KR100622306B1 - Method for decoding video and apparatus thereof - Google Patents

Abstract

The present invention relates to an improved video encoding method. The video encoding method using a plurality of quantization matrices according to the present invention includes selecting one of the plurality of quantization matrices in units of macroblocks based on characteristics of an input image; And performing image transformation on the input image, and performing quantization on the image transformed data using the selected quantization matrix.

Description

Method for decoding video and apparatus

1 is a block diagram showing a conventional MPEG encoding and decoding apparatus.

2 illustrates an approximated generalized Wiener filter applied to an image whose mean is not zero.

3 shows an approximated generalized Wiener filter in the DCT region for non-average images.

4 (a) (b) (c) are diagrams showing the configuration of a filter applied to encoding of an intra block.

5 is a diagram illustrating a general video encoder applied to encoding of an inter block.

6 illustrates a video encoder according to an embodiment of the present invention.

7 is a diagram illustrating a video encoder according to an embodiment of the present invention.

8 illustrates a video decoder according to an embodiment of the present invention.

The present invention relates to a video encoding and decoding method and apparatus, and more particularly, to an encoding and decoding method and apparatus for improving video encoding efficiency by adaptively selecting a quantization matrix based on characteristics of an image input to a video encoder. It is about.

1 is a block diagram illustrating an encoder 120 and a decoder 140 for general video encoding.

For VOD service or video communication, the encoder 120 generates a bit stream encoded by a compression technique, and the decoder 140 performs a function of restoring an image from the input bit stream.

First, the DCT (Discrete Cosine Transform) unit 110 performs a DCT operation on image data input in units of 8 × 8 pixel blocks to remove spatial correlation, and the quantization unit (Q) 120 performs a DCT. A high efficiency lossy compression is performed by performing quantization on the DCT coefficients obtained by the unit 110 and expressing them as some representative values.

The inverse quantization unit (IQ) 130 inversely quantizes the image data quantized by the quantization unit 120. The IDCT unit 140 performs Inverse Discrete Cosine Transform (IDCT) transformation on the image data dequantized by the inverse quantization unit 130. The frame memory unit 150 stores image data converted by IDCT in the IDCT unit 140 in units of frames.

The motion estimation and compensation unit (ME / MC) 160 uses the image data of the input current frame and the previous frame random image data stored in the frame memory unit 150 to obtain a motion vector per macroblock (MV). ) And the sum of absolute difference (SAD) corresponding to the block matching error.

The variable length coding unit VLC 170 removes statistical redundancy from the DCT and quantized data according to the motion vector MV estimated by the motion estimator 160.

The bit stream encoded by the encoder 120 includes a decoder including a variable length decoder 142, an inverse quantizer 144, an IDCT unit 146, a frame memory unit 148, and a motion compensation unit 150. 140).

Such a video encoder is also disclosed in US Pat. No. 6,480,539.

Recently, set-top boxes that receive analog terrestrial broadcasting and encode and store them using compression methods such as MPEG 2 and MPEG 4 have been introduced. However, in the case of terrestrial broadcasting, the video at the receiving end is often distorted with noise by the channel. For example, there may be a case where white gaussian noise is added to the entire image. If such an image is compressed as it is, the compression efficiency is lowered due to noise.

Therefore, in the conventional video encoding method, a preprocessing filter is applied to an input of an encoder to remove noise. However, when using a preprocessing filter, there is a problem that an additional operation process is required for video encoding.

In addition, in the conventional video encoding method, quantization is performed by applying a quantization matrix determined in a picture unit irrespective of characteristics of an input image. However, in this case, there is a problem that video encoding is not performed efficiently.

An object of the present invention is to improve a conventional video encoding and decoding method as described above, and to provide a video encoding and decoding method and apparatus for improving video compression efficiency and performance.

Another object of the present invention is to provide a video encoding and decoding method and apparatus for improving the conventional video encoding and decoding method and performing video encoding and decoding by removing noise without increasing an amount of computation. do.

In the video encoding method using a plurality of quantization matrix according to the present invention, the step of selecting one of the plurality of quantization matrix in macroblock units based on the characteristics of the input image, and the image for the input image And performing a transformation and performing quantization on the image transformed data using the selected quantization matrix.

The present invention also provides a video decoding method using a plurality of quantization matrices, the method comprising: variable length decoding input coded image data, and classifying the input coded image data in association with characteristics of an input image from the variable length decoded image data. Extracting index information specifying one of a plurality of quantization matrices in macroblock units, selecting one of the plurality of quantization matrices using the extracted index information, and using the selected quantization matrix, And performing inverse quantization on a macroblock basis for the variable length decoded image data.

In addition, the above object is to provide a video encoding apparatus using a plurality of quantization matrix, to select one of the plurality of quantization matrix in macroblock units based on the characteristics of the input image, and to obtain index information indicating the selected quantization matrix A quantization matrix determiner generated in units of macroblocks and a plurality of quantization matrices classified in association with characteristics of an image are stored, and the quantization matrixes are output in macroblock units based on index information generated in the quantization matrix determiner. A quantization matrix storage unit, an image transformation unit performing image transformation on the input image, and a quantization matrix output from the quantization matrix storage unit on the image-converted image data to quantize the macroblock. It is achieved by an encoding device including a quantization unit for performing the operation.

The present invention also provides a video decoding apparatus using a plurality of quantization matrices, wherein the encoded video stream is input, variable length decoded, and one of the plurality of quantization matrices classified in association with characteristics of the input image is identified. A variable length decoder for extracting index information in units of macroblocks, the plurality of quantization matrices, and selecting one of the stored plurality of quantization matrices using the index information extracted by the variable length decoder, Achieved by a decoding apparatus including an output quantization matrix storage unit and an inverse quantization unit for performing inverse quantization of the variable length decoded image data in macroblock units using the quantization matrix output from the quantization matrix storage unit do.

Hereinafter, a noise removing method used in the video encoding method according to the present invention will be described with reference to FIGS. 2 to 5.

In video encoding, preprocessing filtering is very important in that encoding efficiency is improved by removing noise mixed in an image. Unlike conventional preprocessing filtering for noise removal, which is performed mostly in the spatial pixel region, the noise removing method according to the present invention is characterized in that it is performed in the DCT region in the encoder.

An embodiment for explaining the noise cancellation method used in the present invention uses approximated generalized Wiener filtering. Here, the approximated generalized Wiener filtering refers to a filtering method that enables to implement Wiener filtering by using fast unitary transformation such as DCT. In addition to selectively approximated Wiener filtering, it is also possible to use other filtering techniques for filtering in the DCT domain.

2 shows an approximated generalized Wiener filtering block diagram for image data whose mean value is not zero.

은 필터 링된 영상 블록의 행-정렬된 열 벡터(row-ordered column vector)를 의미한다. 입력 영상 v는 보통 0이 아닌 평균을 갖는 영상 블록이므로, 평균값 예측부(210)에서는 해당 블록의 평균값

을 예측하고, 감산부(220)에서는 입력 영상 v로부터 예측된 평균값

을 감산한다.In FIG. 2, the input v denotes an image block including noise, and thus an output.

Denotes a row-ordered column vector of the filtered image block. Since the input image v is usually an image block having an average other than 0, the average value predictor 210 determines the average value of the block.

And the subtractor 220 estimates an average value estimated from the input image v.

Subtract

를 출력한다. 가산부(240)는 필터링된 데이터

에 평균값 계산부(210)에 의해 계산된 해당 블록의 평균값

를 가산하여 원하는 필터링된 데이터

를 출력한다.The output value z in the subtractor 220 is filtered by the filtering unit 230, and the filtering unit 230 filters the filtered data.

Outputs The adder 240 filters the filtered data.

The average value of the corresponding block calculated by the average value calculation unit 210

To add the desired filtered data

Outputs

Hereinafter, a generalized Wiener filtering method for an image model having an average value of 0 will be described.

Generalized Wiener filtering for an image model having an average value of 0 may be described as in Equation 1 below.

,

,

, Z=Az이고,

은 노이즈 분산값(noise variance)이다. 또한, A는 유니터리 변환을 의미하는데, 본 실시예에서는 DCT 변환을 사용하므로, 본 실시예의 경우에는 A는 DCT 변환을 의미한다. 또한,

가 8×8 DCT 행렬을 의미하고,

가 크로네크(kronecker) 곱 연산자를 의미한다고 할 때, A = (

)으로 표현할 수 있다.From here,

,

,

, Z = Az,

Is a noise variance. In addition, A means unitary conversion. In this embodiment, since DCT conversion is used, A means DCT conversion in this embodiment. Also,

Denotes an 8 × 8 DCT matrix,

Is a kronecker multiplication operator, A = (

Can be expressed as

이 대부분의 유니터리 변환에서 근사적으로 대각화(diagonal)되기 때문에, 수학식 1은 아래의 수학식 2로 표현될 수 있다.

Equation 1 can be expressed by Equation 2 below because it is approximately diagonal in most of the unitary transformation.

From here,

 to be.

Therefore, by applying Equation 2 to an 8x8 block, Equation 3 below can be obtained.

은 수학식 4와 같다.From here,

Is the same as Equation 4.

이다.

to be.

은

의 대각선에 놓이는 정규화된 요소값들이며,

은 원 영상 y의 분산값을 의미한다. 하지만, 일반적으로

을 알 수 없으므로, z의 분산값에서 노이즈 분산값을 뺀 값을

으로 대체한다.

silver

Normalized element values on the diagonal of,

Denotes the variance of the original image y. But generally

Since we don't know, the variance of z minus the noise variance

Replace with

을 곱함으로써 이루어짐을 알 수 있다. 일단,

이 결정되면, 최종 필터링된 영상은

에 평균값

을 더함으로써 얻어진다.Equation 3 shows that for an averaged zero image block, the approximated generalized Wiener filtering is applied to two-dimensional DCT coefficients.

It can be seen that by multiplying. First,

Once this is determined, the final filtered image

Average value on

It is obtained by adding

Hereinafter, a generalized Wiener filtering scheme for an image model whose average value is not 0 will be described.

Assuming that the average block is obtained by multiplying S (k, l) by the noisy input DCT block, i.e., Equation 5 holds, Figure 3 is a structure including addition and subtraction of the mean value in the DCT domain. It may be reconfigured as shown in FIG.

Using Equations 5 and 3 according to the above assumptions, the filtered image block in the DCT region may be expressed as Equation 6 below.

Here, F (k, l) is expressed as in Equation 7 below.

That is, it can be seen from Equation 6 that the entire filtering process is unified with F (k, l) and one multiplication. In Equation 7, F (k, l) is determined by a signal to noise ratio (SNR), a covariance matrix, and an average matrix.

First, it is necessary to find the average matrix S (k, l). In this embodiment, S (k, l) that satisfies the expression (5) is selected. In this embodiment, the DC value is taken from the DCT block, which is the simplest method, as shown in Equation 8 below.

Hereinafter, the preprocessing process in the video encoder will be described with reference to FIGS. 4 and 5.

In the above description, the generalized Wiener filtering, which is approximated even for an image block whose average of the input image is not 0, may be performed as a multiplication of DCT coefficients.

4 shows that the approximated generalized Wiener filtering scheme is applied to a video encoder. 4 shows an encoder structure for an intra block. 4 (a) and 4 (b) show that in the case of an intra block, quantization and VLC may be performed immediately after performing filtering in the DCT region without performing an IDCT process.

That is, the filtering is completed by multiplying the DCT coefficients by F (k, l). On the other hand, the quantization process is also a task of multiplying each of the DCT coefficients by a specific value according to the quantization table. Therefore, the operation of multiplying F (k, l) by each of the DCT coefficients and the quantization operation may be combined into one operation as shown in FIG.

The concept applied to FIG. 4 may be equally applied to inter blocks under the assumption that the noise prediction block information p (m, n) by motion compensation has already been removed, as shown in FIG. 5. .

In general, since the covariance ψ (k, l) is determined according to whether the input image block is an inter block or an intra block, F (k, l) of Equation 7 is also changed and applied according to the block mode.

Hereinafter, a method of obtaining the variance prediction value for the intra and inter blocks in which the average value is subtracted will be described with reference to Equation (9). When S is an N × N (N = 8) block in which the mean is subtracted, the variance matrix of the block can be obtained by Equation (9).

Equation 9 is IEEE Trans. Circ. Syst. "Covariance analysis of motion-compensated frame differences" by W. Niehsen and M Brunig, for Video Technol.

Meanwhile, the variance prediction value may be obtained by applying Equation 9 to various experimental images. In the case of the intra block, the original image is divided into 8 × 8 blocks, and in the case of the inter block, the global prediction method is used to collect only the blocks selected as the inter blocks and apply Equation 9 to calculate the variance prediction value.

을 구하고, 다시 R을 DCT 변환하여

가 얻어진다.From the calculated variance prediction

And then again convert DCT to R

Is obtained.

를 구하는 방법을 설명한다.In the following,

How to obtain.

에서 노이즈 분산값

은 노이즈 예측기로부터 얻어질 수 있다. 또한, 노이즈와 원 영상 화소들은 랜덤 변수(random variable)이라고 가정할 때 통상 독립적인 것으로 간주되므로, 원 영상의 분산값

의 예측치

은 아래 수학식 10을 사용하여 예측된다.

Noise variance

Can be obtained from a noise predictor. In addition, since the noise and the original image pixels are generally regarded as independent when a random variable is assumed, the variance value of the original image

Estimate

Is predicted using Equation 10 below.

은 매크로블록(macro block: MB)의 분산값으로서, 통상의 동영상 부호화기에서는

은 매크로블록 단위로 계산된다. 본 실시예에서는 동일 매크로블록에 포함된 8 ×8 블록들은 동일한 분산값을 갖는 것으로 가정한다. 따라서, 분산값 계산을 위해 별도의 연산량을 요구하지는 않는다.From here,

Is a variance value of a macro block (MB). In a typical video encoder,

Is calculated in macroblock units. In this embodiment, it is assumed that 8 x 8 blocks included in the same macroblock have the same dispersion value. Therefore, it does not require a separate calculation amount for calculating the variance value.

6 is a diagram illustrating an embodiment of a video encoder in consideration of characteristics of an input image according to the present invention.

In the present embodiment, the quantization matrix is adaptively applied in consideration of the degree of noise included in the input image among the characteristics of the input image.

Hereinafter, an embodiment of the present invention will be described with reference to FIGS. 1 and 6.

The video encoder employing the noise elimination scheme according to the present invention is provided in the general video encoder of FIG. 1 by a noise estimation unit 680, a quantization weight matrix determiner 692, and a quantization weight matrix storage unit 694. It includes.

The DCT unit 610, the IDCT unit 640, the frame memory unit 650, and the motion prediction and compensation unit 660 perform the same operations as the corresponding functional units in the general video encoder of FIG. 1. Detailed descriptions are omitted for simplicity.

와 움직임 예측 및 보상부(660)로부터 전송된

에 기초하여, 해당 매크로블록에 대응하는 양자화 가중치 행렬을 결정하고, 이에 대응하는 인덱스 정보를 양자화 가중치 행렬 저장부(694) 및 가변 길이 부호화부(variable length coding unit: VLC)(670)로 전송한다.The quantization weight matrix determiner 692 is a noise variance value transmitted from the noise predictor 680.

And motion transmitted from the motion prediction and compensation unit 660.

The quantization weighting matrix corresponding to the macroblock is determined based on the quantization weighting matrix, and the corresponding index information is transmitted to the quantization weighting matrix storage unit 694 and the variable length coding unit (VLC) 670. .

와 움직임 예측 및 보상부(660)로부터 전송된

에 기초하여 각각의 매크로블록에 대응하는 양자화 가중치 행렬을 결정하는 과정을 설명한다.Hereinafter, the noise variance value transmitted from the noise predictor 680 by the quantization weight matrix determiner 692.

And motion transmitted from the motion prediction and compensation unit 660.

A process of determining a quantization weight matrix corresponding to each macroblock will be described based on the following.

은 양자화 과정에서 양자화 가중치 행렬에 의해 나누어진다. In accordance with Equation 8 and the description associated with FIGS. 4 and 5 above, F (k, l) is determined by Equation 7. Once F (k, l) is determined, as shown in Fig. 4 (c), the DCT transformed coefficients V (k, l) of the 8x8 block are each multiplied by the corresponding F (k, l), As a result

Is divided by the quantization weight matrix in the quantization process.

을 양자화 가중치 행렬로 나누는 과정을 하나의 과정으로 합쳐서 수행한다. 즉, 양자화 가중치 행렬(QT)의 (k,l) 위치 성분을 Q(k,l)이라고 하는 경우, 새로운 양자화 가중치 행렬(QT')의 (k,l)의 위치 성분은 Q(k,l)/F(k,l)이 된다.In the video encoding apparatus according to the present invention, the DCT transformed coefficients V (k, l) are multiplied by F (k, l),

The process of dividing into a quantization weight matrix is performed by combining the process into one process. That is, when the (k, l) position component of the quantization weight matrix QT is called Q (k, l), the position component of (k, l) of the new quantization weight matrix QT 'is Q (k, l). ) / F (k, l).

와

에 따른 여러 개의 F 행렬들을 미리 계산하여, 이에 다른 새로운 양자화 가중치 행렬(QT')들을 계산하여 양자화 가중치 행렬 저장부(694)에 저장한다.In this embodiment, by using the method of combining the two processes into one process,

Wow

Calculate a plurality of F matrices in advance, calculate other new quantization weight matrices QT ', and store them in the quantization weight matrix storage unit 694.

와

에 따른 5개의 새로운 양자화 가중치 행렬 (QT')이 양자화 가중치 행렬 저장부(694)에 저장된다.

와

이 결정되면, 수학식 10을 사용하여

을 구할 수 있다.In this embodiment,

Wow

Five new quantization weighting matrices QT 'are stored in the quantization weighting matrix storage unit 694.

Wow

Once this is determined, use Equation 10

Can be obtained.

에 의해 결정되는데, S(k,l)은 수학식 8에 의해 결정되고, Ψ(k,l)은 입력 영상이 인터 블록인 지 인트라 블록인지 여부에 따라 결정되는 블록 모드에 따라 변경되어 적용된다. 따라서, F(k,l)을 결정하기 위한 유일한 변수는

이다. 본 실시예에서는,

을 5가지 경우로 분류하여, 다섯개의 새로운 양자화 가중치 행렬(QT')를 생성하여, 이를 양자화 가중치 행렬 저장부(694)에 저장한다.Referring to Equation 7, F (k, l) is S (k, l), Ψ (k, l), and

S (k, l) is determined by Equation 8, and Ψ (k, l) is changed and applied according to a block mode determined according to whether an input image is an inter block or an intra block. . Thus, the only variable for determining F (k, l) is

to be. In this embodiment,

Are classified into five cases, and five new quantization weighting matrices QT 'are generated and stored in the quantization weighting matrix storage unit 694.

와 움직임 예측 및 보상부(660)로부터 입력된 해당 매크로블록의

에 기초하여

를 양자화한다. 이때, 양자화된 값은 해당 매크로블록에 대응하는 양자화 행렬의 인덱스 정보로서 양자화 가중치 행렬 저장부(692) 및 가변 길이 부호화부(670)로 전송된다. The quantization weight matrix determiner 692 may determine the macroblock input from the noise predictor 680.

And the macroblock inputted from the motion prediction and compensation unit 660.

Based on

Quantize In this case, the quantized value is transmitted to the quantization weight matrix storage unit 692 and the variable length encoder 670 as index information of the quantization matrix corresponding to the macroblock.

에 기초해서 5가지 경우로 분류되는 경우, 상기 양자화는 5 단계로 양자화 되고, 인덱스도 0, 1, 2, 3, 4 중 하나의 값이 된다.For example, the quantization weight matrix stored in the quantization weight matrix storage unit 694 is

Based on five cases, the quantization is quantized in five steps, and the index is one of 0, 1, 2, 3, and 4, respectively.

이 매우 커지게 된다.

이 매우 큰 값을 갖는 경우, 수학식 7 및 8로부터 알 수 있는 바와 같이, 대부분의 F(k,l)이 0에 가까운 수가 되어 블록화 현상이 심화되는 경향이 있다. 따라서, 이런 현상을 막아주기 위해 아래 수학식 11에서와 같이 T_Cutoff를 사용한다.Meanwhile, in the case of blocks having a small variance value in an image having a high noise, that is, a large noise variance value,

This becomes very large.

In the case of having this very large value, as can be seen from equations (7) and (8), most of F (k, l) becomes a number close to zero, and the blocking phenomenon tends to be intensified. Therefore, T _Cutoff is used as shown in Equation 11 to prevent this phenomenon.

Usually, T _Cutoff uses the value of about 1-2.

The quantization weight matrix storage unit 694 transmits the quantization weight matrix to the quantization unit 620 and the inverse quantization unit 630 using the index information input from the quantization weight matrix determination unit 692.

The quantization unit 620 performs quantization on the macroblock using the input quantization weight matrix.

The inverse quantization unit 630 also performs inverse quantization on the macroblock using the input quantization weight matrix.

The variable length encoder 670 variable-length encodes the input image data quantized by the quantizer 620, and also obtains index information of the quantization weight matrix corresponding to the corresponding macroblock input from the quantization weight matrix determiner 692. Insert into macroblock header.

In this embodiment, the index of the quantization weight matrix is inserted into the macroblock header and transmitted. In this case, when the number of matrices stored in the quantization weight matrix storage unit 694 is 10, additional information amount of 4 bits per macroblock is required.

Adjacent macroblocks have similar characteristics of the input image, so that index values are related. Therefore, it is also possible to selectively use the difference value obtained by subtracting the index of the current index and the neighboring macroblock as index information. When using the same quantization weight matrix for the entire sequence, it is possible to significantly reduce the amount of index information transmitted.

In the present embodiment, it is assumed that a plurality of quantization weight matrix information stored in the quantization weight matrix storage unit 694 is also stored in the decoder. However, optionally, the plurality of quantization weight matrix information is transmitted to the decoder in a picture unit through a picture extension header or a plurality of quantization weight matrix information in sequence units through a sequence extension header. It is also possible to use a quantization weight matrix transmitted to the decoder and transmitted in a picture unit or a sequence unit.

As described above, by adaptively applying the quantization matrix in macroblock units in consideration of the degree of noise included in the input image among the characteristics of the input image, it is possible to remove the noise and improve the coding efficiency.

In addition, the new quantization weight matrices can be arbitrarily determined by the user. Although the present embodiment shows a noise canceling method in the DCT region with respect to the Y component of the input image block, the same apparatus can be applied to the U and V components other than the Y component. At this time, separate weight matrices for the U and V components are needed.

7 is a diagram illustrating another embodiment of a video encoder in consideration of characteristics of an input video according to the present invention.

The characteristic of the input image considered in this embodiment is the edge characteristic in each macroblock of the input image.

The video encoder according to the present embodiment further includes a quantization matrix determiner 780 and a quantization matrix storage 790 in the general video encoder of FIG. 1. The DCT unit 710, the IDCT unit 740, the frame memory unit 750, the motion prediction and compensation unit 760, and the variable length encoder 770 perform the same functions as those of the general video encoder of FIG. 1. Detailed description is omitted for the sake of simplicity.

The quantization matrix determiner 780 selects an optimal quantization matrix in units of macroblocks based on characteristics of the input image, and stores index information corresponding to the selected quantization matrix by the quantization matrix storage 790 and the variable length encoder 770. To send).

In the present embodiment, the quantization matrix determiner 780 considers edge characteristics of the macroblock as a reference for selecting one of a predetermined number of quantization matrices.

Hereinafter, a method of selecting a quantization matrix based on the edge characteristics of the macroblock will be described in detail.

First, if the corresponding macroblock of the input image is an intra block, an existing known edge detector such as a sobel operator is used to calculate the edge size per pixel in the macroblock, and the The direction is also calculated. The Sobel operator is shown in Equation 12 below.

The quantization matrix determiner 780 calculates the vertical edge size and the horizontal edge size using Equation 12, and calculates the strength of the edge and the direction of the edge in the macroblock using the equation 12. One of the predetermined numbers of quantization matrices is selected in consideration of coding efficiency based on the calculated strength and direction of the edge of the macroblock. That is, for a macroblock having a vertical edge or a macroblock having a horizontal edge, a quantization matrix capable of performing quantization while selecting these edges is selected.

On the other hand, in the case of the inter block, it is also possible to obtain the edge size and direction using an edge detector such as Sobel.

In this embodiment, the Sobel detector is used to calculate the edge characteristics of the macroblock, that is, the strength and directionality of the edges. However, other arbitrary spatial filters such as a difference filter and a Roberts filter may be selectively used. It is also possible to use.

In the present embodiment, the quantization matrix is selected in consideration of the strength and the direction of the edge of the macroblock as characteristics of the input image. However, it is also possible to selectively select predetermined quantization matrices in macroblock units in consideration of other characteristics of the input image in terms of encoding efficiency or quality improvement.

The quantization matrix storage unit 790 selects the quantization matrix based on the quantization matrix index information input from the quantization matrix determination unit 780 and transmits the quantization matrix to the quantization unit 720 and the inverse quantization unit 730.

The quantization unit 720 performs quantization using the input corresponding quantization matrix.

The inverse quantization unit 730 also performs inverse quantization using the input corresponding quantization matrix.

The variable length encoder 770 variable length encodes the input image data quantized by the quantizer 720, and also variable lengths an index of the quantization matrix corresponding to the corresponding macroblock input from the quantization weight matrix determiner 780. Encode At this time, the index information is inserted into the macroblock header.

In this embodiment, the index of the quantization weight matrix is inserted into the macroblock header and transmitted. However, it is also possible to selectively use the difference value obtained by subtracting the index of the current index and the adjacent macroblock as index information.

In the present embodiment, it is assumed that a plurality of quantization weight matrix information stored in the quantization matrix storage 790 is also stored in the decoder. However, it is also possible to selectively transmit a plurality of quantization weight matrix information to a decoder in picture units through a picture extension header or to transmit a plurality of quantization weight matrix information to a decoder in sequence units through a sequence extension header. Do.

8 is a diagram illustrating a video decoder according to an embodiment of the present invention.

The video decoder according to the present invention further includes a quantization weight matrix storage unit 860 in the general video decoder 140 of FIG. 1.

Since the IDCT unit 830, the frame memory 840, and the motion compensator 850 perform the same functions as the general video decoder in FIG. 1, detailed descriptions thereof will be omitted for simplicity.

The variable length decoding unit (VLD) 810 variable length decodes the input stream, extracts index information of the quantization weight matrix corresponding to the macroblock from the macroblock block header of the input stream, and extracts the index information. The index information is output to the quantization weight matrix storage unit 860.

The quantization weight matrix storage unit 860 outputs the quantization weight matrix corresponding to the index to the inverse quantization unit 820 based on the input index information. Here, the quantization weight matrix storage unit 860 has a characteristic of an input image processed by an encoder, for example, a noise degree such as a ratio between a noise variance value and an input image variance value, and a spatial characteristic such as an edge degree of the input image. A plurality of weighting matrices classified according to are stored.

Optionally, the plurality of quantization weight matrix information stored in the quantization weight matrix storage unit 860 may be transmitted in picture units through a picture extension header or in sequence units through a sequence extension header. In this case, the plurality of transmitted quantization weight matrix information is input from the variable length encoder 810 to the quantization weight matrix storage unit 860 as indicated by a dotted line in FIG. 8.

The present invention is not limited to the above-described embodiment, and of course, modifications may be made by those skilled in the art within the spirit of the present invention. In particular, the present invention can be applied to all video encoding and decoding methods and apparatuses such as MPEG-1, MPEG-2, MPEG 4, and the like.

The invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, hard disk, floppy disk, flash memory, optical data storage, and also carrier wave (for example, transmission over the Internet). It also includes the implementation in the form of. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

As described above, in the video encoding and decoding method according to the present invention, it is possible to improve coding efficiency and performance by adaptively applying a quantization matrix on a macroblock basis based on characteristics of an input image.

Claims (6)

A method of performing video decoding by adaptively applying a plurality of quantization weight matrices to a macroblock, Receiving an encoded input video stream comprising at least one macroblock; Performing variable length decoding on the encoded stream; Performing inverse quantization on each of the variable length decoded macroblocks using a plurality of quantization weighting matrices specified for each macroblock, The plurality of quantization weighting matrices includes a quantization weighting matrix used for the U component and the quantization weighting matrix used for the V component, and the header of each macroblock includes a plurality of quantization weighting matrices for the Y component. And index information for specifying one of the quantization weighting matrices. The method of claim 1, The variable length decoding step includes extracting index information of each quantization weight matrix from a macroblock header. The method of claim 2, And the index information is a difference between an index of a quantization weighting matrix corresponding to one macroblock and an index of a quantization weighting matrix corresponding to a macroblock adjacent to the macroblock. An apparatus for performing video decoding by adaptively applying a plurality of quantization weighting matrices to a macroblock, A variable length decoder which receives an encoded input video stream including at least one macroblock and performs variable length decoding on the received input video stream; A quantization weighting matrix storage unit for storing the plurality of quantization weighting matrices, selecting one corresponding to each component of the plurality of quantization weighting matrices, and outputting the selected quantization weighting matrix; An inverse quantization unit configured to perform inverse quantization on each of the variable-length decoded macroblocks using the quantization weighting matrix output from the quantization weighting matrix storage unit, The plurality of quantization weighting matrices includes a quantization weighting matrix used for the U component of the input video and a quantization weighting matrix used for the V component, and a header of each macroblock of the encoded input video stream is used for the Y component. And index information for specifying one quantization weighting matrix among a plurality of quantization weighting matrices, wherein the quantization weighting matrix storage selects one of the plurality of quantization weighting matrices for a Y component based on the index information. A video decoding apparatus. The method of claim 4, wherein The variable length decoder extracts index information of each quantization weight matrix from a macroblock header. The method of claim 5, And the index information is a difference between an index of a quantization weighting matrix corresponding to one macroblock and an index of a quantization weighting matrix corresponding to a macroblock adjacent to the macroblock.

Images