KR100496098B1 - Method and Apparatus for Determining Modes To Reduce Transmission Error for Use in Video Encoder - Google Patents

Abstract

The present invention relates to a method and apparatus for mode determination in a video encoder for overcoming a transmission error environment.

A method of mode determination in a video encoder for overcoming a transmission error environment,

The intra cost function value in the intra mode is calculated and it is determined whether or not the pixel to be subjected to the mode belongs to the still picture, and as a result of the determination, in the inter mode when there is no motion vector when the pixel belongs to the still picture. The inter cost function is calculated, and if the pixel does not belong to the still picture, the inter cost function is calculated in the inter mode when there is a motion vector. The intra cost function value is compared with the inter cost function value, and as a result, if the intra cost function value is greater than the inter cost function value, the inter mode function is determined. If the intra cost function value is less than or equal to the inter cost function value, the intra mode function is determined. The present invention provides a method for mode determination in a video encoder for overcoming a transmission error environment.

Therefore, by using the technique of the present invention, it is possible to effectively prevent the propagation of an error when there is an error in a transmission environment such as a wireless environment in a digital video encoder using a video compression technique. This has the effect of contributing to the performance improvement in video transmission systems including video on demand (VOD), mobile broadcast systems and video telephony.

Description

Mode and Apparatus for Determining Modes To Reduce Transmission Error for Use in Video Encoder

The present invention relates to a method and apparatus for mode determination in a video encoder for overcoming a transmission error environment. More specifically, the present invention relates to a method and apparatus for determining a mode so that a video encoder can cope with an error environment efficiently in order to overcome image quality degradation in a decoder due to a transmission error in a transmission error environment.

In order to realize audio and video services, it is essential to develop and standardize compression techniques for effectively storing or transmitting large amounts of data. Advances in digital imaging technology began with still images and moved to video, eventually building a platform for multimedia services. However, in order to efficiently provide multimedia services, not only technologies for compressing data such as audio and video, but also effectively link data, facilitate interactive manipulation by users, and efficiently search for compressed data. You also need a tool to help. Standardization of related technologies has been conducted around the International Organization for Standardization / International Electrotechnical Commission (ISO / IEC), which includes H.261, H.262, H.263 and H.264, and MPEG-1, MPEG-2, International video compression standards, including MPEG-4, have all been established for the purpose of compression storage and transmission of video.

However, if there is an error in the transmission environment, the consistency of the video compression standard algorithm itself is severely affected, and if no action is taken, it will cause a fatal deterioration of image quality. In order to prevent the occurrence of such an error from affecting the overall video quality, it is also a trend of the current video standard to propose effective techniques for preventing the propagation of an error without reducing the compression performance as much as possible. These techniques make video codecs more resistant to errors, and code / designs are designed to minimize image quality degradation of decoder playback images even when errors are caused by channel uncertainty during bitstream delivery. Decoding methodologies are used. That is, the encoder has been encoding for the highest efficiency without any consideration of the state or operation of the decoder until now, but in the channel environment where transmission errors occur, the encoder can guarantee the image quality of the image reproduced by the decoder. In the channel environment in which transmission error occurs, it means that special considerations must be added to overcome the transmission error problem from the encoding process.

The error tolerance methodology considering this is largely divided into an error concealment technique in an image decoder and a bit stream generation technique having an error tolerance in an image encoder.

Here, the error concealment technique in the image decoder refers to a technique for localizing and minimizing the effect of the error on the image quality as soon as an error is detected in the received bit stream. Localizing / minimizing the effects of errors in the error concealment decoder has a large performance difference depending on the content of the bit stream. The error concealment technique has the advantage that the decoder can continue to play the video even if an error occurs in the bit stream.However, once the error concealment process has been performed, even if it finds a synchronization signal and starts a new decoding. In many cases, since the state of the decoder (the content of the frame memory, etc.) does not match the state of the virtual decoder included in the encoder, the error continues to propagate until the contents match. This is a problem that occurs due to insufficient consideration of these points in the encoding process.

On the other hand, a bitstream generation technique having an error tolerance in a video encoder means that the error concealment decoder intercepts the necessary information in the middle of the bitstream so that the influence of the error can be prevented from propagating in time and space as soon as possible. It refers to a method of constructing a bit stream by including it in an appropriate manner.

Specifically, as a technique for preventing the propagation of an error, proposed in MPEG-4 includes a resynchronization technique, a data partitioning technique, a reversible VLC technique, and the like.

The resynchronization technique refers to a technique for preventing the propagation of an error by using a point that can start decoding again when an error is mixed in the bit stream. In MPEG-4, this is achieved by inserting a special code called a Resync Marker (Resync_Marker) into the bit stream. This resynchronization marker is generated only when an error is mixed in the bit stream, but not from the combination of other codes. A group of macroblocks (MBs) leading the resynchronization marker is called a video packet, and information necessary for decoding is described in the header portion so that decoding can be performed again from the beginning of the video packet. When decoding, when the resync marker (Resync_Marker) is found in the bit stream, the interval until the next resynchronization marker is found is regarded as one packet and decoded again so that the data from the place where the error is mixed to the resync marker is It is discarded. Meanwhile, since only motion vectors predicted based on previous data exist in a motion vector data part (MVDP) consisting of motion vectors, the motion vector can be used only when there is a previously decoded motion vector. Do. Therefore, if an error is found in the motion vector data region (MVDP) in the bit stream, there is a problem in that a large amount of information is lost because the entire packet must be ignored until a resynchronization marker at which the next packet starts is found.

The data partitioning technique refers to transmitting motion data and texture data information of all macro blocks in a video packet separately. In order to easily understand this technique, FIG. 1 shows an example of a video data packet generated by a coding method for an error-tolerant mode. As shown in FIG. 1, it can be seen that the coding method separates coding into motion data and texture data. Even if such texture data information is lost due to an error, if only the motion data information is alive, it can be expected to reproduce to some extent with the motion data information. In addition, since the data amount of motion data information is overwhelmingly small compared to the texture data information, the probability of an error occurring is low. Also, if the motion data information is immediately after the packet header, it can be quickly reproduced when an error occurs in the texture data information, thereby reducing the loss due to the error. In the case of data partitioning, a special code called Motion_Marker (MM) is inserted between the motion data information and the texture data information. As with the resynchronization markers, this occurs only when an error is mixed in the motion data information or the texture data information, but never from the combination of other codes. In this way, the MM functions like a resynchronization marker. According to this technique, there is a problem in that a large amount of information is lost because data between the position where the error occurs and the next MM is discarded.

In order to compensate for this problem, RVLC (Reversible VLC) technique is to make it possible to instantaneously decode data between the next resynchronization marker or the next MM from the position where an error occurs. By using this, reverse decoding from the next resynchronization marker or the MM makes it possible to recover data to be discarded to some extent. However, this technique is also impossible to recover completely, there is still a problem that the discarded data is lost.

In order to solve the above problem, in the present invention, in view of the fact that the current image frame is encoded in association with the previously encoded frames, an error occurs at the position of each pixel in the frame based on the error probability. In this case, the effect on the image quality deterioration is quantified. The basic object of the present invention is to efficiently determine intra refresh, which takes into account the influence of noise values on image quality deterioration as described above. The ultimate object of the present invention is to propose a method for efficient mode determination using the influence of noise value.

In order to achieve the above object, the present invention provides a mode determination method in a video encoder for overcoming a transmission error environment, comprising the steps of: (a) calculating an intra cost function value in an intra mode; determining whether the pixel belongs to a still picture; (c) calculating an inter cost function value in an inter mode when there is no motion vector when the pixel belongs to the still picture as a result of the determination in step (b); (d) calculating an inter cost function value in an inter mode when there is a motion vector when the corresponding pixel does not belong to the still picture as a result of the determination in step (b); (e) comparing the intra cost function value with the inter cost function value; (f) determining the inter mode if the intra cost function value is greater than the inter cost function value as a result of the comparison in step (e); And (g) determining the intra mode if the intra cost function value is less than or equal to the inter cost function value as a result of the comparison in step (e). It is characterized by providing a method of mode determination.

In addition, according to another object of the present invention, in the video compression apparatus having a processor for determining the mode in the video encoder to overcome the transmission error environment by using a frame reconstructed by decoding in the encoder for motion compensation encoding An intra mode calculator for calculating a cost function value of the intra mode; A determination unit that determines whether the pixel to be the mode to be determined belongs to the still picture; An inter mode calculation unit for calculating an inter mode cost function value when there is no motion vector when the pixel belongs to a still picture, and an inter mode cost function value when there is a motion vector if the pixel does not belong to a still picture ; A comparison unit for comparing the result value in the intra mode calculator and the result value in the inter mode calculator; And a determination unit configured to determine the inter mode when the result value in the intra mode calculation unit is larger, and determine the intra mode when the result value in the intra mode calculation unit is smaller than or equal to the video encoder. And an image compression device having a processor for determining a mode.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. First of all, in adding reference numerals to the components of each drawing, it should be noted that the same reference numerals are used as much as possible even if displayed on different drawings. In addition, in describing the present invention, when it is determined that the detailed description of the related well-known configuration or function may obscure the gist of the present invention, the detailed description thereof will be omitted.

2 is a block diagram briefly illustrating an operation model of a video encoder and a decoder in the case of a bitstream transmission error.

Assuming a channel in which a transmission error occurs, the operations of the encoder 200 and the decoder 210 can be modeled as shown in FIG. 2. All currently adopted international video compression standards adopt a structure of encoding in units of frames, but the basic structure can be represented as shown in FIG. In intra mode, macroblocks are generally encoded through transform, quantization, and variable length encoding, in which case motion estimation and compensation are not needed. On the other hand, in the inter mode, motion estimation is performed on the current block, and the motion vector is obtained as the displacement of the predicted block with respect to the current block.

The main difference between assuming a channel with a transmission error and a channel without a transmission error is that the former uses a decoder with error concealment capability. Because a decoder with error concealment capability is used, it can be assumed that even if a part of the transmitted bit stream is lost or modified, the decoder will faithfully reproduce the image. As shown in FIG. 2, if an error exists in the transmission channel between the encoder 200 and the decoder 210. and This will have a different value, and thus will have different reference frames between the encoder 200 and the decoder 210. This results in the recognition of different images between the encoder 200 and the decoder 210 by referring to non-identical reference frames, which ultimately leads to deterioration of image quality. Such deterioration of image quality can be quantified by the following equation.

First, the n-th decoded frame reconstructed for the motion compensation encoding by the encoder 200 may be expressed as Equation 1, which is used as a reference image of the next frame.

here Is the estimated motion vector, and Quantization noise generated in the process of encoding in inter mode and intra mode, respectively, Reconstructs the nth decoded frame for motion compensation encoding, Is the nth input frame, Denotes the nth decoded frame after reconstruction for motion compensation encoding.

The decoding process in the decoder 210 is basically an inverse process of the coding process. If the reverse process is normally performed, as shown in Equation 2, the decoder 210 reconstructs the nth output frame for motion compensation encoding in the encoder and recognizes the same frame as the nth decoded frame.

here Is the nth output frame in decoder 210, Denotes the n-1 th output frame in the decoder 210.

However, if a transmission error is detected, the decoder 210 will have to switch to the error concealment mode according to the above assumptions. That is, in the decoder 210, it is most common to attempt to overcome an error by replacing information of a previous frame instead of information of a part damaged due to an error as shown in Equation 3 below.

Once the decoder 210 experiences the error concealment mode, not only distortion occurs in the reproduced image (pixel), but also distortion due to the image reproduced in the error concealment mode being used for prediction of the inter mode macroblock of the next frame. Propagates beyond the frame. This distortion is based solely on reference images to be used for motion compensation of the encoder 200 and decoder 210, i.e. Wow It is caused by the difference of and is caused by previous transmission errors. The difference may be expressed as in Equation 4, which is defined as drift noise.

If the quantization noise and the drift noise are assumed to be uncorrelated, the difference between the decoded image and the original image, that is, the total noise of the nth coded image May be expressed as shown in Equation 5.

From here Is Represent quantization noise and drift noise, respectively. That is, it may be expressed as in Equation 6.

In this case, quantization noise refers to noise generated during compression, and drift noise refers to noise caused by errors as mentioned above. In addition, since error concealment is a signal processing process performed by the decoder 210 when a transmission error occurs, a probability that a new drift noise occurs is equal to a loss probability of a corresponding packet. In addition, the intensity of the drift noise generated at this time is proportional to the spatiotemporal activity of the image, and shows a difference according to the type of error concealment technique. In addition, since the drift noise inherits the noise of the reference frame referenced by the pixel currently being encoded, the drift noise is affected by the drift noise of the reference frame. Therefore, if there is an error in the reference pixels of the reference frame, it continues to affect until an intra refresh occurs and becomes a major factor of image quality deterioration. In order to prevent this, when the n-th frame is encoded, the code mode of the corresponding macroblock may be determined by quantifying how the drift noise will change for each mode.

First, when the inter mode block information is normally transmitted, the decoder 210 will proceed with inter mode decoding in a normal condition. In this case, the drift noise included in the reproduced video may be expressed as in Equation 1 through Equation 2 through Equation 7.

That is, the drift noise of the previous frame compensated by the motion vector is inherited.

Next, if the intra mode block information is decoded under a normal condition, the drift noise to be included in the reproduced video may be expressed from Equation 1 and Equation 2 as shown in Equation 8.

On the other hand, if the current block information is lost regardless of whether it is inter mode or intra mode and the current video is reproduced by the error concealment technique, the drift noise included in the reproduced video is expressed as shown in Equations 1 and 3 through Equation 9. do.

That is, the new drift noise is added to the drift noise of the previous frame to appear as the drift noise of the current frame.

Using the probability of occurrence in each case to quantify the expected value of the drift noise of the current frame, the probability of packet loss is In this case, the expectation of the drift noise in the inter mode block can be numerically expressed as in Equation 10 from Equation 7 and Equation 9.

here Is a conversion constant that converts the squared value of the decoded frame difference signal into drift noise. This value depends on the type of error concealment technique, and becomes 1 if purely time-base error concealment is performed.

In addition, the new drift noise in the intra mode block can be quantified as in Equation 11 from Equations 8 and 9.

Equations 10 and 11 are the basis of the drift noise propagation model.

On the other hand, the drift noise propagation model described above generally matches well with the actual situation when the packet loss is well-dispersed, but is transmitted in the frame where the screen transition occurs in the case of the image which is still maintained after the screen transition occurs. When an error occurs, the error has a weakness that continues to propagate for a long time. The reason for this is as follows. As the drift noise calculated in Equation 11 proceeds to the frame, the drift noise is updated to Equation 10. Since the pixel belongs to the still picture, the difference signal between the motion vector and the reconstructed frame becomes zero, and thus the drift noise increases. The same value will be used over and over again. Once the first value does not make the coding mode of the block intra, after that, it will continue to be encoded in inter mode (inter mode without exact motion vector) as long as the still picture continues. Thus, the estimated drift noise will result in poor modeling of the actual drift error. Of course, the probability is very low, but when this actually happens, it can be seen that the image quality of the playback video is significantly lower than other cases. In order to solve this problem, the present invention modifies and uses the model to forcely increase the estimated value of the drift noise according to the number of consecutive inter modes. In other words, the drift noise is propagated and the stop area and the motion area are divided. or Multiply by Diffusion Constant, which is expressed as, to propagate to the next frame. here Is the index of the most recently encoded frame in intra mode Is the index of the current frame. The results are shown in the following equations (12), (13) and (14) for updating the drift noise and updating the emphasis constant for each mode.

First, when an inter mode without a motion vector, that is, a pixel belongs to a still picture and a difference signal between a motion vector and a reconstructed frame becomes 0, the drift noise value is numerically expressed as Equation 12.

here Denotes the diffusion constant in the stationary region. Is an element for updating the drift noise value in the static region. However, since there is no difference from the previous frame in the still area, if it is updated as it is, it will not be updated for a long time. Therefore, it should be set so that the value gradually increases over time, that is, as n increases, and the most suitable model is the exponential function. Through this process, the diffusion constant in the stationary region Is expressed by Equation 13.

Where M depends on the size of the image. For example, to be.

Next, in the inter mode when there is a motion vector, the drift noise value is numerically expressed as in Equation 14.

here Denotes the diffusion constant in the motion region. Is the update factor in the motion area, which is also Similarly, we design the value under the assumption that the probability of error propagation increases exponentially with time, and sets the exponential part of the exponential function as a quadratic function over time, spreading faster than in the stationary region. It is designed to be possible. Through this process, the diffusion constant in the motion region Is expressed by Equation 15.

Lastly, in the intra mode, the drift noise value is numerically expressed as in Equation 16 below.

In addition, the final intra mode frame index is updated as shown in Equation 17.

3 is a flowchart illustrating an optimized code mode determination process according to a preferred embodiment of the present invention.

According to the rules given in Equations 12, 14, and 16, the drift value of each pixel may be digitized in the nth frame. The encoder matches the drift noise value with the quantization noise value and the number of bits needed to encode the macroblock, so that the code mode of the macroblock is inter mode when there is a motion vector, inter mode when there is no motion vector, or It is possible to determine whether to encode in intra mode.

For example, the optimal mode may be encoded by calculating a value of the Cost function, which is a cost function defined as in Equation 18, and determining the mode in which the value is the smallest.

Where R represents the number of bits needed for encoding, and Lagrangian multiplier Is a constant that must be determined in advance according to an encoding parameter (quantization class, etc.) before the image is actually encoded.

Inter mode when there is a motion vector / inter mode or intra mode when there is no motion vector Since the algorithm has a value, the cost function in each mode is calculated and compared to each other to determine the optimal mode. The value is not specifically considered. In the present invention, used in the cost function We digitized the value of and consider only this value to determine the optimization mode.

An optimization mode determination process will be described with reference to the flowchart of FIG. 3. First, a cost function Cost_intra value in the intra mode is calculated (S300). At this time, The value of can be obtained from Equation 16. As described above, if the pixel belongs to the still picture and the difference signal between the motion vector and the reconstructed frame becomes 0, the drift noise will not increase and the same value will be used repeatedly. The drift noise was obtained by dividing by equation (14). Therefore, next, it is determined whether the corresponding pixel belongs to the still picture (S302) and the optimization mode determination process is performed. If the pixel belongs to the still picture, the cost function Cost_inter value in the inter mode when there is no motion vector is calculated (S304). At this time, The value of can be obtained from Equation 12. If the pixel does not belong to the still picture, the cost function Cost_inter value in the inter mode inter mode when there is a motion vector is calculated (S306). At this time, Can be obtained from Equation 14. Now, the optimal mode is determined by comparing the obtained cost_intra value with the cost_inter value (S308). Cost function Since the case where the value of the Cost function is the minimum is the optimized sign mode, if the Cost_intra value is greater than the Cost_inter value, the code is determined by the inter mode (S310). If the Cost_intra value is less than or equal to the Cost_inter value, the intra mode is determined (S312). Encode This determines the optimization mode to overcome the transmission error environment in the video encoder.

The above description is merely illustrative of the present invention, and those skilled in the art to which the present invention pertains may various modifications without departing from the essential characteristics of the present invention. Accordingly, the embodiments disclosed herein are not intended to limit the present invention but to describe the present invention, and the spirit and scope of the present invention are not limited by these embodiments. It is intended that the scope of the invention be interpreted by the following claims, and that all descriptions within the scope equivalent thereto shall be construed as being included in the scope of the present invention.

As described above, conventional error-tolerance methods have a problem that errors are continuously propagated by concealing an error, and there is a problem that there is a lot of loss of information when using a specific marker even in the case of error concealment. A special consideration for overcoming an error problem, that is, an efficient mode decision can be made by quantifying the effect that an error occurs at the position of each pixel in a frame on the basis of an error probability, thereby affecting image quality deterioration.

Therefore, in case of errors in the transmission environment, such as wireless environment, in the digital video encoder using the video compression technique, a method for effectively preventing the propagation of errors is proposed. It can contribute to the improvement of the performance in the video transmission system, including.

1 is a view showing an example of a video data packet generated in a conventional video coding apparatus,

2 is a block diagram briefly illustrating an operation model of a video encoder and a decoder in the case of a bitstream transmission error;

3 is a flowchart illustrating an optimized code mode determination process according to a preferred embodiment of the present invention.

Claims (26)

A method of mode determination in a video encoder for overcoming a transmission error environment, (a) calculating an intra cost function value in the intra mode; (b) Determining whether or not the pixel targeted for mode determination belongs to a still picture step; (c) when the pixel belongs to the still picture as a result of the determination in step (b) Calculating an inter cost function value in an inter mode when there is no motion vector; (d) As a result of the determination in step (b), the pixel does not belong to the still picture. Calculating an inter cost function in inter mode when there is a motion vector; (e) comparing the intra cost function value with the inter cost function value; (f) determining the inter mode if the intra cost function value is greater than the inter cost function value as a result of the comparison in step (e); And (g) determining the intra mode if the intra cost function value is less than or equal to the inter cost function value as a result of the comparison in step (e). Mode determination method in a video encoder for overcoming transmission error environment comprising a. The method of claim 1, wherein in step (a) The intra cost function is a video encoder for overcoming a transmission error environment, wherein the drift noise value in the intra mode, the quantization noise value generated during compression in the intra mode, and the number of bits required to encode the macro block are added. How to determine mode in The method of claim 2, The drift noise value is a noise value generated in the intra mode when the nth frame is encoded, and is expressed as a product of a probability of packet loss and a square value of a frame difference signal decoded according to an error concealment technique in the intra mode. A method of mode determination in a video encoder for overcoming a transmission error environment, characterized in that. The method of claim 3, wherein The decoded frame difference signal is a drift noise value of a previous frame. A method of mode determination in a video encoder for overcoming a transmission error environment, characterized in that a new drift noise value E is summed. The method of claim 4, wherein remind Is the nth decoded frame reconstructed for motion compensation encoding by the encoder, and the nth input frame and the quantization noise generated during the encoding in the intra mode are combined. Method of mode determination in encoder. The method of claim 1, wherein in step (c), The inter cost function is a sum of the drift noise value in the inter mode in the absence of the motion vector, the quantization noise value generated in the compression in the inter mode, and the number of bits required to encode the macroblock. Mode Decision Method in Video Encoder for Overcoming Error Environment. The method of claim 6, The drift noise value is a noise value generated in the inter mode when the n-th frame is encoded. The drift noise value is a previous value generated by the difference between the 1-p probability that the packet is not lost and the reference image to be used for the motion compensation of the encoder and the decoder. In the video encoder for overcoming the transmission error environment characterized in that the product of the transmission error square value of and the probability of packet loss and the product of the squared value of the decoded frame difference signal according to the error concealment technique in the inter mode is summed. How to determine the mode. The method of claim 6, The drift noise value is a noise value generated in the inter mode when an nth frame is encoded, and a diffusion constant in a still region. And a mode determination method in a video encoder for overcoming a transmission error environment, which is represented by drift noise of a previous frame. The method of claim 1, wherein in step (d), The inter cost function includes a drift noise value in an inter mode in the presence of the motion vector, a quantization noise value generated during compression in the inter mode, and a number of bits required to encode the macroblock. Mode Decision Method in Video Encoder for Environment Overcoming. The method of claim 9, The drift noise value is a motion vector due to previous transmission errors generated by a probability 1-p of packet loss and a difference between reference images to be used for motion compensation of an encoder and a decoder. Overcoming the transmission error environment characterized in that the product of the transmission error square value of the previous frame compensated by as much as the probability of packet loss and the product of the square value of the decoded frame difference signal according to the error concealment technique in the inter mode Method of Mode Determination in Video Encoder. The method of claim 9, The drift noise value is a noise value generated in the inter mode when the nth frame is encoded, and is a spreading constant in a motion region. And a mode determination method in a video encoder for overcoming a transmission error environment, which is represented by drift noise of a previous frame. The method according to claim 7 or 10, The decoded frame difference signal is a new drift noise value to the drift noise value of the previous frame. Mode is combined in a video encoder for overcoming transmission error environment. The method of claim 12, remind Is the nth decoded frame reconstructed by the encoder for motion compensation encoding. Represented by Is the nth input frame, Is an estimated motion vector, Is a quantization noise generated during the encoding in the inter mode. An image compression apparatus having a processor for determining a mode in a video encoder to overcome a transmission error environment by using a frame reconstructed by decoding in an encoder for motion compensation encoding. An intra mode calculator configured to calculate a cost function value in the intra mode; A determination unit that determines whether a pixel, which is a target of mode determination, belongs to a still picture; An inter mode calculation unit for calculating an inter mode cost function value when there is no motion vector when the pixel belongs to a still picture, and an inter mode cost function value when there is a motion vector if the pixel does not belong to a still picture ; A comparison unit for comparing the cost function value in the intra mode calculator with the cost function value in the inter mode calculator; And  As a result of the comparison in the comparison unit, if the cost function value in the intra mode calculation unit is larger, the inter mode is determined. If the cost function value in the intra mode calculation unit is smaller or equal, the determination is made as the intra mode. part And a processor configured to determine a mode in the video encoder. 15. The method of claim 14, wherein the cost function value of the intra mode is a sum of a drift noise value in the intra mode, a quantization noise value generated during compression in the intra mode, and a number of bits required to encode the macroblock. And a processor for determining a mode in a video encoder. The method of claim 15, The drift noise value is a noise value generated in the intra mode when the nth frame is encoded, and is expressed as a product of a probability of packet loss and a square value of a frame difference signal decoded according to an error concealment technique in the intra mode. And a processor for determining a mode in the video encoder. The method of claim 16, The decoded frame difference signal is a new drift noise value to the drift noise value of the previous frame. And a processor for determining a mode in the video encoder. The method of claim 17, remind Is the nth decoded frame reconstructed for motion compensation encoding by the encoder, and the nth input frame and the quantization noise generated during the encoding in the intra mode are combined to determine the mode. An image compression device having a processor. The method of claim 14, The cost function value of the inter mode in the absence of the motion vector is a drift noise value in the inter mode in the absence of the motion vector, the quantization noise value generated at the time of compression in the inter mode, and the bits necessary for encoding the corresponding macroblock. And a processor for determining a mode in a video encoder characterized by the sum of the numbers. The method of claim 19, The drift noise value is a noise value generated in the inter mode when the n-th frame is encoded. The drift noise value is a previous value generated by the difference between the 1-p probability that the packet is not lost and the reference image to be used for the motion compensation of the encoder and the decoder. The processor for determining the mode is a video encoder characterized in that the product of the transmission error square value of the product and the probability of packet loss p and the product of the square value of the decoded frame difference signal according to the error concealment technique in the inter mode. Image compression device provided. The method of claim 19, The drift noise value is a noise value generated in the inter mode when an nth frame is encoded, and a diffusion constant in a still region. And a processor for determining a mode in the video encoder, characterized by the drift noise of the previous frame. The method of claim 14, The cost function value of the inter mode in the presence of the motion vector is a drift noise value in the inter mode in the presence of the motion vector, the quantization noise value generated at the time of compression in the inter mode, and the number of bits required to encode the corresponding macroblock. And a processor for determining a mode in the video encoder. The method of claim 22, The drift noise value is a motion vector due to previous transmission errors generated by a probability 1-p of packet loss and a difference between reference images to be used for motion compensation of an encoder and a decoder. The product of the video encoder, characterized in that the product of the transmission error square value of the previous frame, and the probability of packet loss p and the product of the squared value of the decoded frame difference signal according to the error concealment technique in the inter mode is summed. The image compression device having a processor for determining. The method of claim 22, The drift noise value is a noise value generated in the inter mode when the nth frame is encoded, and is a spreading constant in a motion region. And a processor for determining a mode in the video encoder, characterized by the drift noise of the previous frame. The method of claim 20 or 23, The decoded frame difference signal is a new drift noise value to the drift noise value of the previous frame. And a processor for determining a mode in the video encoder. The method of claim 25, remind Is the nth decoded frame reconstructed by the encoder for motion compensation encoding. Represented by Is the nth input frame, Is an estimated motion vector, And a processor for determining a mode in the video encoder, wherein the quantization noise is generated in the inter mode encoding.