JP2002369206A - Device and method for selective encoding of dynamic region and static region - Google Patents

Abstract

PROBLEM TO BE SOLVED: To shorten the search time (movement prediction calculating time) for a motion vector by increasing the encoding efficiency. SOLUTION: A method of detecting a dynamic region and a static region and then encoding these regions selectively in video encoding comprises a process of sampling an inputted image into blocks which include two-dimensional arrays of pixels, a process of computing the discrete cosine transforms(DCT) of the sampling blocks of the image data to create DCT coefficients, a process of classifying the sampling DCT blocks into macroblocks, each of which comprises a luminance block and a chrominance block, a process of detecting the macroblocks as a static macroblock or dynamic macroblock, a process of predicting a movement based, on the information about the static or dynamic block, a process of quantizing a current macro block by controlling the rate, and quantization of the DCT coefficients, a process of selectively rejecting scanned quantized DCT factor, and a process of variable-length encoding of the scanned quantized DCT coefficients.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to compression of video signals, and more particularly to motion estimation of moving images.

[0002]

2. Description of the Related Art Video compression has been created to handle ever increasing information. In the race to provide more channels to consumers with higher quality, digital video has made significant progress in terms of quality, transfer and storage. With this in mind, MPEG-2
Memory capacity, transfer and high quality efficiency have been created for that purpose.

[0003] In MPEG-2 video compression, a series of pictures is coded based on picture types, ie, I, B and P pictures. I pictures are intra mode encoded, while P and B pictures are predictive mode encoded. Each picture is first divided into macroblocks to facilitate calculation and processing. The macro block (MB) is composed of 16 × 16 pixels. Each macroblock is further divided into 8 × 8 pixels called blocks.

FIG. 2 shows an implementation of a general MPEG-2 video encoder. In encoding an I-picture, first, a series of pictures is transformed into a frequency domain by a discrete cosine transform (DCT). Next, it is quantized by a quantization step, and finally stored for variable length coding (VLC).
Is encoded using.

In the coding of P and B pictures, a series of pictures is first predicted using motion prediction to generate a motion vector. Next, the difference between the current frame and the predicted position of the reference frame is encoded. Next, DCT, quantization, and variable length coding (VLC) are applied.

A picture consists of three components: luminance (Y), chrominance red (ie, red color difference signal Cr), and chrominance blue (ie, blue color difference signal Cb). MPE
In G-2, the Y component is sampled at all frequencies, while the chrominance is sampled horizontally and / or vertically at all or half frequencies.

The Y component has four blocks, whereas the Cb and Cr components have one, two or four blocks depending on the sampling used. Every block has 8 × 8 pixels.

[0008]

In a series of pictures, there are portions of a picture that normally do not move (ie, background) and portions that move temporarily (ie, foreground). However, the general motion estimation algorithm used does not recognize these characteristics and re-creates all of the pictures as if they were moving due to small changes such as changes in brightness or camera noise. Encode. This unnecessarily encodes parts that are indistinguishable by the viewer, and wastes valuable bits that are preferably used for moving parts. The reason for this is probably due to particulates on the smooth surface of the picture. FIG. 1 shows an enlarged macroblock from a picture showing texture formation. The human eye can only see pictures with a smooth surface, but the enlarged macroblock is as shown. By zooming into the enlarged macroblock area, the particles can be seen. However, encoding these movements does not improve the fines in quality.
On the other hand, the introduction of “jiglingness” degrades the picture quality. This is MPEG-2
This is shown in a table tennis sequence which is a test sequence for the game. In this sequence, the stationary wall moves after encoding.

Another problem arising from the above situation is that the wrong large motion vector is derived from the wrongly identified motion due to small bit changes in the still region.
This uses more valuable bits for encoding and wastes time searching for motion vectors.

SUMMARY OF THE INVENTION It is an object of the present invention to improve the coding efficiency and reduce the search time (motion estimation calculation time) for a motion vector.

[0011]

SUMMARY OF THE INVENTION A method for detecting and selectively encoding moving and still regions in video coding according to the present invention comprises the steps of sampling an input image into blocks containing a two-dimensional array of pixels. , A discrete cosine transform (DCT) on the sampling block of pixel data
Generating a DCT coefficient by performing the above-described steps, classifying the sampled DCT block into a macro block composed of a luminance block and a chrominance block, and detecting the macro block as a still macro block or a moving macro block. Performing motion prediction based on information of a still or moving macroblock, performing quantization by rate control on the current macroblock, quantizing the DCT coefficient, and selectively scanning the scanned quantized DCT coefficient. A method comprising the steps of discarding and performing variable length coding of the scanned quantized DCT coefficients, thereby achieving the above object.

The step of detecting a macroblock as a still or moving macroblock includes dividing the luminance and chrominance blocks in the macroblock into microblocks, which are smaller blocks; Calculating the difference between the microblocks, recording the difference between the reference frame and the microblock in the current frame, and determining the attribute of the macroblock using the recorded difference, Assigning still and moving macroblock labels to the attribute of the macroblock based on the recorded difference.

The step of performing motion prediction includes the steps of performing motion prediction on all the macroblocks having a moving macroblock attribute in the sampled image, and all the macroblocks having a still macroblock attribute in the sampled image. Skipping the motion prediction.

The discarding includes scanning each of the stationary microblocks having quantized DCT coefficients using a different scanning method depending on the DCT type, and scanning each block of the microblock within a particular location. Comparing the quantized DCT coefficient with a preset threshold; determining that the encoded block is a coded block if the comparison satisfies a first necessary condition; Discarding the block if satisfied.

[0015] The step of dividing may include a step of dividing the luminance block and the chrominance block into sub-micro blocks, which are small two-dimensional pixel blocks.

Calculating the difference between the reference frame and the current frame in the microblock; calculating the difference between the reference frame and the current frame in the luminance block; and calculating the difference between the reference frame and the current frame. Calculating a difference between the microblocks of the chrominance block between; assigning the macroblock having a difference value between the microblocks based on a preset condition; Using the difference between macroblocks.

Assigning the still and moving macroblock labels to the attributes of the macroblock, comparing the recorded difference of the macroblock of the luminance block with a first preset threshold value; The recorded difference of the block is a preset first
Marking the macroblock as a moving macroblock if the condition is satisfied; and defining the macroblock as a still macroblock if the recorded difference of the luminance block satisfies a second predetermined condition. Marking a block; and comparing the recorded difference of the macroblock of the chrominance block with the second preset threshold value;
Marking the macroblock as a moving macroblock if the comparison of the recorded difference of the chrominance block satisfies a third preset condition; and Marking the macroblock as a still macroblock if the comparison satisfies a preset fourth condition.

An apparatus for detecting and selectively encoding moving and still regions in video encoding according to the present invention comprises:
A sampling unit for applying a discrete cosine transform (DCT) to the sampling block of pixel data to generate a DCT coefficient; A classification unit that classifies the macroblocks into chrominance blocks, a detection unit that detects the macroblock as a still macroblock or a moving macroblock, and a motion prediction unit that performs motion prediction based on information of the still or moving macroblock. A quantization unit for performing quantization on the current macroblock by rate control to quantize the DCT coefficients, a post-processing unit for selectively discarding the scanned quantized DCT coefficients, A coding unit for performing variable length coding of the coefficients. The purpose is achieved.

A detecting unit that divides the luminance block and the chrominance block in the macro block into micro blocks, which are smaller blocks, and calculates a difference between the reference frame and the current frame in the current frame;
Record the difference between the reference frame and the microblock of the current frame, use the recorded difference to determine the attribute of the macroblock, and based on the recorded difference, store the attribute of the macroblock in the attribute. And a moving macroblock label may be assigned.

The motion prediction unit performs motion prediction on all the macroblocks having the moving macroblock attribute in the sampled image, and skips the motion prediction on all macroblocks having the still macroblock attribute in the sampled image. May be.

The discarding post-processor, which selectively discards the quantized DCT coefficients, scans each of the stationary microblocks having the quantized DCT coefficients using a different scanning method depending on the DCT type, and specifies a specific position. Comparing the quantized DCT coefficient of each block of the micro block with a preset threshold value, and if the comparison satisfies the first necessary condition, determines that the block is a coded block. 2
The block may be discarded when the required condition is satisfied.

[0022] The detecting section may divide the luminance block and the chrominance block into sub-micro blocks which are small two-dimensional pixel blocks.

The detection unit calculates a difference between the micro blocks of the luminance block between the reference frame and the current frame, calculates a difference between the micro blocks of the chrominance block between the reference frame and the current frame, and The macroblock having a difference value between the microblocks may be allocated based on a set condition and used as a difference between the macroblocks of a reference frame and a current frame.

When assigning the still and moving macroblock labels to the attributes of the macroblock, the detecting section compares the recorded difference of the macroblock of the luminance block with a first threshold value set in advance, and Marking the macroblock as a moving macroblock if the recorded difference of the block satisfies a preset first condition, and wherein the recorded difference of the luminance block is a predetermined second condition. If the condition is satisfied, mark the macroblock as a still macroblock, compare the recorded difference of the macroblock of the chrominance block with the second preset threshold, and A moving macroblock if the comparison of the recorded differences satisfies a third preset condition; Marking the macroblock and marking the macroblock as a still macroblock if the comparison of the recorded differences of the chrominance blocks satisfies a fourth preset condition. .

In the present invention, the moving area of the picture is
It has a macroblock known as a moving macroblock. The static areas of the picture have macroblocks known as still macroblocks.

Knowing that the general motion estimation algorithm is based on checking the value difference between the current macroblock in the current frame and the predicted macroblock in the reference frame, Check to see if the block gives the smallest difference. Due to changes in brightness and camera noise, this increases the difference in values even if the current macroblock is not moving from the reference frame to the current frame. As a result, the still area is also coded as being regarded as moving, and the reconstructed image looks like a moving object with respect to the still area.

Since the conventional motion estimation cannot distinguish between a moving area and a still area, this proposal aims to add a pre-processing step and a post-processing step before and after the motion estimation. At the same time, small changes have been made to the motion estimation process.

Before encoding a picture, the picture is first scanned to identify moving and still macroblocks. The marked moving macroblock undergoes motion prediction, while the still macroblock skips the motion prediction step. A certain threshold is determined and added to clearly distinguish between a moving macroblock and a still macroblock.

Since the I picture is encoded as it is,
These pictures do not go through the pre-processing steps. For P and B pictures, said pictures have different thresholds for the set of Y and UV planes. The B picture has a smaller threshold value than the P picture because the time difference between the reference frame and the B or P picture is smaller.

During motion estimation, macroblocks marked as still do not undergo motion estimation. The macroblock skipped in the motion prediction is encoded as a skipped macroblock or an encoded macroblock depending on the difference between the reference and the currently encoded macroblock. But,
There is obviously no motion vector.

The operation will be described below. A preprocessing step is added before the motion prediction step (FIG. 3). The pre-processing step checks for differences in pixel values between the current microblock in the current frame and the predicted block in the same location in the predicted frame.

The difference is calculated based on a number of small areas in the same microblock, the size of the microblock being the same as that used in standard, such as 16 × 16 in the case of MPEG2. When the four differences are calculated here, the size of the small area may be four 4 × 4 blocks.

The highest difference among the four differences is selected as a representative value, and it is determined whether or not a moving object exists in the 16 × 16 micro block. The same process is performed for both luminance and chrominance.

Based on the difference, a comparison is made with a defined threshold (FIG. 4). A stationary microblock is identified if the highest difference is much less than the threshold, while a dynamic microblock is identified if the highest difference is much greater than the threshold. The thresholds used here may be different for P or B, luminance or chrominance.

In this step, the identification for stationary or moving microblocks is recorded and marked, and in the next step it is used for motion estimation.

In the next step of motion estimation, those microblocks marked as stationary are skipped. These still microblocks are further encoded as skipped or encoded microblocks according to the standard encoding block rules (FIG. 5).

A post-processing step is added before variable length coding. In this post-processing step, blocks that do not contribute to the improvement of picture quality are discarded.

[0038]

DETAILED DESCRIPTION OF THE INVENTION The present invention solves the problem of incorrect motion vectors generated by conventional motion estimation and incorrect allocation of bits to macroblocks that are part of a still region in video. Things.

FIGS. 2 to 6 are views showing a preferred embodiment of the present invention. First, FIG. 2 is a block diagram showing the configuration and data flow of the encoder of the present invention. In FIG. 2, a block sampling module 201 samples an input video in a digital format in the form of 8 × 8 pixels. The motion prediction module 202 performs prediction on the current frame based on the reference frame stored in the reference frame memory module 208 in advance. The motion compensation module 203 reverses the motion prediction process of the motion prediction module 202 in order to store it as a reference frame. In the case of a P picture, only prediction mode coding is performed. For a B picture, it can be predicted from one frame or interpolated from two frames as in the frame interpolation module 209.

The result of the motion compensation module 203 is stored in the local decoded frame memory module 210 for future reference. In the DCT module 204,
The result of prediction transforms the time domain picture into the frequency domain via DCT (Discrete Cosine Transform). Next, it is quantized in quantization module 205 using a quantization step adjusted by rate control in rate control module 206.

After quantization, the run-length coding and variable-length coding module 207 performs run-length coding and variable-length coding (VLC), and then stores the picture as a bit stream. In the case of an I or P picture, the quantized picture undergoes an inverse DCT in an inverse DCT module 211 and is stored in a local frame memory for subsequent reference in a local decoding frame memory module 210.

FIG. 3 is a block diagram showing the configuration and data flow of an MPEG-2 video encoder to which pre-processing and post-processing steps have been added. Components with the same names perform the same operation. As shown in FIG. 3, in the present invention, a preprocessing step (preprocessing module 302) is added between the block sampling module 301 and the motion prediction module 303. In this step, macroblocks having different attributes are detected and marked. The input video sampled by the block sampling module 301 passes through a pre-processing module 302.
The input to the pre-processing module 302 is a macro block. The operation of the preprocessing module 302 will be described later. After pre-processing module 302, all macroblocks
It is an assigned still or moving macroblock.

The motion prediction module 303 performs motion prediction to derive a motion vector for a picture. DCT
Module 305 provides the D of the sampled block.
Calculate CT. The quantization module 306 calculates the prediction DC
Quantize the T coefficient. The quantization step is set by the rate control in the rate control module 307.

Next, the quantized DCT coefficient proceeds to a post-processing step (post-processing module 308). Blocks that do not affect picture quality are discarded. The blocks are later encoded into a bitstream after the run-length and variable-length encoding steps of the run-length encoding and variable-length encoding module 309.

After the quantization module 306, the quantized coefficient DCT passes through the inverse DCT module 312 to transform the picture back to the time domain, and stores this picture in the local frame memory of the local decoding frame memory module 311. .

FIG. 4 shows a detailed view of the pre-processing step. Module 401 first classifies each macroblock into two groups, a luminance group and a chrominance group. The luminance part consists of all luminance blocks, whereas the chrominance part consists of all Cr and Cb components.

Two separate checks are made for the luminance and chrominance groups. The checks are identical except for the threshold used for identification.

Each group is divided into smaller blocks called microblocks to increase the sensitivity of the detection method. For each microblock, the absolute difference between the microblock at the same position in the current frame and the microblock in the reference frame is calculated and the sum (SAD) is calculated. P
For pictures, only one set of SAD calculations is performed, ie, between the current frame and the reference frame. For a B picture,
Two sets of SAD calculations are performed. One is a forward reference frame and the other is a backward reference frame. However, only the maximum SAD value from the two calculations is taken to represent the currently sought macroblock. Macroblocks from an encoded I-picture do not undergo the pre-processing step.

The equation of SAD is as follows.

[0050]

(Equation 1)

Here, C and R indicate the current frame and the reference frame, and i and j indicate the positions of the pixels in the block.

After calculating the SAD for all microblocks, only the SAD value with the highest value is used to represent the SAD for the macroblock. The selected SAD value is
It is compared with a threshold value chosen depending on the type of picture.

The luminance SAD is checked against a predefined threshold (module 402), and if the SAD value is greater than the threshold, the macroblock is marked as a moving macroblock as in module 403.

[0054]

(Equation 2)

If the above check fails (when the SAD value is smaller than the threshold), the SAD for chrominance
Check the value. This is done because there is no significant change in luminance, but there is a large change in the color (chrominance) of the macroblock that indicates motion. This allows
The detection accuracy is improved.

Finally, if one or more of the above check routines succeeds, the macroblock being sought is marked as a moving macroblock (module 403); otherwise, the still macroblock (module 40) is marked.
Mark as 4). The module 405 checks whether there is another macroblock in the picture to be processed. If so, the next macroblock is processed according to modules 401-404. After processing all macro blocks, the next step is started.

Since this algorithm is used in conjunction with the motion prediction algorithm, some modifications are required when performing motion prediction. When checking the attribute of a macroblock as stationary, the motion prediction for the current macroblock is skipped. Thus, for the current picture being coded, there is no motion vector for the current macroblock. As a result, the motion estimation time is reduced.

When a macroblock is marked as a moving macroblock, a normal motion prediction algorithm is applied. The motion vectors generated from this step are normally encoded according to the standard.

Next, the following considerations will be given with reference to FIGS. Before writing the prediction results to the bitstream,
Further optimization takes place in the coding of the macroblock. This is a post-processing step of quantization. FIG. 5 is a diagram showing a procedure for selecting a quantized block for P and B pictures.

The module 501 checks to determine whether the current macroblock is an intra or moving macroblock. If the block is an intra or moving macro block, this step is skipped. If the macroblock is not an intra or moving macroblock, the method proceeds to module 502. In module 502, the first n
Check the quantized DCT coefficient at the _v position.

[0061] When more than the first n _v threshold v _s of both preset the coefficient at position coding a block (module 503). On the other hand, after confirming that all of the coefficients at the first _nv position are below the threshold, the block is discarded (module 504). After completion of the processing for the current block, the next block is processed until all blocks in the macroblock have been processed (modules 505 to 506).

The reason for performing the post-processing is as follows. For still macroblocks, P and B pictures are predicted from I pictures and are therefore inefficient to encode the difference. This is because human eyes cannot distinguish small changes in luminance or chrominance of the picture. However, the motion estimation algorithm is very sensitive to small changes in the characteristic. Therefore, in the present invention, such a block is selected and discarded.

However, the disposal is not performed properly.
Only the value at the first _nv position (where _nv > 0) is
It is determined by checking whether a preset threshold value v is less than _s are values. If any coefficient exceeds v _s , encode the block (module 503), otherwise continue to the next position. Check all coefficients at n _v locations and if no coefficients exceed v _s , discard block (module 504).

FIG. 6 shows the scanning method used and the method of discarding blocks. FIG. 6A is a diagram showing a general scanning order, and FIG. 6B is a diagram showing a coefficient order after scanning. As described in these figures, the coefficient of n _v position is rearranged so as to utilize the scanning order of the standards. Significant coefficients are placed first, and the significance of the values decreases as the position increases. In other words, low frequencies are given higher priority. (C) the threshold _{v s}
FIG. 11 shows the first _nv positions for all coefficients less than.
(D) is the first n _v for some coefficient above the threshold v _s
It is a figure showing a position. (E) is a diagram showing a preset threshold value with respect to v _s and n _v. The MPEG2 standard has two scanning methods, zigzag scanning and alternate zigzag scanning. The determination using any one of the scanning methods is as follows.
DC assigned to macroblock by encoder
Determined by T type.

[0065]

(Equation 3)

During the quantization of the picture, the internal bit counter of the encoder keeps track of the total bits allocated to each macroblock and each picture. As the number of motion vectors decreases (compared to coding without the described method), rate control
More bits can be provided to more required macroblocks such as edges and the like. This helps to emphasize the edges of the picture.

Also, due to the availability of bits due to discarding of block coding, bits available for allocation to more requested macroblocks (macroblocks having high activity such as edges, color change areas, etc.) become available. To increase.

According to such a method, the following effects can be obtained as shown in Table 1. Table 1 shows the results of tests performed using the algorithm according to the present invention. The test used contrasting foreground and background.

[0069]

[Table 1]

Norway is the result of having a stationary background of the building and a foreground of a person moving on the market. Fountai
n is a computer-generated image with a fixed background and a fountain of a fountain and a foreground of a moving rainbow. Popple is a blue background picture with three rotating objects. Poppl
et is the same picture as Popple, but has panning text. In addition, MSE (Mean Square Erro
r) is the mean square error, PSNR (Peak Signal to Noise)
Ratio) is an index for quantitatively evaluating the image quality of the compressed image. For the above experiments, the following thresholds were used. That is, for a P picture, V _{threshold_luminance} = 550, V
_{threshold_chrominance} = 85, for B picture, V _{threshold_luminance} = 650, V
_{theshold_chrominance} = 85.

The threshold value is selected from poplet coding. Optimally, each picture should have its own threshold depending on the content of the picture to be coded.

FIG. 7 shows a macroblock with motion detected before the motion estimation step. This is the output of the pre-processing step. FIG. 8 shows a motion vector detected by motion prediction with a conventional encoder. FIG. 9 shows a motion vector detected by motion prediction using the method according to the present invention.

[0073]

According to the present invention, each picture is first scanned for motion and still macroblock detection.
Next, the macroblock identified as a moving macroblock undergoes a normal motion estimation step, while the identified still macroblock is skipped for the motion estimation step. That is, it is determined whether the region is a still region or a moving region, and no motion prediction is performed on the still macroblock. As a result, the coding efficiency can be improved and the motion prediction calculation time can be reduced. The bits pre-allocated to the still macroblock can be used successfully for coding of the moving macroblock, while at the same time solving the instability problem around the still picture area of the reconstructed picture.

[Brief description of the drawings]

FIG. 1 shows an enlarged macroblock from a picture showing texture formation.

FIG. 2 is a block diagram showing the configuration and data flow of the encoder of the present invention.

FIG. 3 MPEG with pre-processing and post-processing steps added
FIG. 2 is a block diagram illustrating a configuration and data flow of a -2 video encoder.

FIG. 4 is a detailed flowchart of a preprocessing step.

FIG. 5 is a diagram showing a procedure for selecting a quantized block for P and B pictures.

FIG. 6 shows the scanning method used and the block discarding method used. (A) is a figure which shows a general scanning order.
(B) is a diagram showing the coefficient order after scanning. (C) is a diagram showing a first n _v position for all coefficients below a threshold value v _s. (D) is a diagram showing a first n _v position for a certain factor exceeds a threshold v _s. (E) is a diagram showing a preset threshold value with respect to v _s and n _v.

FIG. 7 is a diagram showing a macroblock having a motion detected before the motion prediction step.

FIG. 8 is a diagram showing a motion vector detected by motion prediction by a conventional encoder.

FIG. 9 is a diagram showing a motion vector detected by motion prediction using the method according to the present invention.

[Explanation of symbols]

 301 Block sampling module 302 Pre-processing module 303 Motion prediction module 304 Motion compensation module 305 DCT module 306 Quantization module 307 Rate control module 308 Post-processing module 309 Run-length encoding and variable length encoding 310 Frame interpolation module 311 Locally decoded frame Memory 312 Inverse DCT 313 Reference frame memory

 ──────────────────────────────────────────────────の Continued on the front page (72) Inventor Shen Mei Shen Singapore 534415 Singapore, Thailand Sen Avenue, Blocks 1022, 04-3530, Thai Sen Industrial Estate, Panasonic Singapore Research Institute, Inc. (72) Inventor Tech We Who Singapore 534415 Singapore, Thai Sen Avenue, Blocks 1022, 04-3530, Thai Sen Industrial Estate, Panasonic Singapore Research Institute F-term (reference) 5C059 KK17 LB01 MA00 MA23 MC11 MC38 ME01 ME05 NN24 NN28 PP05 PP06 PP07 PP16 PP26 PP27 PP29 TA00 TA45 TA66 TB07 TC04 TD11 UA02 5J064 AA02 BA09 BA16 BB03 BC14 BC16 BC25 BD01

Claims (14)

[Claims] 1. A method for detecting and selectively encoding moving and still regions in video encoding, comprising: sampling an input image into blocks containing a two-dimensional array of pixels; Applying a discrete cosine transform (DCT) to the sampling block to generate DCT coefficients; classifying the sampling DCT block into a macroblock consisting of a luminance block and a chrominance block; Detecting as a moving macroblock, performing motion prediction based on information of a still or moving macroblock, performing quantization on the current macroblock by rate control, and quantizing the DCT coefficient, Quantized DCT coefficients Method comprising the steps of discarding the 択的, and performing variable length coding of the scanned quantized DCT coefficients. 2. The method of claim 1, wherein detecting the macroblock as a still or moving macroblock comprises: dividing the luminance and chrominance blocks in the macroblock into smaller microblocks; Calculating the difference between the microblocks of the frame; recording the difference between the reference frame and the microblock of the current frame; and determining the attribute of the macroblock using the recorded differences. 2. The method of claim 1, comprising: assigning still and moving macroblock labels to the attributes of the macroblock based on the recorded differences. 3. The step of performing motion prediction includes: performing motion prediction on all the macroblocks having a moving macroblock attribute in the sampling image; and performing all motion prediction on all the macroblocks having a still macroblock attribute in the sampling image. Skipping motion estimation for the macroblock. 4. The step of discarding comprises: scanning each of the stationary microblocks with quantized DCT coefficients using a different scanning method depending on the DCT type; Comparing the quantized DCT coefficient of the block with a preset threshold; determining that the block is a coded block if the comparison satisfies a first requirement; Discarding the block if the condition is satisfied. 5. The method of claim 2, wherein the step of dividing comprises dividing the luminance block and the chrominance block into sub-micro blocks that are small two-dimensional pixel blocks. 6. The step of calculating a difference between a reference frame and a microblock of a current frame, the step of calculating a difference between the microblock of the luminance block between a reference frame and a current frame; Calculating a difference between the microblocks of the chrominance block between current frames; allocating the macroblock having a difference value between the microblocks based on a preset condition; Using as a difference between the macroblocks. 7. allocating still and moving macroblock labels to macroblock attributes: comparing the recorded difference of the macroblock of the luminance block to a first preset threshold; Marking the macroblock as a moving macroblock when the recorded difference of the luminance block satisfies a preset first condition; and wherein the recorded difference of the luminance block is preset. Marking the macroblock as a still macroblock when the second condition is satisfied, and comparing the recorded difference of the macroblock of the chrominance block with the preset second threshold. Step and said comparing said recorded difference of said chrominance blocks is preset Marking the macroblock as a moving macroblock when the third condition is satisfied, and when the comparison of the recorded difference of the chrominance block satisfies a fourth condition set in advance. Marking the macroblock as a still macroblock. 8. An apparatus for detecting a moving region and a still region in video encoding and selectively encoding the moving image and the still region, comprising: a sampling unit configured to convert an input image into a block including a two-dimensional array of pixels; A transform unit that performs a discrete cosine transform (DCT) on the blocks to generate DCT coefficients; a classifying unit that classifies the sampling DCT blocks into macroblocks including a luminance block and a chrominance block; Or a detection unit that detects a moving macroblock, a motion prediction unit that performs motion prediction based on information of a still or moving macroblock, a quantization unit that performs quantization on the current macroblock by rate control, and quantizes the DCT coefficient. And selectively discarding the scanned quantized DCT coefficients Apparatus having a post-processing unit, and an encoding unit that performs variable length coding of the scanned quantized DCT coefficients. 9. The detecting unit divides the luminance block and the chrominance block in the macro block into micro blocks, which are smaller blocks, calculates a difference between the reference frame and the current frame in the current frame, and calculates a reference frame. And the difference between the microblocks in the current frame is recorded, the attribute of the macroblock is determined using the recorded difference, and the attribute of the macroblock is statically and dynamically moved based on the recorded difference. The apparatus according to claim 8, wherein the apparatus assigns a macroblock label. 10. The motion prediction unit performs motion prediction on all the macroblocks having a moving macroblock attribute in the sampled image, and performs motion prediction on all macroblocks having a still macroblock attribute in the sampled image. 9. The apparatus according to claim 8, wherein said step is skipped. 11. A discard post-processor for selectively discarding quantized DCT coefficients, wherein each of the stationary microblocks having quantized DCT coefficients is scanned using a different scanning method depending on a DCT type. Comparing the quantized DCT coefficient of each block of the micro block in the position of the micro block with a preset threshold value, and when the comparison satisfies a first necessary condition, determines that the block is a coded block; 9. The apparatus according to claim 8, wherein the block is discarded when the second requirement is satisfied. 12. The apparatus according to claim 9, wherein the detection unit divides the luminance block and the chrominance block into sub-micro blocks, which are small two-dimensional pixel blocks. 13. The detecting unit calculates a difference between the micro blocks of the luminance block between a reference frame and a current frame, and calculates a difference between the micro blocks of the chrominance block between the reference frame and the current frame. Assigning the macroblock having a difference value between the microblocks based on a preset condition;
The apparatus according to claim 9, wherein the apparatus is used as a difference between a reference frame and the macroblock of a current frame. 14. The method according to claim 1, wherein the detecting unit assigns the still and moving macroblock labels to the attributes of the macroblock, and sets the recorded difference of the macroblock of the luminance block to a predetermined first one.
Comparing the recorded difference of the luminance block with the macro block if the recorded difference of the luminance block satisfies a preset first condition. Marks the macroblock as a still macroblock if a second predetermined condition is satisfied, and compares the recorded difference of the macroblock of the chrominance block with the second predetermined threshold. Comparing, marking the macroblock as a moving macroblock if the comparison of the recorded difference of the chrominance block satisfies a third preset condition; and If the comparison satisfies a preset fourth condition, the macro is regarded as a still macroblock. Mark lock, according to claim 9.