer reproduction image D1. Optional FGS decoding method. 20. The method of claim 17, wherein The base layer decoding step Decode the basic information with H.264 algorithm Optional FGS decoding method.