KR100586099B1 - Moving picture coding method - Google Patents

Abstract

In accordance with another aspect of the present invention, there is provided a video coding method for performing error concealment in a video transmission system, comprising: extracting synchronization information of a predetermined macroblock in a GOB unit with respect to an image input from an encoder; Transmitting through data hiding during encoding on the synchronization information of a predetermined macroblock constituting the extracted GOB unit; Extracting synchronization information of a predetermined macroblock of a GOB unit transmitted by data hiding at a decoder, and performing error concealment using the extracted synchronization information; It includes.

According to the present invention, the synchronization information of a predetermined macroblock forming a GOB unit indicates a start position of a sixth macroblock on a bit stream in a GOB unit.

In addition, according to the present invention, in performing data hiding, data hiding is performed by using a quantization parameter for the input image and / or a level value (a value obtained by dividing the DCT coefficient by the quantization parameter) in a block in which DCT is performed.

Further, according to the present invention, in performing data concealment on the synchronization information of the predetermined macroblock constituting the extracted GOB unit, the synchronization information of the predetermined macroblock extracted one GOB after the GOB from which the synchronization information of the predetermined macroblock is extracted Hides.

Description

Moving picture coding method

1 is a view schematically showing a process in which coding is performed by a general video coding method.

2 is a diagram for explaining an image layer in a general video coding method.

3 is a view for explaining the effect of an error occurs in each video layer in a general video coding method.

4 is a view for explaining the effect when the synchronization information of each GOB is provided by the video coding method according to the present invention.

5 is a view for explaining a process of data concealment is performed using a quantization parameter by the video coding method according to the present invention.

6 is a view for explaining a process of data hiding is performed by using a level value by the video coding method according to the present invention.

7 is a flowchart illustrating a process in which coding is performed in an encoder by a video coding method according to the present invention.

8 is a flowchart illustrating a process in which coding is performed in a decoder by a video coding method according to the present invention.

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to video coding. In particular, when a multimedia video is transmitted through a network, the video coding method can efficiently process error concealment with respect to errors generated in a decoding process by hiding data in an encoding process. It is about.

Recently, as high-speed mobile communication such as IMT2000 becomes visible, multimedia services such as video, as well as voice and still images, are expanding. Examples of multimedia services of video include one-way services such as VOD and two-way services such as video calls. Video communication basically performs video compression because it has much more data than voice calls.

In the general video coding process, as shown in FIG. 1, a motion vector is extracted through motion compensation and motion estimation on an input image, and a DCT and quantization are performed on the difference image. Perform the process. VLC encoding is performed to more efficiently process data compression on the quantized data.

Then, the inverse quantization and the inverse DCT (IDCT) are combined with the previous image to restore the existing image, and the difference image of the next image is obtained. In this way video data compression is achieved.

Then, after the video compression, a call is made with the other party through the network. At this time, noise is generated in the network to which data moves. That is, the data to be communicated is not transmitted to the receiving end as it is, and the data is damaged or lost. This loss or corruption of transmitted data is called an error.

This error affects not only the frame in which the error occurred, but also continuously thereafter. The reason for this is that compression is performed on the video, where spatiotemporal motion prediction causes errors in the data that occur once to affect successive frames. In addition, the variable coding scheme loses sync when a data error occurs and affects until the next sync is found in the frame. In other words, the video compression causes the picture quality to be lost to data very large and continuous.

In particular, when an error occurs in data compression, all information becomes inaccurate until the next synchronization information is found. Therefore, the error propagation can be greatly reduced as synchronization information frequently appears. However, because the synchronization information must be different from other encoded bitstreams, the information is long and cannot be frequently entered into video compression. For example, in H.263, synchronization information is included only before the picture header and the GOB header. Since the synchronization stream is not frequently included in the compressed stream, when an error occurs, all the data is damaged until the next synchronization information is met, and the image quality deteriorates.

The image quality deterioration is a major obstacle to video services. In the video compression standard, there are methods for restoring and concealing the above errors, but each has its limitations.

The recovery method for an error is mainly performed by inserting additional data in an encoder, and is a method for preventing an error propagation in a decoder when an error occurs. This is a method to prevent the propagation of an error by inserting additional Sync bits in the stream to be encoded or cutting the encoded stream into bundles of the same length. This error recovery method has a disadvantage in that the decoder cannot perform decoding unless the decoder knows the error recovery method of the encoder.

Another method of error recovery includes a method in which an encoder and a decoder perform simultaneously. This method uses a back channel, and when the decoder informs the encoder where the error occurs, the encoder encodes the information in such a way that the error can be recovered. In this case, there is a disadvantage in that the system needs to additionally support the back channel.

In addition, the error concealment method is a method performed by the decoder, and is a method of making the portion where an error occurs as similar to the original image as possible by using the image of the decoder. In this case, since the error is concealed using only limited information in the decoder, the accuracy of the error concealment is limited.

Only the method of using the error recovery method and the error concealment method described above has a disadvantage in that it is not possible to eliminate all the errors.

According to the present invention, when a multimedia image is transmitted through a network, by providing data synchronization information of a predetermined macroblock forming a GOB unit through data hiding in an encoding process, error concealment is efficiently processed for errors generated during a decoding process. The purpose is to provide a video coding method that can be done.

In order to achieve the above object, a video coding method according to the present invention provides a video coding method for performing error concealment in a video transmission system.

Extracting synchronization information of a predetermined macroblock in a GOB unit with respect to an image input from an encoder;

Transmitting through data hiding during encoding on the synchronization information of a predetermined macroblock constituting the extracted GOB unit; And

Extracting, by the decoder, synchronization information of a predetermined macroblock forming a GOB unit transmitted by the data hiding, and performing error concealment using the extracted synchronization information; Its features are to include.

According to the present invention, the synchronization information of a predetermined macroblock constituting the GOB unit is characterized in that it indicates the start position of the sixth macroblock on the bit stream in the GOB unit.

Further, according to the present invention, in performing data hiding on the synchronization information of the predetermined macroblock of the extracted GOB unit, the feature is that the synchronization information of the predetermined macroblock of the GOB unit is repeatedly hidden several times. There is this.

According to the present invention, in performing the data concealment, data concealment is performed using a quantization parameter and / or a level value (a value obtained by dividing the DCT coefficient by the quantization parameter) in the block in which the DCT is performed. There is a characteristic in that.

Further, according to the present invention, in performing data concealment on the synchronization information of the predetermined macroblock constituting the extracted GOB unit, the extracted predetermined macroblock is one GOB after the GOB from which the synchronization information of the predetermined macroblock is extracted. Its characteristics are that it hides the motivational information of the.

According to the present invention, in performing error concealment by using the synchronization information of a predetermined macroblock extracted by the decoder, when an error occurs in a macroblock of an input GOB, a macro that forms a GOB extracted from hidden data is formed. The feature is that the decoding of the macroblock of the GOB in which the error occurs is performed again by referring to the synchronization information of the block.

According to the present invention, when the synchronization information of a predetermined macroblock extracted by the decoder is the synchronization information of a sixth macroblock forming one GOB, the macroblock in which the error occurs is earlier than the sixth macroblock. In this case, the decoding is performed again from the sixth macroblock of the GOB in which the error is generated by referring to the extracted synchronization information.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The present invention provides a video coding method that can more effectively process concealment of an error when an error occurs in a decoder by hiding and transmitting synchronization information of a predetermined macroblock forming a GOB unit through data hiding in an encoding process. .

As is known, in video image communication, video is compressed and transmitted because of the large amount of data. When all the video data is compressed, the amount of compressed data is also large. Therefore, a motion vector is found to find a part similar to the previous image, only the difference value is encoded, and the variable length encoding is performed.

When video data is compressed in this manner, data compression can be performed a lot, but when an error occurs, error propagation occurs. That is, even if an error occurs in one bit because it is compressed through variable-length encoding, the error continues until the next sync code is met, and since the encoding is referred to the previous image, the error occurs in the previous image until the next image. The error will be propagated.

In addition, the general video compression method that compresses through motion prediction and motion compensation divides an image into layers and performs compression using layers, because using layers can include common information of lower layers in a header. . The layer of H.263, which is a general motion prediction and motion compensation video compression method, is shown in FIG. 2 is a diagram for describing an image layer in a general video coding method.

As shown in Fig. 2, the image layer is divided into four layers: a picture layer, a GOB layer, a macro block (MB) layer, and a block layer. Here, the picture means one frame and includes nine GOBs. And, GOB means one row in macroblock units and includes 11 macroblocks. The macro block includes four luminance blocks and two color difference blocks in units of finding a motion vector. The block contains 8 * 8 pixels as the most basic unit of data compression.

More specifically, headers are attached to each layer to represent structural information of the lower layer. The picture header contains information on the entire skeleton of one frame, such as a picture encoding type, whether to use an additional mode, a reference frame number, and a size of an image. The GOB header contains information such as the GOB start code, GOB number, and frame ID. That is, information on one column of macro block units is included. The macro block header includes information such as a motion vector and a macro block encoding type in units of 16x16 pixels.

As explained above, each layer has a header that contains the structural information of its lower layer. When an error occurs in the header of each layer in a video communication environment, the image quality deterioration is much greater than when an error occurs in data. FIG. 3 shows what quality degradation occurs when an error occurs in the header of each layer, and how much the effect thereof is.

As shown in Fig. 3, when an error occurs in the picture header which is the uppermost layer, the entire frame cannot be decoded and a large image quality deterioration occurs in the next frame referencing the frame. In particular, when the movement is large, the image quality deteriorates even more when the previous image cannot be referred to.

In addition, when an error occurs in the GOB header, one line of the GOB cannot be decoded normally. When an error occurs in the macroblock header, normal decoding is not performed from the macroblock in which the error occurs to the last macroblock.

In the decoding process, since the picture header and the GOB header include a sync code, when the encoded data encounters an error and loses synchronization and cannot be decoded normally, the picture header and the GOB header are synchronized. Can be decoded normally. However, since the macroblock header does not contain the synchronization code, if an error occurs in the header or data in units below the macroblock, variable-length encoding cannot correctly decode the content of the data.

Therefore, all data is discarded until a picture header or GOB header with the next synchronization code appears. In other words, even if the generated error is simply one bit of error, as shown in Fig. 4, all data until the next synchronization code appears cannot be restored.

In the present invention, in order to prevent such deterioration of image quality, by using a data concealment technique, the synchronization information of a predetermined macroblock constituting a GOB unit is hidden in the encoded data and transmitted to perform image reconstruction for errors generated during decoding.

Then, the data hiding performed in the present invention will be described.

Data hiding is a technology that hides the desired information in digital multimedia and extracts the hidden information when needed. The technique of hiding data can be divided into those that require originals and those that do not require originals to extract largely hidden information. In the present invention, it is intended to use a method that does not require the original data as a data hiding method to be used.

Data hiding in the present invention is performed during video compression. Data hiding may be performed by modifying a parameter used for data compression or a value dependent on the input image when compressing the input image. In order to be able to hide data when performing video compression, the data should be hidden so that the original parameter or the original data is not changed, but it does not affect the quality or the amount of data to be compressed. Such values include the quantization parameter (QP) and the 'level' value obtained by dividing the DCT coefficient by the quantization parameter.

Performing data hiding using the quantization parameter is shown in FIG. 5. 5 is a diagram for describing a process of data hiding using a quantization parameter by a video coding method according to the present invention.

As is known, controlling the amount of data to be encoded is a quantization parameter. The quantization parameter is a parameter that divides the coefficient value obtained by DCT of the difference between the input image and the input image. The larger the quantization parameter is, the larger the value of dividing the DCT coefficients is. Therefore, the amount of data to be encoded is reduced.

And when sending and receiving video and video over the network, the data is compressed according to the network bandwidth. The wider the bandwidth of the network, the greater the amount of data that can be communicated, resulting in higher amounts of encoding and better image quality. Conversely, the narrower the bandwidth of the network, the less data can be communicated, which results in less encoded content and poorer quality. As described above, the amount of data to be encoded is controlled by using the quantization parameter, and the data is hidden.

As shown in FIG. 5, after performing DCT on the input image or the differential image of the input image, an appropriate quantization parameter is set in consideration of the bandwidth of the network. The quantization parameter enters the macroblock header and is used to decode the compressed image. In this case, data hiding is performed before quantization using the actual quantization parameter. Here, data hiding can be performed as follows.

QP_new% 2 == Hide bit [k]

QP_new no change

QP_new% 2! = Hide bit [k]

QP_new = QP_new + 1;

Hide bit [k]: Bit stream of data to hide

When the data is hidden as described above, the decoder can decode the compressed data and look at the value of the quantization parameter to extract the hidden data. If the quantization parameter is even in the decoder, the hidden data becomes '0', and if the quantization parameter is odd, the hidden data becomes '1'.

At this time, the encoder slightly changes the quantization parameter to hide data. That is, if the data to be hidden and the quantization parameter are divided by 2, and the remainder is not the same, the quantization parameter value is increased by 1 and divided by 2 makes the remainder the same. This increases the value of dividing the DCT coefficients, thus reducing the amount of data encoded. However, the image quality felt by the human eye is hardly generated.

Meanwhile, FIG. 6 illustrates that data hiding is performed on a 'level' value, which is a quantized value of a DCT coefficient value of an input image or a differential image. FIG. 6 is a diagram for describing a process of data hiding using a level value by a video coding method according to the present invention.

According to the video coding method according to the present invention, DCT is performed on an input video or a differential video for compression. Then we look at the bandwidth to be transmitted and specify the quantization parameter to adjust the amount of bits to be encoded. The value divided by the quantization parameter is the DCT coefficient.

In FIG. 6, the DCT is performed in units of 8 * 8 blocks. A quantization parameter is applied to a block on which DCT is performed to divide coefficient values. The quotient divided by the quantization parameter is called 'level'. Data hiding is performed using 'level' which is the quotient of each block. Here, the data can be hidden as follows.

LevelSum% 2 == Hide Bit [k]

No level change

LevelSum% 2! = Hide Bit [k]

Decrease the lowest value of level among the levels by 1

LevelSum: The sum of all block levels

The reason for hiding data at the sum of 'level' in units of blocks without hiding data at each 'level' is that data hiding for each 'level' affects the image quality. Since 'Level' is the quotient of the DCT coefficient divided by the quantization parameter, the larger the quantization parameter is, the larger the range of DCT coefficients that 'level' means.

For example, even if the 'level' value is equal to '1', the DCT range is 10 to 19 when the quantization parameter is 10, and the DCT range is 30 to 59 when the quantization parameter is 30. Therefore, after calculating 'sum' of 'level' in 8 * 8 block units, if 'sum' of 'level' is divided by 2 and the data bits to be hidden are the same, there is no change in 'level' and the least important if not Decrease the value of 'level' by 1.

Here, the 'level' of the least importance refers to the part having the least sensitivity to the human eye. The human eye is less sensitive to the high frequency range, so if the data bits to hide and the 'level sum' divided by 2 are not equal, reduce the value of 'level' with the highest frequency.

That is, if the 'level sum' is '16' and the bit to be inserted is '1', the block 'level sum' becomes '15' by reducing the value of 'level', which is the highest frequency region of the block, to '1'. Divide the beat and the 'level sum' by two to make the rest the same.

However, if 'level sum' is '1', there is a problem. If it is 'Level sum' and the bit to hide is '1', the sum of all blocks is '0' by reducing '1' to 'level' of the least sensitive part. However, a block whose original sum of blocks is '0' does not hide data because it can no longer reduce the 'level'. In this case, the original 'level sum' is '1' and the data is hidden so that the block having 'level sum' of '0' and the level sum '0' from the beginning cannot be distinguished from the block without data hiding.

In the present invention, the following method was used to prevent such an error. If the sum of 'Level' is '1' and the data to hide is '1', the data is inserted into the block. However, the sum of 'levels' does not change. If the sum of 'level' is '1' and the data to hide is '0', change the sum of 'level' of the block to '0' and do not hide data. If the sum of 'level' is '0', the decoder assumes that there is no data hiding in the block.

In this way, except for blocks where the sum of 'levels' is '1', all blocks have the same level of data to hide and the remainder of their blocks divided by two, so the 'level' of the block remains unchanged; You can hide the data by reducing the lower level by one.

A block whose sum of 'Level' is '1' hides the data if the data to be hidden is '1'. If the hidden data is '0', the block's 'level' value is changed to '0' without hiding the data. Even when the sum of 'Level' is 0 ', data hiding is not performed.

Meanwhile, in the present invention, in performing data hiding using a quantization parameter or 'level' value, data hiding is performed for synchronization information of a predetermined macroblock forming a GOB unit.

Here, the synchronization information of a predetermined macroblock forming a GOB unit, for example, includes 11 macroblocks in one GOB, which means a start position on a bitstream of a sixth macroblock, which is a middle macroblock among them. The intermediate position information of each GOB is inserted after one GOB and transmitted.

In other words, hiding and transmitting the start position of the sixth macroblock on the bit stream means inserting synchronization information for each position of the sixth macroblock on the bit stream.

As described above, when an error occurs in the bit stream, all data cannot be decoded normally until the next synchronization code is met. However, according to the present invention, by transmitting the start position of the sixth macroblock on the bit stream, when an error occurs during decoding at the decoder, synchronization can be achieved in the sixth macroblock instead of the next GOB header. Better image quality can be obtained and the propagation of errors can be greatly reduced.

For example, as shown in FIG. 4, if an error occurs in the third macroblock, the existing method cannot normally decode all the data following until the next GOB. However, as shown in the present invention, if the start position of the sixth macroblock can be known using the synchronization information hiding, the third to fifth macroblock data cannot be decoded normally, but the sixth macroblock can be decoded normally. do.

On the other hand, in terms of an encoder that performs data hiding on synchronization information, when the data compressed in the unit of GOB is compressed by performing variable length encoding, the positions of the sixth GOB in the unit of GOB are different. Accordingly, the sixth GOB position is extracted from the bits encoded in the GOB unit, and the extracted position information is inserted when the next GOB is encoded. At this time, the start position of the sixth macroblock is hidden in the quantization parameter and the level value of the DCT. This series of processes is shown in FIG. 7 is a flowchart conceptually illustrating a process in which coding is performed in an encoder by a video coding method according to the present invention.

In addition, when the decoder side process of extracting synchronization information is described, the hidden information is extracted in units of GOB when decoding is performed to find the start position of the sixth macroblock of the previous GOB. If an error occurs in the previous GOB before the sixth GOB, the extracted information is normally decoded again from the sixth macroblock of the previous GOB. This series of processes is shown in FIG. 8. 8 is a flowchart illustrating a process in which coding is performed in a decoder by a video coding method according to the present invention.

On the other hand, the error rate in the video communication varies depending on the environment. Some environments are badly 10E-3 and almost error free. At this time, since the hidden data may be broken by an error in an environment where the error occurrence rate is high, it should be sent repeatedly. If an error occurs in the restored data, the hidden data is extracted and the error is concealed. Since the data hiding itself is hidden in the encoded bit stream, the hidden data is also broken if the encoded data is broken by an error.

Therefore, when the error concealment is performed using the broken information, the image quality deterioration becomes more severe, so that the duplicated data is repeatedly transmitted several times so that the duplicated data can be used even when the hidden data in the decoder is broken by an error.

On the other hand, the conventional error concealment method is to reconstruct as much as possible to the original image using only the image received from the decoder. That is, since the information available is limited, the reconstructed image is different from the original image. In general, the video compression method compresses by using a difference from the previous video. If an error occurs at a certain point, the error is propagated to the next video unless the error is completely removed. In other words, it is necessary to eliminate the error at the time of occurrence.

Using the video coding method proposed in the present invention, the error concealment is not only performed at the receiving end, but the additional error can be eliminated as much as possible because the receiving end receives additional information. Using the proposed algorithm, when the error occurs, synchronization can be found not only in the picture header and the GOB header, but also in the middle of the GOB, thereby greatly preventing error propagation.

Using the video coding method according to the present invention, when an error occurs in a video communication service such as IMT-2000, error concealment can be more efficiently, and thus a general video codec standard (H.26X, MPEG 1,2,4) It can be effectively used for a video communication service that uses a.

As described above, according to the video coding method according to the present invention, when a multimedia image is transmitted through a network, by providing the synchronization information of a predetermined macroblock forming a GOB unit through data hiding in the encoding process, There is an advantage that the error concealment can be efficiently processed for the generated error.

Claims (7)

In a video coding method for performing error concealment in a video transmission system, Extracting synchronization information of a predetermined macroblock in a GOB unit with respect to an image input from an encoder; The predetermined value is determined using a quantization parameter for the image input during encoding and / or a level value (a value obtained by dividing the DCT coefficient by the quantization parameter) for the image input during encoding and the synchronization information of the predetermined macroblock constituting the extracted GOB unit. Transmitting the synchronization information of the macroblock through data hiding which inserts the synchronization information of the extracted predetermined macroblock after one GOB rather than the extracted GOB; And Extracting, by the decoder, synchronization information of a predetermined macroblock forming a GOB unit transmitted by the data hiding, and performing error concealment using the extracted synchronization information; Video coding method comprising a. The method of claim 1, The synchronization information of a predetermined macroblock constituting the GOB unit indicates a start position of a sixth macroblock on a bit stream in a GOB unit. The method of claim 1, And performing data hiding on the synchronization information of the predetermined macroblock of the extracted GOB unit, wherein the synchronization information of the predetermined macroblock of the GOB unit is repeatedly hidden. delete delete The method of claim 1, In performing error concealment using the synchronization information of the predetermined macroblock extracted by the decoder, when an error occurs in the macroblock of the input GOB, the synchronization information of the macroblock constituting the GOB extracted from the hidden data is referred to. And decoding the macroblock of the GOB in which the error occurs again. The method of claim 6, When the synchronization information of the predetermined macroblock extracted by the decoder is the synchronization information of the sixth macroblock constituting one GOB, the extraction is performed when the macroblock in which the error occurs is a macroblock earlier than the sixth macroblock. And decoding from the sixth macroblock of the GOB in which the error occurs with reference to the received synchronization information.

Images