KR20140019012A - High definition video encoding/decoding method and apparatus - Google Patents

Abstract

The present invention relates to a method and an apparatus for encoding/decoding high-definition video. The present invention provides an image encoding method characterized by predicting a current block to generate a prediction block, subtracting the prediction block from the current block to generate a residual block, determining types of transform and quantization depending on the block type of the current block, transforming and quantizing the residual block according to the determined types of transform and quantization, and encoding the transformed and quantized residual block. According to the present invention, the encoding can be performed using high correlation of temporally/spatially neighboring pixels shown on an image, thereby improving the encoding efficiency. Further, the block distortion can be reduced, thereby improving the compression efficiency. Further, the number of times the filtering is performed can be decreased, thereby reducing the degree of complexity of the apparatus for encoding/decoding an image. [Reference numerals] (AA) Filtering when various types of transform and quantization are applied according to an embodiment of the present invention; (B1, B2, B5, B6, B18, B16, B17, B24, B27) 16 X 16 transform; (B10, B11, B23, B25, B26) 16 X 8 transform; (B12, B13, B14, B15) 4 X 8 transform; (B18, B20, B21, B22) 16 X 16 transform; (B28) 32 X 16 transform; (B3, B4, B9) 32 X 8 transform; (B7, B8) 8 X 16 transform

Description

High Definition Video Encoding / Decoding Method and Apparatus

The present invention relates to a high resolution image encoding / decoding method and apparatus. More specifically, the present invention relates to a method and apparatus for improving coding efficiency by performing encoding and decoding on a block basis of various types, and performing transformation, quantization, scanning, and filtering on a block type suitable for the same.

The Moving Picture Experts Group (MPEG) and the Video Coding Experts Group (VCEG) have developed superior video compression techniques that are superior to the existing MPEG-4 Part 2 and H.263 standards. This new standard was called H.264 / AVC (Advanced Video Coding) and jointly announced as MPEG-4 Part 10 AVC and ITU-T Recommendation H.264. In the H.264 / AVC (abbreviated as 'H.264') standard, intra prediction / inter prediction is performed in units of macroblocks having various types of subblocks. The residual signal may be generated, and the generated residual signal may be transformed, quantized, and encoded to further reduce the number of bits of the encoded data.

Referring to the encoding process of a conventional macroblock unit from the encoder side, the input image is divided into macroblock units, and each macroblock is divided into subblocks that may have an inter mode or an intra mode. A residual block is generated by performing prediction by magnitude, and the generated residual block is transformed by applying an integer transform designed based on a Discrete Cosine Transform (DCT) of 4x4 or 8x8 units. A transform coefficient is generated, and the transform coefficient is quantized according to a given quantization parameter (QP). Blocking caused by the transform process and the quantization process is reduced through loop filtering.

In a typical image compression technique such as H.264, encoding is performed by dividing an image to be encoded into 16x16 macroblock units, and a unit of transformation is also fixed to 4x4 block size and 8x8 block size, so that the image having high correlation between pixels In this case, there is a problem that the coding efficiency is lowered. That is, when there is a high correlation between pixels, it may be more efficient to predict macroblock units or macroblock units of various types than macroblocks having a 16x16 block size. Although it may be efficient to predict the shape, the conventional video compression technique has a problem in that the encoding efficiency is deteriorated because the size of the macroblock and the size of the transform are fixed, so that adaptive encoding is impossible according to the characteristics of the image.

In order to solve the above problems, the present invention performs encoding in units of macroblocks of various types suitable for high resolution images, and improves compression efficiency by performing various types of prediction, transformation and quantization, scanning, and filtering corresponding thereto. Has a main purpose.

In order to achieve the above object, the present invention provides a method of encoding an image, the method comprising: generating a prediction block by predicting a current block; Generating a residual block by subtracting a current block from a prediction block; Determining a transform and quantization type according to the block type of the current block, and transforming and quantizing the residual block according to the determined transform and quantization type; And encoding the transformed and quantized residual block.

According to another object of the present invention, there is provided an apparatus for encoding an image, comprising: a predictor for predicting a current block to generate a predicted block; A subtractor for subtracting a current block from a prediction block to generate a residual block; A converter and quantizer for determining a transform and quantization type according to the block type of the current block, and for transforming and quantizing a residual block according to the determined transform and quantization type; And an encoder for encoding the transformed and quantized residual blocks.

According to another object of the present invention, there is provided a method of decoding an image, the method comprising: restoring a transformed and quantized residual block by decoding encoded data; Inverse quantization and inverse transformation on the transformed and quantized residual blocks according to the transform and quantization type to recover the residual block; Predicting a current block to generate a predicted block; And restoring a current block by adding a residual block and a prediction block to be reconstructed.

According to still another object of the present invention, there is provided an apparatus for decoding an image, comprising: a decoder configured to decode encoded data to restore a transformed and quantized residual block; An inverse quantizer and an inverse transformer for reconstructing the residual block by inverse quantization and inverse transformation on the transformed and quantized residual block according to the transform and quantization type; A predictor for predicting a current block to generate a predicted block; And an adder for adding a residual block to be reconstructed and a prediction block to reconstruct a current block.

As described above, according to the present invention, encoding can be performed using a high correlation between temporally and spatially adjacent pixels appearing in an image, thereby increasing encoding efficiency and reducing block distortion to improve compression efficiency. In addition, since the number of times of filtering may be reduced, the complexity of implementing the apparatus for encoding and decoding an image may be reduced.

1 is a block diagram schematically illustrating a configuration of an image encoding apparatus according to an embodiment of the present invention;
2 to 4 are exemplary diagrams illustrating an intra prediction mode according to a macroblock type used in conventional video encoding.
5 is an exemplary diagram illustrating an inter prediction mode according to a macroblock type used in conventional video encoding;
6 is an exemplary view showing a macroblock of an MxN block size according to an embodiment of the present invention;
FIG. 7 is an exemplary diagram illustrating various types of sub macroblocks that a macroblock of an MxN block size may have according to an embodiment of the present invention; FIG.
8 is an exemplary diagram illustrating MF for 8x4 transform according to an embodiment of the present invention;
9 is an exemplary diagram for describing filtering when various transform and quantization types are applied according to an embodiment of the present invention;
10 is a diagram illustrating a process of deblocking filtering on a block boundary according to an embodiment of the present invention.
11 is an exemplary diagram for explaining a process of performing de-ring filtering according to an embodiment of the present invention;
12 is an exemplary view showing a scanning order according to a transform and quantization type according to an embodiment of the present invention;
13 to 18 are exemplary diagrams for explaining a CAVLC application method according to a transform and quantization type according to an embodiment of the present invention;
FIG. 19 is a flowchart illustrating an image encoding method according to an embodiment of the present invention. FIG.
20 is a block diagram schematically illustrating an image decoding apparatus according to an embodiment of the present invention;
21 is a flowchart illustrating an image decoding method according to an embodiment of the present invention.

Hereinafter, some embodiments of the present invention will be described in detail with reference to exemplary drawings. It should be noted that, in adding reference numerals to the constituent elements of the drawings, the same constituent elements are denoted by the same reference symbols as possible even if they are shown in different drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

In addition, in describing the component of this invention, terms, such as 1st, 2nd, A, B, (a), (b), can be used. These terms are intended to distinguish the constituent elements from other constituent elements, and the terms do not limit the nature, order or order of the constituent elements. When a component is described as being "connected", "coupled", or "connected" to another component, the component may be directly connected or connected to the other component, Quot; may be "connected," "coupled," or "connected. &Quot;

1 is a block diagram schematically illustrating a configuration of an image encoding apparatus according to an embodiment of the present invention.

The image encoding apparatus 100 according to an embodiment of the present invention is an apparatus for encoding an image, and includes a predictor 110, a subtracter 120, a transformer and quantizer 130, and a scanner. The scanner 140 may include an encoder 150, an inverse quantizer and a transformer 160, an adder 170, and a filter 180. The video encoding apparatus 100 may be a personal computer (PC), a TV, a notebook computer, a personal digital assistant (PDA), a portable multimedia player (PMP), or a PlayStation Portable (PSP). PlayStation Portable), a mobile communication terminal, and the like, a communication device such as a communication modem for communicating with various devices or a wired / wireless communication network, a memory for storing various programs and data for encoding an image, Means a variety of devices including a microprocessor for executing and operating a program.

The input image to be encoded may be input in units of blocks, and the block may be a macroblock. In an embodiment of the present invention, the shape of the macroblock may be M × N. Here, M and N may be an integer having a value of 2 ⁿ (where n is an integer of 1 or more), in particular M and N may be an integer greater than or equal to 16. In addition, a different type of block may be used for each frame to be encoded, and information about a block type, which is information about the frame type, may be encoded for each frame to determine a form of a block of a frame to be decoded when the image decoding apparatus decodes the encoded data. You can decide. Determination of which type of block to use can be made by selecting the type of block that has the best efficiency by encoding the current frame into various types of blocks, or by selecting the type of block according to the analyzed characteristics by analyzing the characteristics of the frame. have. For example, if the image of the frame has a high correlation in the horizontal direction, a block having a long shape in the horizontal direction may be selected, and if the image has a high correlation in the vertical direction, a block having a long shape in the vertical direction may be selected.

To this end, the image encoding apparatus 100 may further include a block type determiner (not shown) that determines the block type and encodes information about the block type in the encoded data.

The predictor 110 generates a predicted block by predicting a block to be currently encoded (hereinafter, referred to as a current block) in the input image. That is, the predictor 110 predicts a current block from an input image by using intra prediction or inter prediction, and has a predicted pixel value as a pixel value of each pixel. Create

In this case, in order to optimize the predicted pixel value, the block may be divided into smaller blocks and predicted as necessary. That is, the prediction block may be generated in units of subblocks divided from the blocks. Herein, the block may be an MxN block having a square or rectangular shape as described above, and the subblock may have a QxP shape having a size of 2 ⁿ horizontally and vertically within a range not exceeding the size of the block (or macroblock). It may be a block of.

The subtractor 120 subtracts the prediction block from the current block to generate a residual block. That is, the subtractor 120 calculates a difference value between an original pixel value of each pixel of the current block and a predicted pixel value of each pixel of the prediction block to generate a residual block having a residual signal. .

The transformer and quantizer 130 determines a transform and quantization type according to the block type of the current block, and transforms and quantizes the residual block according to the determined transform and quantization type. That is, the transformer and quantizer 130 transforms the residual signal of the residual block into a frequency domain to generate and quantize a residual block having a transform coefficient, and transform and quantize the quantized transform coefficient. Create a residual and transformed block.

When the transformer and the quantizer 130 transform and quantize the residual block, the transformation is completed only when the quantization is completed because the transformation process is included in the quantization process. In this case, the video signal in the spatial domain, such as Hadamard Transform, Discrete Cosine Transform Based Integer Transform (hereinafter abbreviated as 'Integer Transform'), is used for the frequency domain. A quantization scheme may be used, and various quantization techniques such as Dead Zone Uniform Threshold Quantization (DZUTQ) or Quantization Weighted Matrix (DZUTQ) may be used. Can be used.

In addition, the transform and quantization type may be variously modified such as transform and quantization of the PxQ block size within a range not exceeding the size of the current block. Here, the transform and quantization of the PxQ block size may be a transform and quantization of a subblock size that a current block having an MxN block size may have in addition to the transform and quantization of a 4x4 block size and the transform and quantization of an 8x8 block size. .

To this end, the converter and quantizer 130 may transform and quantize the residual block by determining a transform and quantization type according to the block type of the current block. In this case, when the block type of the current block is an intra block type, the transform and quantization type may be determined to be the same as the block size of the intra block type, and when the block type of the current block is an inter block type, a plurality of encoding may be performed using an encoding cost. One of the transform and quantization types may be determined. Here, the plurality of transform and quantization types may be the same transform and quantization type as the block size of the subblock, but may also be a transform and quantization type having other various block sizes. A process of transforming and quantizing the residual block by the transform and quantizer 130 will be described in detail later.

Here, the current block refers to a block to be currently encoded by the video encoding apparatus 100. A block type is a type of a block specifying a method of encoding a block and a block size, and may be classified into an intra block type and an inter block type. The intra block type may be further divided into an intra 4x4 block type, an intra 8x8 block type, an intra 16x16 block type, and may be further divided into various intra block types such as an intra 32x32 block type and an intra 16x32 block type. The inter block type can be divided into inter 4x4 block type, inter 8x4 block type, inter 4x8 block type, inter 8x8 block type, inter 8x16 block type, inter 16x8 block type, inter 16x16 block type, and the like. It can be divided into various inter block types such as block type, inter 16x32 block type, inter 32x32 block type.

The scanner 140 scans the quantized transform coefficients of the residual block transformed and quantized by the transformer and quantizer 130 to generate a quantized transform coefficient sequence. In this case, the scanning method considers characteristics of a transform technique, a quantization technique, and a block (macroblock or subblock), and the scanning order is determined so that the scanned quantization transform coefficient sequence has a minimum length. In FIG. 1, the scanner 140 is illustrated and described as being implemented independently of the encoder 150, but the scanner 140 may be omitted, and its function may be integrated into the encoder 150.

The encoder 150 encodes the transformed and quantized residual block to generate encoded data. That is, the encoder 150 generates encoded data by encoding the quantized transform coefficient string generated by scanning the quantized transform coefficients of the residual block transformed and quantized by the transformer and the quantizer 130, or scans by the scanner 140. The encoded quantized transform coefficient sequence is generated to generate encoded data.

As the encoding technique, an entropy encoding technique may be used, but various encoding techniques may be used without being limited thereto. In addition, the encoder 150 may include not only a bit string encoding the quantized transform coefficient sequence but also various pieces of information necessary to decode the encoded bit sequence in the encoded data. Here, the various pieces of information necessary to decode the encoded bit strings include information about a block type, information about an intra prediction mode when the prediction mode is an intra prediction mode, and information about a motion vector when the prediction mode is an inter prediction mode. Information, information on transform and quantization types, and the like, but also various other information.

The inverse quantizer and inverse transformer 160 inverse quantizes and inverse transforms the residual block transformed and quantized by the transformer and quantizer 130 to reconstruct the residual block. Inverse quantization and inverse transformation may be performed by inversely performing a transform process and a quantization process performed by the transformer and quantizer 130. That is, inverse quantizer and inverse transformer 160 use transform and quantization information transmitted from transformer and quantizer 140 (eg, information about the type of transform and quantization) to transform and quantizer 130. Inverse transformation and quantization may be performed inversely to perform inverse quantization and inverse transformation.

The adder 170 reconstructs the current block by adding the prediction block predicted by the predictor 110 and the residual block inversely quantized and inversely transformed by the inverse quantization and inverse transformer 170.

The filter 180 filters the current block restored by the adder 170. In this case, the filter 180 is a ringing effect occurring at a block boundary due to block-by-block transformation and quantization of an image, and ringing noise generated around an edge of the image due to high frequency loss during quantization. Reduce noise. Here, a deblocking filter and a deringing filter may be used to reduce the blocking phenomenon and the ringing noise, respectively. Among the two methods, a deblocking filter and a deringing filter may be used. You can optionally choose to use only one filter, or not both. In addition, the deblocking filtering may selectively use one of applying a filter to a boundary between subblocks and a macroblock or applying a filter only to a boundary between macroblocks.

Meanwhile, in a typical image encoding, a macroblock used for image encoding is 16x16 pixels having the same width and height, and one or more of intra prediction and inter prediction may be performed on each macroblock to generate a prediction block. have. Encoding an image in units of macroblocks is an encoding method that is widely used because efficient encoding considering local characteristics of an image is possible. In addition, since various methods of intra prediction or inter prediction are used to generate a prediction block, high coding efficiency is shown.

2 to 4 are exemplary diagrams illustrating an intra prediction mode according to a macroblock type used in conventional video encoding.

2 is an exemplary diagram illustrating nine intra prediction modes when the macroblock type is an intra 4x4 macroblock, and FIG. 3 is an exemplary diagram illustrating nine intra prediction modes when the macroblock type is an intra 8x8 macroblock. 4 is an exemplary diagram illustrating four intra prediction modes when the macroblock type is an intra 16x16 macroblock.

If the macroblock type is an intra block type, the macroblock to be encoded is predicted using intra prediction. Intra block types are further classified into intra 4x4 macroblocks, intra 8x8 macroblocks, intra 16x16 macroblocks, and the like. In each case, macroblocks are already encoded and decoded according to prediction modes as shown in FIGS. 2 to 4. And predicted using the neighboring pixels of the restored neighboring block.

5 is an exemplary diagram illustrating an inter prediction mode according to a macroblock type used in conventional image encoding.

If the macroblock type is the inter prediction type, the macroblock to be encoded is predicted using inter prediction. In this case, as shown in FIG. 3, the macroblock is predicted to a 16x16, 16x8, 8x16 or 8x8 block size using a frame that has been previously encoded, decoded and reconstructed to generate a prediction block. The macroblock is an 8x8 block. In the case of predicting the size, each 8x8 block is predicted to have an 8x8, 8x4, 4x8 or 4x4 block size to generate a predicted block.

However, if a high resolution image is encoded in a macroblock unit having a 16x16 block size as in conventional image encoding, efficient encoding using high correlation between pixels, which is a characteristic of a high resolution image, cannot be performed. Since the prediction accuracy of the prediction block generated in the macroblock unit of the extended MxN block size is similar to that of the prediction block generated in the macroblock of the normal 16x16 block size, the image is recorded in the macroblock unit of the 16x16 block size. This is because, in the case of encoding, the number of macroblocks to be encoded increases, thereby reducing the encoding efficiency.

In addition, conventional video encoding uses an integer transform of 4x4 block size or 8x8 block size based on discrete cosine transform. Integer transform has an advantage in terms of coding efficiency and complexity because it performs only integer unit operations without performing real unit operation, which is a disadvantage of discrete cosine transform, while maintaining the characteristics of the discrete cosine transform. Blocking phenomena and ringing noise caused by block-by-block conversion can be minimized by using filtering.

However, the image encoding apparatus 100 encodes a high resolution image using only transformation and quantization of 4x4 block size or 8x8 block size, which is conventional to encode a high resolution image by additionally using more various PxQ block sizes. Even more efficient. When encoding a large area in which high correlation pixels, which are characteristic of a high resolution image, are concentrated, a blockage phenomenon occurs when only a typical 4x4 block size transform and quantization or 8x8 block size transform and quantization are used. This is because a lot of loss occurs in the high frequency component.

On the other hand, if transform and quantization can be selectively performed using various types of transform and quantization of PxQ block sizes, the blockage phenomenon can be reduced and the loss in the high frequency component is higher than that of conventional block size conversion and quantization. It can also reduce ringing noise. Accordingly, the number of filtering operations may be reduced, thereby reducing the complexity for the filtering operation, which takes up a large part in terms of the complexity of the image encoding apparatus 100 and the image decoding apparatus to be described later. In addition, since various types of transform and quantization of PxQ block sizes are used, the scanner 140 scans the quantized transform coefficients generated by the transformer and the quantizer 130 in a manner suitable for the block size for transform and quantization. Encoding efficiency can be improved.

Therefore, in one embodiment of the present invention, unlike a conventional video encoding method of encoding an image with a macroblock of 16x16 block size, the residual block is generated, transformed, and predicted in units of macroblocks having an extended MxN block size. The block size for quantization is not only using a 4x4 block size or an 8x8 block size, but also performs transform and quantization according to various types of PxQ block sizes, and performs filtering and scanning suitable for the block size of transform and quantization.

6 is an exemplary diagram illustrating a macroblock of an M × N block size according to an embodiment of the present invention.

According to an embodiment of the present invention, as shown in FIG. 6, a macroblock of 32x32 block size, a macroblock of 32x16 block size, or the like, in addition to the macroblock of 64x64 block size, the macroblock of 128x128 block size, the 64x128 block size, etc. Images may be encoded in units of macroblocks of various block sizes such as macroblocks. As shown in FIG. 6, the macroblock of the M × N block size according to an embodiment of the present invention may have a rectangular shape as well as a square shape.

7 is an exemplary diagram illustrating various types of sub macroblocks that a macroblock of an M × N block size may have according to an embodiment of the present invention.

FIG. 7 exemplarily illustrates types of block sizes of sub macroblocks that a macroblock may have when the macroblock has a 32x16 block size. Prediction of macroblocks using such sub-macroblocks can generate prediction blocks that are more similar to the original macroblocks, thereby further enhancing coding efficiency. In this case, M and N, which determine the size of the macroblock of the MxN block size, may be determined as 2 ⁿ , and J and K, which determine the block size of the JxK sub-macroblock, do not exceed M and N within 2 ⁿ units. Has

On the other hand, the transformer and quantizer 130 transforms and quantizes the residual block according to the transform and quantization type, thereby converting and quantizing the residual signals of the residual block into transform coefficients to generate quantized transform coefficients. In this case, the converter and the quantizer 130 may differently determine the block size of the transform and quantization to be used for transform and quantization according to the block type of the block to be encoded.

For example, when the block type is an intra block type, a block size of transform and quantization may be determined as a block size for performing intra prediction. That is, 4x4 block size transform and quantization is used for intra 4x4 prediction, 8x8 block size transform and quantization is used for intra 8x8 prediction, 16x16 block size transform and quantization type is used for intra 16x16 prediction, and intra 16x8 prediction is used. If 16x8 block size transform and quantization can be used. Therefore, if the size of performing intra prediction is a PxQ block size, it may be determined as a transform and quantization type of the PxQ block size.

As another example, when the block type is an inter block type, it may be determined as a block size of transform and quantization having a minimum coding cost among a plurality of block sizes of transform and quantization. That is, 4x4 block size transform and quantization, 8x8 block size transform and quantization, 16x16 block size transform and quantization, 32x16 block size transform and quantization, 8x16 block size transform and quantization, 16x8 block size transform and quantization, etc. The transform and quantization of one blow size may be selected to transform and quantize the residual block using transform and quantization of the selected block size. In addition, when the block type is an inter block type, when the current block has a subblock (that is, when the current block is a macroblock and has a sub macroblock), a plurality of blocks are used for transforming and quantizing the size of the subblock. The transformation and quantization of a block size having the smallest coding cost by being included in the transformation and quantization of the size may be determined as the transformation and quantization type. Here, the encoding cost may be a rate-distortion cost, but is not necessarily limited thereto, and may be any cost as long as it is required to encode the residual block according to the transformation and quantization of the corresponding block size.

Hereinafter, assuming that the transform and quantization type is determined by 8x4 block size transform and quantization, a transform and quantization process according to an embodiment of the present invention will be described.

The 8x4 block size transform may be designed by combining an integer transform of 4x4 block size and an integer transform of 8x8 block size, and may be represented by Equation 1.

In Equation 1, X denotes a residual block of 8x4 block size generated by 8x4 block size prediction block, A denotes a matrix for integer conversion of 4x4 block size, and B denotes a matrix for integer conversion of 8x8 block size. T denotes a transpose matrix that is a matrix in which rows and columns of the corresponding matrix are changed, and Y denotes a transformed residual block that is a result of performing an 8x4 block size transform on a residual block X having an 8x4 block size.

In the above Equation, when A and B ^T represent elements, they can be expressed as Equation 2.

&Quot; (2) "

, B =

A =

, B =

, y는

, z는

이고, 행렬 B의 a는

, b는

, c는

, d는

, e는

, f는

, g는

이다. 여기서, 이산 코사인 변환의 특성인 직교성을 유지하면서 정수 연산을 수행하기 위해 각 4x4 크기의 정수 변환과 8x8 크기의 정수 변환은 수학식 3과 같이 분해되고 근사화된다.In equation (2), x of matrix A is

, y is

, z is

Where a in matrix B is

, b is

, c is

, d is

, e is

, f is

, g is

to be. Here, each 4x4 integer transform and 8x8 integer transform are decomposed and approximated as in Equation 3 to perform integer operations while maintaining orthogonality, which is a characteristic of discrete cosine transform.

&Quot; (3) "

A =

B =

y는 ,w(=

로 각각 근사화 된다. 행렬 B의, a는 그대로

이고, b는

, c는

, K(=

)는

, L(=

)는

, M(=

)는

로 각각 근사화 된다. 이러한 과정을 통하여 수학식 1은 다음의 수학식 4와 같이 나타낼 수 있다.In Equation 3, in the matrix A , x is as it is

y is , w (=

Each is approximated by In the matrix B , a remains the same

And b is

, c is

, K (=

)

, L (=

)

, M (=

)

Each is approximated by Through this process, Equation 1 may be expressed as Equation 4 below.

연산은

의 결과 행렬과 행렬 E의 같은 위치의 계수들끼리 곱해주는 연산을 나타낸다. 행렬 E는 수학식 1을 수학식 4로 근사화 및 분해하는 과정에서 유도된 8x8 행렬을 나타내며, 행렬의 엘레멘트는 수학식 4과 같다.In Equation 4, X and Y are the same as Equation 1, C represents the right 4x4 matrix in matrix A of Equation 3, and D ^T represents the left 8x8 matrix in matrix B of Equation 3. And

The operation

The operation of multiplying the coefficients of the same position of the matrix E and matrix E. The matrix E represents an 8x8 matrix derived in the process of approximating and decomposing Equation 1 into Equation 4, and the elements of the matrix are shown in Equation 4.

&Quot; (5) "

=

=

In Equation 5, x, y, a, b, c of the matrix E is equal to the value of Equation 3. On the other hand, in Equation 5 it can be seen that the coefficients of the matrix C , D ^T , E is not an integer, each matrix is scaled as shown in Equation 6 for the integer operation.

&Quot; (6) "

C =

D =

E =

When the scaling process of Equation 6 is completed, an 8x4 block sized integer transform is designed, and the final 8x4 block sized integer transform is designed by including the matrix E in the quantization process for integer conversion.

The basic quantization process may be represented as shown in Equation 7.

In Equation 7, Y _ij represents the elements represented by the matrix after transforming the residual block, and Qstep represents the size of the quantization step.

A basic quantization operation performed as shown in Equation 7 may be represented by Equations 8 and 9 to be applied to an actual quantization process for transform and quantization of an 8x4 block size.

이고, 인터 예측으로 예측된 경우에는

으로 고정된다. 여기서, qbits는

이며, floor는 내림(Round Down) 연산을 의미하는데, qbits는 변환 후의 변환 계수가 갖는 최대값 및 최소값에 따라 변경될 수 있다.In Equation 8, W _ij represents elements represented by a matrix after transforming each residual signal of the residual block, MF represents a multiplication factor determined according to a quantization parameter, and f represents quantization. This is a factor that determines the size of the dead zone and errors due to rounding in the process. When the current block is predicted by intra prediction,

If predicted by inter prediction

. Where qbits is

Floor is a round down operation. Qbits may be changed according to the maximum and minimum values of the transform coefficient after the conversion.

In this case, the matrix E in the equation (6) is included in the MF MF is equal to the equation (10).

In Equation 9, PF denotes the matrix E , and since PF changes according to the form and approximation of the transform, MF suitable for each transform must be obtained and used.

8 is an exemplary diagram illustrating MF for 8x4 transform according to an embodiment of the present invention.

Since the MF value changes according to the coefficient position of the matrix E of Equation 6, the first row of FIG. 8 represents the position of the coefficient according to the matrix E of Equation 6, and the first column of FIG. 8 means Qstep . Here, the MF value used is a mathematically derived value, but since the integer transform is not an optimal transform for image encoding, the MF value can be modified.

The above-described 8x4 block size transform and quantization can be applied according to various types of transform and quantization such as 4x8 block size transform and quantization, 16x8 block size transform and quantization, such as PxQ block size transform and quantization. Can be done.

9 is an exemplary diagram for describing filtering when various transform and quantization types are applied according to an embodiment of the present invention.

FIG. 9 illustrates a current block in which a residual block when the current block is a macroblock having a size of 32x32 blocks is transformed and quantized according to various types of transform and quantization types, inversely quantized and inverse transformed, and then added with a prediction block to be recovered. As shown.

In FIG. 9, a solid line indicates a boundary of a macroblock that is a current block or a sub macroblock of the current block. When filtering is performed on the boundary inside the area indicated by the solid circle, the filtering is performed only on the block boundary indicated by the dotted line. In this case, the filtering may be performed using deblocking filtering that reduces blocking. In this case, the deblocking filter used is a 1D low pass filter. Filtering is performed on block boundaries in the vertical direction and then on block boundaries in the horizontal direction.

In a typical video encoding method, since the transform is quantized using only an integer transform having a 4x4 block size or an integer transform having an 8x8 block size, the number of times that the filtering is performed because the boundaries of blocks to be deblocked and / or deringed are increased. Will increase. However, when the PxQ block size is transformed and quantized according to an embodiment of the present invention, since the number of times to perform filtering is reduced compared to a general video encoding method, the image encoding apparatus 100 and the image decoding apparatus for filtering The implementation complexity can be reduced, and less blocking occurs, thereby improving coding efficiency.

In addition, when the PxQ block size is transformed and quantized according to an embodiment of the present invention, the block size to be transformed and quantized may be increased, thereby increasing the number of pixels referred to for filtering. Can be more accurate to further reduce blocking or ringing.

10 is an exemplary diagram for describing a process of deblocking filtering on a block boundary according to an embodiment of the present invention.

In FIG. 10, a, b, c, d, e, and f represent pixels before deblocking filtering is performed on the block boundary, and b ', c', d ', and e' perform deblocking filtering on the block boundary. After the pixels are shown, the vertical position of the pixels indicates the brightness of the pixels. The solid line in the middle of the c and d pixels represents the block boundary. As can be seen from FIG. 10, before the deblocking filtering is performed on the block boundary, a blockage phenomenon occurs because the brightness difference of the pixel increases at the block boundary (a sharp increase in the brightness difference between pixels c and d).

In order to reduce this blocking phenomenon, deblocking filtering is performed at the block boundary to correct the brightness of pixels b, c, d, and e before the filtering is performed using the brightness of the surrounding pixels as shown in FIG. Pixels of ', c', d ', and e' may be generated.

In this case, a strong filter and a weak filter may be used as the one-dimensional low pass filter used for the deblocking filtering. The strong filter may be implemented as shown in Equation 11, and the weak filter may be implemented as shown in Equation 12.

In Equations 10 and 11, b, c, d, e, f, and d ' denote pixels in FIG. 8 and α denotes a constant for rounding.

As can be seen from equations (10) and (11), filtering with a strong filter affects the pixels that are filtered much from neighboring pixels, and filtering with a weak filter gives weight to the pixels that are filtered. Is less affected. By applying this concept, the weights applied to the strong and weak filters may be changed and filtered. The number of adjacent pixels and the number of pixels to be filtered may be selectively applied and filtered. Therefore, according to an embodiment of the present invention, since the number of adjacent pixels to be referred to when performing deblocking filtering on the block boundary by performing PxQ size conversion and quantization can be used more, the filtering effect can be further enhanced. .

11 is an exemplary diagram for describing a process of performing de-ring filtering according to an embodiment of the present invention.

In order to perform the de-ringing filter, an edge must be found in the reconstructed image, and an edge detection process such as a Sobel operation is performed to find the edge. 11 shows a block in which an edge is detected. In FIG. 11, black filled pixels represent edges detected in an image, A represents pixels to be filtered, and B and C represent adjacent pixels when filtered in the horizontal direction. To reduce ringing noise, de-ringing filtering is performed on the pixels of the block where the edge is detected. In this case, the filter used may be implemented as shown in Equation 13, and each pixel of the block where the edge is detected is filtered in the vertical direction and then filtered in the horizontal direction.

In Equation 13, A denotes a pixel to be filtered, B and C denote adjacent pixels when filtered in the horizontal direction, or adjacent pixels when filtered in the vertical direction, and A ' denotes a pixel on which the filtering is performed. Indicates. β, γ, and δ each represent weights that are differently applied depending on whether the B , A , and C pixels are edges, α represents a rounding constant, and λ represents a sum of β, γ, and δ .

The de-ringing filtering is performed on pixels that are not edges. When B or C is an edge, B and C are edges, and B and C are not edges, the weights assigned to the pixels may vary. . For example, if the B or C edging, B is the weight of the edge face C C increases the greater the weight of the side edges B. When B and C are edges, the weights of B and C are zero, and when B and C are not edges, the weight of A is the largest.

By applying the above concept, the weighting applied to each pixel may be changed to perform de-ringing filtering, and the number of adjacent pixels to be referred may be selectively used. Therefore, according to an embodiment of the present invention, since the number of adjacent pixels to be referred to when performing the de-ringing filtering on the block having an edge by performing the conversion and quantization of the PxQ block size, the filtering effect is further increased. It can increase.

12 is an exemplary diagram illustrating a scanning order according to a transform and quantization type according to an embodiment of the present invention.

According to an embodiment of the present invention, since the transform and quantization are performed according to various transform and quantization types, the scanner 140 or the encoder 150 scans the quantized transform coefficients of the transformed and quantized residual block according to the transform and quantization type and quantizes them. Generate the transformation coefficients. FIG. 12 exemplarily illustrates a procedure of scanning quantized transform coefficients when the transformed and quantized residual blocks have an 8x4 block size and a 4x8 block size.

The scanning sequence illustrated in FIG. 12 may be applied and used according to various transform and quantization types used according to an embodiment of the present invention. If the quantization transform coefficients are scanned according to a suitable scanning sequence, encoding efficiency may be further increased. . In the case of the discrete cosine transform, since the transform coefficient after the transformation is usually collected in the frequency portion having the low energy, the transform coefficient tends to gather at the upper left side of the block (to the DC coefficient side). Because it shows the same propensity. Therefore, it may be efficient to scan in the direction in which the energy of the transform coefficient decreases from the coefficient at a position close to the DC coefficient with the position of the DC coefficient as a starting point.

The encoder 150 may generate a bit string by encoding the quantized transform coefficient sequence generated by scanning according to the scanning sequence shown in FIG. 12 or a similar scanning sequence in various ways. However, the encoder 150 may apply existing context-based adaptive variable length coding (CAVLC) to scan and encode residual blocks that have been transformed and quantized. .

13 to 18 are diagrams for describing a CAVLC application method according to a transform and quantization type according to an embodiment of the present invention.

In typical image encoding, CAVLC is performed only for blocks having a 4x4 block size. However, in an embodiment of the present invention, CAVLC may be performed for blocks having a 4x4 block size or more as shown.

13 to 18, when the transformed and quantized residual blocks have sizes of 8x4, 4x8, 8x8, 16x8, 8x16, and 16x16, respectively, the transformed and quantized residual blocks are decomposed into blocks of 4x4 block size to apply CAVLC. An example is shown. 13 to 18, numbers 1, 2, 3, and 4 shown in each pixel are given to indicate the position of each pixel, and are collected in blocks having a size of 4x4 by collecting only pixels having the same number.

For example, in the 8x4 block sized transformed and quantized residual block shown in FIG. 13, the numbers 1 and 2 are assigned as partitions, and 4 columns with the number 1 are collected separately and 4 with the number 2 is assigned. Columns can be collected separately, resulting in two 4x4 block sized transformed and quantized residual blocks. The encoder 150 may obtain a bit string by encoding two 4 × 4 block sizes obtained as illustrated and transforming the quantized residual blocks using CAVLC.

As another example, in the 8x8 block size transformed and quantized residual block shown in FIG. 15, pixels numbered 1, pixels numbered 2, pixels numbered 3, numbers 4 Each of the given pixels may be separately collected to obtain four 4 × 4 block sized transformed and quantized residual blocks. The encoder 150 may obtain the bit string by encoding the transformed and quantized residual blocks having four 4x4 block sizes obtained using CAVLC as illustrated.

FIG. 19 is a flowchart illustrating an image encoding method according to an embodiment of the present invention.

According to an image encoding method according to an embodiment of the present invention, the image encoding apparatus 100 generates a prediction block by predicting a current block (S1910), and generates a residual block by subtracting the current block and the prediction block (S1920). A transform and quantization type are determined according to the block type of the current block (S1930), and the residual block is transformed and quantized according to the determined transform and quantization type (S1940). Herein, the current block is a macroblock having an MxN block size, and M and N may be larger than 16.

In operation S1910, the image encoding apparatus 100 may generate a prediction block by dividing the current block into a plurality of subblocks and combining the prediction subblocks predicting the divided plurality of subblocks. To this end, the image encoding apparatus 100 may determine a block type for each frame of an image. In this case, the current block may be a macroblock according to the determined block type. In determining the block type, the image encoding apparatus 100 may determine the block type by using the encoding cost required to encode the frame according to the plurality of block types, but may also determine the block type according to the characteristics of the frame. have. The characteristics of such a frame may be one or more of the horizontal and vertical correlations of the frame. In addition, the image encoding apparatus 100 may encode information about a block type and additionally include the encoded data.

In operation S1930, when the block type is an intra block type, the image encoding apparatus 100 may determine transform and quantization of a block size equal to the block type size as a transform and quantization type, and the block type is an inter block type. In case of, the transform cost and the quantization type of one block size among the transform and quantization of the plurality of block sizes may be determined as the transform and quantization types by using the encoding cost. Here, the transformation and quantization of the plurality of block sizes may include the same transformation and quantization as the block size of the subblock.

In operation S1940, the image encoding apparatus 100 scans the quantized transform coefficients of the transformed and quantized residual block in order of decreasing energy of the quantized transform coefficients from the quantized transform coefficients at positions close to the DC coefficients. The transform coefficient string may be encoded.

In addition, the image encoding apparatus 100 may reconstruct the residual block by inverse quantization and inverse transformation of the transformed and quantized residual block, and may reconstruct the current block by adding the reconstructed residual block and the prediction block, and reconstruct the current block. Can be filtered according to the transform and quantization type. In this case, the image encoding apparatus 100 filters the reconstructed current block, deblocking filtering the boundary of the reconstructed current block according to the transform and quantization type, and de-rings the reconstructed current block according to the transform and quantization type. You can filter. In this case, the image encoding apparatus 100 may perform deblocking filtering and / or deringing filtering differently according to a transform and quantization type. However, both the deblocking filtering and the de-ringing filtering need not be performed, and only one of the two filtering may be selectively performed or both may not be performed.

In addition, the image encoding apparatus 100 may encode information about the transform and quantization type determined in operation S1930 and additionally include the encoded data.

As described above, when the image encoding apparatus 100 and the image encoding method according to the embodiment of the present invention are used, a macroblock of MxN block size, a transform and quantization of a PxQ block size, and a transform and quantization type are suitable. Scanning and filtering can be performed to encode high resolution images more efficiently. As described above, the image encoded by the encoding data by the image encoding apparatus 100 may be connected to a wired or wireless communication network such as the Internet, a near field communication network, a wireless LAN network, a WiBro network, or a mobile communication network in real time or in real time, or may be connected to a cable or a universal serial bus. The image decoding apparatus may be transmitted to a video decoding apparatus to be described later through a communication interface such as a universal serial bus (USB), decoded by the video decoding apparatus, and restored and reproduced.

20 is a block diagram schematically illustrating an image decoding apparatus according to an embodiment of the present invention.

The image decoding apparatus 2000 according to an embodiment of the present invention may include a decoder 2010, an inverse scanner 2020, an inverse quantizer and an inverse converter 2030, a predictor 2040, an adder 2050, and a filter 2060. It may be configured to include). In this case, the inverse scanner 2020 and the filter 2060 are not necessarily included and may be selectively omitted depending on the implementation manner. When the inverse scanner 2020 is omitted, the function is integrated in the decoder 2010. Can be implemented.

The decoder 2010 decodes the encoded data to restore the transformed and quantized residual block. That is, the decoder 2010 decodes the encoded data to restore the quantization transform coefficient sequence. When the function of the scanner 140 is integrated with the encoder 150 in the image encoding apparatus 100, the image encoding apparatus 2000 may be implemented. Since the inverse scanner 2020 is omitted and its function is integrated in the decoder 2010 and implemented, the decoder 2010 may inversely scan the reconstructed quantized transform coefficient sequence to restore the transformed and quantized residual block. have.

Also, the decoder 2010 may decode and extract encoded data to decode or extract not only residual and transformed quantized blocks, but also information necessary for decoding. The information necessary for decoding refers to information necessary for decoding the coded bit string in the encoded data. For example, information about a block type, information about an intra prediction mode when the prediction mode is an intra prediction mode, and an inter prediction mode In the case of the prediction mode, the information may be information on a motion vector, information on a transform and quantization type, or the like.

Information about the block type may be passed to the inverse quantizer and inverse transformer 2030 and the predictor 2040, and information about the transform and quantization type may be passed to the inverse quantizer and inverse transformer 2030. Information necessary for prediction, such as information about a prediction mode and information about a motion vector, may be transmitted to the predictor 2040.

The inverse scanner 2020 restores and transfers the quantized transform coefficient sequence by the decoder 2010 to restore the transformed and quantized residual blocks by inversely scanning the quantized transform coefficient sequence. In addition, as described above, when the function of the scanner 140 is integrated with the encoder 150 in the image encoding apparatus 100, the inverse scanner 2020 is omitted in the image encoding apparatus 2000 so that the function is decoded. Can be incorporated in 2010. In addition, the decoder 2010 or the inverse scanner 2020 may transform and quantize the residual block according to the transform and quantization type identified by the information on the transform and quantization type, which is recovered by decoding the encoded data in the decoder 2010. Reverse scan. Here, the method of inverse scanning according to the transform and quantization type by the inverse scanner 2020 reverses the method in which the scanner 1040 described above with reference to FIGS. 1 and 12 scans the quantized transform coefficients of the transformed and quantized residual block. Since it is the same as or similar to that performed, a detailed description of the method of reverse scanning is omitted.

Inverse quantizer and inverse transformer 2030 inverse quantizes and inverse transforms the reconstructed transformed and quantized residual block to recover the residual block. In this case, the inverse quantizer and inverse transformer 2030 inversely quantizes and inverse transforms the residual blocks that are transformed and quantized according to the transform and quantization type identified by the information about the transform and quantization type transferred from the decoder 2010. . Here, the method of inverse quantization and inverse transformation by the inverse quantizer and inverse transformer 2030 according to the transform and quantization type of the transformed and quantized residual block may be performed by the transformer and quantizer 1030 of the image encoding apparatus 100. Since the process of transforming and quantizing according to the quantization type is the same as or similar to performing inverse, detailed description of the method of inverse quantization and inverse transformation is omitted.

The predictor 2040 generates a prediction block by predicting the current block. Here, the predictor 2040 predicts the current block by using information on the block type transmitted from the decoder 2010 and information necessary for prediction. That is, the predictor 2040 determines the size and shape of the current block according to the block type identified by the information on the block type, and uses the intra prediction mode or the motion vector identified by the information required for prediction to determine the current block. Predictive to generate predictive blocks. In this case, the predictor 2040 is the same or similar to the predictor 110 of the image encoding apparatus 100. The predictor 2040 divides the current block into subblocks and combines predicted subblocks generated by predicting the divided subblocks to predict the predicted block. Can be generated.

The adder 2050 reconstructs the current block by adding the residual block reconstructed by the inverse quantizer and the inverse transformer 2030 and the predictive block generated by the predictor 2040.

The filter 2060 filters the current block reconstructed by the adder 2050, and the reconstructed filtered current block is accumulated in units of pictures and stored in a memory (not shown) as a reference picture to be stored in the next block or in the predictor 2040. It is used to predict the next picture. Here, when filtering the current block to be reconstructed, the filter 2060 performs filtering according to the transform and quantization type identified by the information on the transform and quantization type transferred from the decoder 2010. In this case, the filter 2060 performs deblocking filtering on the boundary of the current block differently according to the transform and quantization type, or when an edge is detected in the current block, the filter 2060 performs de-ringing filtering on the current block differently according to the transform and quantization type. Blocking at the block boundary can be reduced or ringing around the edges in the block can be reduced. Since the filter 2060 performs the deblocking filtering and the de-ringing filtering, the filter 180 of the image encoding apparatus 100 is the same as or similar to the deblocking filtering and the de-ringing filtering. Omit the description.

21 is a flowchart illustrating an image decoding method according to an embodiment of the present invention.

According to an image decoding method according to an embodiment of the present invention, the image decoding apparatus 2000 decodes encoded data to reconstruct transformed and quantized residual blocks (S2110), and transforms and transforms the transformed and quantized residual blocks. In operation S2120, the residual block is reconstructed by inverse quantization and inverse transform, and a prediction block is generated by predicting the current block (S2130), and the current block is reconstructed by adding the reconstructed residual block and the prediction block (S2140).

The image decoding apparatus 2000 may further reconstruct the information on the transform and quantization types by decoding the encoded data. In this case, the image decoding apparatus 2000 may reconstruct the information on the transform and quantization types in step S2120. Inverse quantize and inverse transform the transformed and quantized residual blocks according to the identified transform and quantization type.

In addition, the image decoding apparatus 2000 may filter the current block reconstructed according to the transform and quantization type in operation S2140. De-ring filtering according to the transform and quantization type. Here, both deblocking filtering and deringing filtering may be performed, but only one may be performed or not all may be performed. When the image decoding apparatus 2000 performs deblocking filtering, the deblocking filtering may be performed differently according to the transform and quantization type. When the de-ringing filtering is performed, the image decoding apparatus 2000 may perform deringing filtering differently according to the transform and quantization type. . Meanwhile, the current block is a macroblock of M × N size, and M and N may be larger than 16.

In operation S2130, the image decoding apparatus 2000 may generate the prediction block by dividing the current block into a plurality of subblocks and combining the prediction subblocks predicting the divided plurality of subblocks. Can be.

In addition, the image decoding apparatus 2000 may further reconstruct the information on the block type for each frame of the image by decoding the encoded data. In this case, the current block is assigned to the block type identified by the information on the reconstructed block type. It can be a macroblock according to.

As described above, according to an embodiment of the present invention, the macroblock of the extended MxN block size and the corresponding PxQ block size are transformed and quantized, the PxQ block size is scanned, and appears in a high resolution image using filtering. Since an image can be encoded by appropriately using high correlation of temporally / spatially adjacent pixels, encoding efficiency can be improved. In addition, the use of extended block size macroblocks and transform and quantization can reduce block distortion, thereby increasing coding efficiency and reducing the distortion applied to the boundaries of transform and quantization blocks that can be performed during encoding / decoding. By reducing the number of companies performing blocking filtering, implementation complexity of the image encoding apparatus 100 and the image decoding apparatus 2000 may be reduced.

While the present invention has been described in connection with what is presently considered to be the most practical and preferred embodiment, it is to be understood that the invention is not limited to the disclosed embodiments. That is, within the scope of the present invention, all of the components may be selectively coupled to one or more of them. In addition, although all of the components may be implemented as one independent hardware, some or all of the components may be selectively combined to perform a part or all of the functions in one or a plurality of hardware. As shown in FIG. Codes and code segments constituting the computer program may be easily inferred by those skilled in the art. Such a computer program can be stored in a computer-readable storage medium, readable and executed by a computer, thereby realizing an embodiment of the present invention. As the storage medium of the computer program, a magnetic recording medium, an optical recording medium, a carrier wave medium, or the like may be included.

Furthermore, the terms "comprises", "comprising", or "having" described above mean that a component can be implanted unless otherwise specifically stated, But should be construed as including other elements. All terms, including technical and scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs, unless otherwise defined. Terms used generally, such as terms defined in a dictionary, should be interpreted to coincide with the contextual meaning of the related art, and shall not be interpreted in an ideal or excessively formal sense unless explicitly defined in the present invention.

The foregoing description is merely illustrative of the technical idea of the present invention, and various changes and modifications may be made by those skilled in the art without departing from the essential characteristics of the present invention. Therefore, the embodiments disclosed in the present invention are intended to illustrate rather than limit the scope of the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be interpreted by the following claims, and all technical ideas within the scope equivalent thereto should be construed as being included in the scope of the present invention.

As described above, the present invention is applied to a field of image compression processing for encoding and decoding a high resolution video, thereby enabling encoding using high correlation between temporally and spatially adjacent pixels appearing in the image, thereby increasing encoding efficiency. In addition, the compression efficiency can be improved by reducing the block distortion, and the number of times of performing filtering can be reduced, thereby generating an effect of reducing the complexity of implementing an apparatus for encoding and decoding an image.

Claims (19)

In the method of decoding an image,
Identifying a size (M × N) and a prediction mode of the current block from block information included in the encoded bit stream;
Restoring a prediction block of the current block according to the prediction mode;
Reconstructing transform subblock partition information of the current block from the encoded bit stream, and in units of sizes (P × Q, where P and Q are each specified within a range not exceeding M and N) for each transform subblock. Restoring the residual block by performing an inverse transform;
Reconstructing the current block by combining the prediction block and the residual block; And
Performing deblocking filtering on a block boundary in the reconstructed image including the current block
And decoding the decoded image. The method of claim 1,
In the deblocking filtering, two filtering methods having different filtering intensities are selectively performed, and the number of pixels to be filtered is different according to the selected filtering intensities. The method of claim 1,
The deblocking filtering is performed on a boundary in a vertical direction among the block boundaries, and then on a boundary in a horizontal direction. The method of claim 1,
And detecting an edge using three or more pixels adjacent to each other in the deblocking filtered reconstructed image, and adjusting an at least one pixel value among the three or more pixels. 5. The method of claim 4,
In the edge filtering step, the pixel value of the pixel located at the center of the three or more pixels is adjusted. 5. The method of claim 4,
The direction in which the three or more pixels are arranged includes a horizontal direction and a vertical direction. An apparatus for decoding a block-based coded image,
A decoder for identifying a size (M × N) and a prediction mode of a current block from block information included in an encoded bit string, and reconstructing transform subblock partition information of the current block from the encoded bit string;
A predictor for reconstructing the prediction block of the current block according to the prediction mode;
Reconstructing transform subblock partition information of the current block from the encoded bit stream, and in units of sizes (P × Q, where P and Q are each specified within a range not exceeding M and N) for each transform subblock. An inverse transformer for restoring a residual block by performing an inverse transform;
An adder for combining the prediction block and the residual block to reconstruct the current block; And
A first filtering unit that performs deblocking filtering on a block boundary in the reconstructed image including the current block.
And an image decoding unit for decoding the image. The method of claim 7, wherein
The deblocking filtering is performed by selectively performing two filtering methods having different filtering intensities, and the number of pixels to be filtered is different according to the selected filtering intensities. The method of claim 7, wherein
The deblocking filtering is performed on a boundary in a vertical direction among the block boundaries, and then on a boundary in a horizontal direction. The method of claim 7, wherein
And a second filtering unit configured to detect an edge by using three or more pixels adjacent to each other in the deblocking filtered reconstructed image, and to adjust at least one pixel value among the three or more pixels. . 11. The method of claim 10,
And the second filtering unit adjusts a pixel value of a pixel located at the center of the three or more pixels. The method of claim 10, wherein the second filtering unit,
A mode for detecting an edge using the three or more pixels adjacent to each other in a horizontal direction;
And a mode for detecting an edge using the three or more pixels adjacent to each other in the vertical direction. 11. The method of claim 10,
And the second filtering unit determines whether to perform the edge filtering according to a setting. An apparatus for encoding an image, the apparatus comprising:
A predictor for predicting a current block having an M × N size to generate a predicted block;
A subtractor for generating a residual block by subtracting the current block and the prediction block;
Converting the residual block in units of a size (P × Q, where P and Q are specified in a range not exceeding M and N, respectively) for each transform subblock of the current block, and quantizing the transformed residual block Transducers and quantizers;
An encoder for encoding information for identifying the size of the current block, a prediction mode, transform subblock partition information of the current block, and the quantized residual block;
An inverse quantizer and an inverse transformer for inverse quantizing the quantized residual block, and then inversely transforming the quantized residual block by a predetermined size unit for each transform subblock; And
A first filtering unit that performs deblocking filtering on a block boundary in the image including the reconstructed residual block.
Image encoding apparatus comprising a. 15. The method of claim 14,
The deblocking filtering is performed by selectively performing two filtering methods having different filtering intensities, and the number of pixels to be filtered is different according to the selected filtering intensities. 15. The method of claim 14,
The deblocking filtering is performed on a boundary in a vertical direction among the boundaries of the transform subblock, and then on a boundary in a horizontal direction. 15. The method of claim 14,
And a second filtering unit which detects an edge by using three or more pixels adjacent to each other in the deblocking filtered reconstructed image, and adjusts at least one pixel value among the three or more pixels. . 18. The method of claim 17,
And the second filtering unit adjusts a pixel value of a pixel located at the center of the three or more pixels. The method of claim 17, wherein the second filtering unit,
A mode for detecting an edge using the three or more pixels adjacent to each other in a horizontal direction;
And a mode for detecting an edge by using the three or more pixels adjacent to each other in a vertical direction.

Images