Prediction block generating device

ES 2 634 819 T3

DESCRIPTION

Apparatus for encoding an image

Technical field

The present invention relates to an apparatus for encoding an image.

Background technique

In image compression procedures such as Moving Picture Expert Group (MPEG) -1, MPEG-2, MPEG-4, and H.264 / MPEG-4 Advanced Video Coding (AVC), an image is divide into macro blocks to encode an image. The respective MBs are then encoded using an inter prediction or an intra prediction.

In intra prediction, a current block of a current image is encoded not using a reference image, but using pixel values ​​spatially adjacent to the current block. A low distortion intra-prediction mode is selected by comparing with a generating prediction block using the adjacent pixel values ​​with the original MB. Next, using the selected intra-prediction mode and adjacent pixel values, the prediction values ​​of the current block are calculated. Differences between the prediction values ​​and the pixel values ​​of the original current block are calculated and then encoded via transform encoding, quantization, and entropy encoding. The intra prediction mode is encoded as well.

Intra-prediction modes are generally classified into a 4x4 intra-prediction mode, an 8x8 intra-prediction mode, and a 16x16 intra-prediction mode for luminance components and chrominance components.

In the 16x16 intra-prediction mode according to the prior art, there are four modes, a vertical mode, a horizontal mode, a continuous current (CC) mode, and a flat mode.

In the 4x4 intra prediction mode according to the prior art there are nine modes, a vertical mode, a horizontal mode, a CC mode, a diagonal bottom-left mode, a diagonal bottom-right mode, a vertical right mode, a vertical left mode, a horizontal-up mode and a horizontal-bottom mode.

Each prediction mode is indexed according to the frequency of use of the respective modes. The vertical mode, whose mode number is 0, shows the highest possibility that it is used most frequently to perform intra prediction on a target block, and the horizontal-up mode whose mode number is 8 shows the possibility highest that is used the most infrequently.

According to H.264 standards, a current block is encoded using a total of 13 modes, that is, 4 modes of the 4x4 intra-prediction mode and 9 modes of the 16x16 intra-prediction mode. A bit stream of the current block is generated according to an optimal mode between these modes.

However, when some or all of the values ​​of the pixels adjacent to the current block do not exist or are not already encoded, it is impossible to apply some or all of the intra-prediction modes to the current block. Furthermore, when intra prediction is performed by selecting a prediction mode among the applicable intra mode, a residual signal between a prediction block and a current block becomes large. Therefore, the encoding efficiency degrades.

TUNG NGUYEN AND COL .: Improved Context Modeling for Coding Quantized Transform Coefficients in Video Compression, IMAGE CODIFICATION SYMPOSIUM 2010; NAGOYA, (08-12-2010), XP030082008, reveals the partitioning of blocks with transform coefficients greater than 4x4 into 4x4 sub-blocks in order to improve entropy coding. The sub-blocks are processed in a zigzag scan order and the coefficients of each sub-block are scanned in a reverse zigzag scan.

YEO CY COL .: Mode-Dependent Coefficient Scanning for Intra Prediction Residual Coding, 95. MPEG MEETING; DAEGU; No. m18796, (21-01-2011), XP030047366, discloses a mode-dependent scan order for intra-prediction residual encoding. For each intra mode and for each transform block size one of the four predetermined scan orders is assigned taking into account the prediction direction.

Divulgation

The present invention relates to an apparatus for encoding an image as specified in the appended claims.

Description of the drawings

FIG. 1 is a block diagram illustrating a moving picture coding apparatus in accordance with the present invention.

ES 2 634 819 T3

Figure 2 is a flow chart illustrating an operation of a scanning unit in accordance with the present invention.

Fig. 3 is a block diagram illustrating a moving picture decoding apparatus in accordance with the present invention.

Figure 4 is a block diagram illustrating an intra prediction unit in accordance with the present invention.

Figure 5 is a conceptual diagram showing the positions of the reference pixels used for intra prediction in accordance with the present invention.

FIG. 6 is a flow chart illustrating a reference pixel generation procedure in accordance with the present invention.

Fig. 7 is a block diagram illustrating an intra-prediction unit of a moving picture decoding apparatus in accordance with the present invention.

Invention mode

Hereinafter, the various embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the present invention is not limited to the exemplary embodiments disclosed below, but by the technical characteristics defined in the appended claim 1.

For image coding, each image consists of a plurality of segments, and each segment is made up of a plurality of coding units. Since an image of a degree of high definition (HD) or higher has many uniform regions, an image compression can be improved by encoding the image with encoding units larger than one MB whose size is 16x16.

A size of the encoding unit according to the present invention can be 16x16, 32x32 or 64x64. An encoding unit size can also be 8x8 or less. An encoding unit of the largest size is known as a super macro block (SMB). An SMB size is indicated by a smaller encoding unit size and depth information. The depth information indicates a difference value between the size of the SMB and the smallest size of the encoding unit.

Therefore, the encoding units that will be used to encode the images can be an SMB or a sub-block of an SMB. The encoding units are set to the default value or indicated in a sequence header.

An SMB consists of one or more encoding units. The SMB is in the form of a recursive coding tree in order to include the coding units and a division structure of the coding units. When the SMB is not divided into four sub-coding units, the coding tree may consist of information indicating that the SMB is not divided and a coding unit. When the SMB is divided into four sub-encoding units, the encoding tree may consist of information indicating that the SMB is divided and four sub-encoding trees. Likewise, each subcoding tree has the same structure as the SMB. However, a coding unit of the size of the smallest coding unit (SCU) is not divided into sub-coding units.

Meanwhile, each coding unit in the coding tree undergoes an intra-prediction or an inter-prediction in units of the coding unit itself or a sub-block. A unit in which intraprediction or interprediction is performed is called a prediction unit. A prediction unit size can be 2Nx2N or NxN for intra prediction. A prediction unit size can be 2Nx2N, 2NxN, Nx2N or NxN for inter prediction. In this document, 2N indicates the horizontal and vertical lengths of a coding unit.

A coding unit includes a prediction mode of the prediction unit and the size information (part_mode) in the prediction unit. To improve coding efficiency, prediction mode and size information can be combined and coded together. In this case, each encoding unit includes a co-encoded prediction type (web_type).

An encoding unit includes one or more containers for additional information. Each additional information container contains the additional information necessary to generate a prediction block for each prediction unit. In intra-prediction, the additional information includes the coded intra-prediction information. In inter prediction, the additional information includes the encoded motion information. The motion information includes a motion vector and a reference image index.

A coding unit also includes a residual signal container for residual signals from the coding unit. The residual signal container contains a transform tree, a luminance residual signal container, and two chrominance residual signal containers. The transform tree indicates whether or not the residual signals of the transform units exist in the residual signal container. The residual signal container consists of a recursive tree structure. The residual signal container for the coding unit is an example. If the encoding unit is not divided into four sub-encoding units, the

ES 2 634 819 T3 residual signal container contains quantization information (a residual quantization parameter) and an encoded residual signal. If the coding unit is divided into four sub-coding units, the residual signal container contains the quantization information and four residual signal sub-containers. Each residual signal sub-container has the same structure as the residual signal container of the coding unit, but does not contain the quantization information.

Meanwhile, a single case is described in which the coding unit is equally divided into prediction units. However, when the above-described equal division is used for an image that has a boundary in a specific direction or a specific position according to a characteristic, different prediction units are used for similar pieces of data in the boundary and a signal. residual cannot be effectively reduced.

In this case, to compress a residual signal, it may be more efficient to divide an SMB or an MB in a specific direction according to a shape of the image boundary and perform an intra or inter prediction.

The simplest adaptation mode is to divide a coding unit into two blocks using a straight line in order to extract the statistical dependence of a prediction region from the local topography. A boundary of an image corresponds to the straight line and is divided. In this case, the divisible addresses may be limited to a predetermined number. For example, a procedure of dividing a block may be limited to four directions from the horizontal, vertical, diagonal up, and diagonal down directions. Also, the division may be limited to only the horizontal and vertical directions. The number of divisible addresses can be three, five, seven, and so on. The number of divisible addresses can vary according to a size of the coding block. For example, for an encoding unit of a large size, the number of divisible addresses can be increased relatively.

In an inter prediction, when a coding unit is divided into two prediction units for one more adaptive prediction, motion estimation and motion compensation should be performed in each of the prediction units. The motion information is obtained for each prediction unit and a residual signal is encoded between a prediction block obtained from the motion information for each prediction unit.

After obtaining the residual signals for the respective two divided prediction blocks of a coding unit, the two residual signals can be added to generate a residual signal for a coding unit. The residual signal for a coding unit is transformed and encoded. In this case, there is a high possibility that there is a difference between the global distributions of the residual signals of the two respective prediction blocks with the center of the boundary, and therefore a residual signal of one coding unit can be generated by multiplying a value from any region by a default value. Furthermore, the boundary region of the two residual signals can be made to overlap, and smoothing can be performed in the overlapping boundary region to generate a residual signal.

In another method, a block can be generated by padding according to the respective cleavage regions of the block, and encoded. In other words, when a current split region is encoded between the two split regions, a block can be configured by filling another split region that constitutes the block with a value from the current split region and then undergoes transform encoding. two-dimensional (2D).

FIG. 1 is a block diagram illustrating a moving picture coding apparatus in accordance with the present invention.

Referring to FIG. 1, a moving image coding apparatus 100 according to the present invention includes an image division unit 110, a transform unit 120, a quantization unit 130, a scan unit 131, a entropy encoding unit 140, an intra-prediction unit 150, an inter-prediction unit 160, an inverse quantization unit 135, an inverse transform unit 125, a post-processing unit 170, an image storage unit 180, a subtractor 190, and an adder 195.

Image splitting unit 110 analyzes an input video signal to divide each largest coding unit (LCU) of an image into coding units each of which has a predetermined size, determines the prediction mode of each unit encoding unit, and determines the size of the prediction unit for each coding unit. The image division unit 110 sends the prediction unit to be encoded to the intra-prediction unit 150 or to the inter-prediction unit 160 according to the prediction mode. Furthermore, the image division unit 110 sends the prediction units to be encoded to the subtractor 190.

The transform unit 120 transforms a residual block. The residual block consists of a residual signal between an input original block and a prediction block generated by the intra-prediction unit 150 or the inter-prediction unit 160. The residual block may consist of a coding unit. The residual block consisting of a coding unit is divided into optimal transform units and transformed. One type

Transform matrix ES 2 634 819 T3 can be adaptively determined according to the prediction mode (intra or inter). Furthermore, an intra-prediction residual signal has directivity according to an intra-prediction mode, and therefore, a transform matrix can be adaptively determined according to the intra-prediction mode. The transform unit can be transformed by two one-dimensional (horizontal and vertical) transform matrices (1D). In inter prediction, a predetermined transform matrix type is determined. In intra-prediction, there is a high possibility that the residual block has vertical directivity when the intra-prediction mode is horizontal. Therefore, an integer matrix based on the discrete cosine transform (DCT) is applied to the vertical direction, and a discrete sine transform (DST) or an integer matrix based on the Karhunen Loeve transform (KLT) is applied ) to the horizontal direction. When the intra-prediction mode is vertical, a DST or KLT-based integer matrix is ​​applied to the vertical direction, and a DCT-based integer matrix is ​​applied to the horizontal direction. Furthermore, in intra prediction, the transform matrix can be adaptively determined according to a size of the transform units.

Quantization unit 130 determines a quantization step size size for quantizing the transformed residual block coefficients. The quantization step size is determined by encoding unit of a predetermined size or more. The default size can be 8x8 or 16x16. Using the determined quantization step size and a quantization matrix determined by a prediction mode, the coefficients of the transform block are quantized. The quantization unit 130 uses the quantization step sizes of the coding units adjacent to a current coding unit as a quantization step size predictor of the current coding unit. The quantization unit 130 sequentially retrieves the coding units in the following scan order; a left coding unit of the current coding unit, an upper coding unit of the current coding unit, and an upper left coding unit of the current coding unit. Next, quantization unit 130 generates the quantization step size predictor of the current coding unit using one or two valid quantization step sizes. For example, the first valid quantization step size discovered in the scan order can be determined as the quantization step size predictor. An average of two valid quantization step sizes retrieved in the scan order can be determined as the predictor of the quantization step size and determined as the predictor of the quantization step size when only one quantization step size is valid. When the quantization step size predictor is determined, a difference between the quantization step size and the quantization step size predictor is transmitted to the entropy coding unit 140.

When a segment is divided into coding units, there can be none of a left coding unit, an upper coding unit, and an upper left coding unit of the current coding unit. But, there may be a previous coding unit of the current coding unit in the coding order in the maximum coding unit. Therefore, the coding units adjacent to the current coding unit and the previous coding unit of the current coding unit in the coding order in the maximum coding unit may be candidates. In this case, the previous scan order can be changed to the next scan order; 1) the left coding unit of the current coding unit, 2) the upper coding unit of the current coding unit, 3) the upper left coding unit of the current coding unit, and 4) the previous coding unit of the current encoding unit. The scan order can be changed, or the upper left coding unit can be omitted from the scan order.

The quantized transform block is provided to inverse quantization unit 135 and scan unit 131.

The scan unit 131 scans the coefficients of the quantized transform block, thereby converting the coefficients to 1D quantized coefficients. Since the distribution of the transform block coefficients after quantization may depend on an intra-prediction mode, a coefficient scan pattern is determined according to the intra-prediction mode. The coefficient scan pattern can also be determined according to the size of the transform unit.

The inverse quantization unit 135 inverse quantizes the quantized coefficients. The inverse transform unit 125 restores a residual block of the spatial domain from the inversely quantized transform coefficients. The adder generates a reconstructed block by adding the residual block reconstructed by the inverse transform unit 125 and the prediction block from the intra prediction unit 150 or the inter prediction unit 160.

The post-processing unit 170 performs an unblocking filtering procedure to remove the blocking disturbance generated in a reconstructed image, an adaptive compensation application procedure to complement a difference between the reconstructed image and the original image per pixel, and a procedure of adaptive loop filter to supplement a difference between the reconstructed image and the original image in one encoding unit.

ES 2 634 819 T3

The unblocking filtering procedure can be applied to a boundary between prediction units having a predetermined size or more and a boundary between transform units. The default size can be 8x8. The unblocking filtering procedure includes a step of determining a limit to be filtered, a step of determining the limit filtering resistance to apply to the limit, a step of determining whether or not an unblocking filter is applied, and a step of selecting a filter to apply to the limit when it is determined that the unblocking filter is applied.

Applying the unblocking filter or not is determined according to i) if the limit filtering resistance is greater than 0 or not and ii) if a value that indicates the difference between the limit pixels of block P and block Q is smaller or not than a first reference value determined according to a quantization parameter.

There can be two or more filters. When an absolute value of a difference between two pixels adjacent to the block boundary is equal to or greater than a second reference value, a weak filter is selected. The second reference value is determined by the quantization parameter and the limit filtering resistance.

The adaptive compensation application method is intended to reduce a difference (distortion) between a pixel subjected to the unblocking filter and the original pixel. It can be determined whether or not the adaptive compensation application procedure is performed according to images or segments. An image or segment can be divided into a plurality of compensation regions, and a compensation mode can be determined by the compensation region. There can be four edge compensation modes and two band compensation modes. In the case of an edge compensation type, an edge type to which each pixel belongs is determined and an offset corresponding to the edge type is applied. The border type is determined based on the distribution of two pixel values ​​adjacent to a current pixel.

The adaptive loop filter procedure can be performed based on a value obtained by comparing an original image and a reconstructed image to which the deblocking filtering procedure or the adaptive compensation application procedure is applied. An adaptive loop filter (ALF) is detected through a Laplacian activity value based on a 4x4 block. The determined ALF can be applied to all pixels included in a 4x4 block or in an 8x8 block. Whether to apply an ALF or not can be determined according to each coding unit. The size and coefficients of a loop filter can vary according to each coding unit. A segment header may include information indicating whether or not to apply the ALF to each encoding unit, the filter coefficient information, and the filter shape information, and so on. In the case of chrominance components, it can be determined whether or not the ALF is applied in imaging units. Unlike luminance, the loop filter can be rectangular in shape.

Image storage unit 180 receives post-processing image data from post-processing unit 160 and stores the image in image units. An image can be an image in a frame or a field. Image storage unit 180 has a buffer memory (not shown) capable of storing a plurality of images.

The interprediction unit 160 performs motion estimation using one or more reference images stored in the image storage unit 180 and determines the reference image indices that indicate the reference images and motion vectors. According to the reference image index and the motion vector, the interprediction unit 160 extracts a prediction block corresponding to a prediction unit to be encoded from a reference image selected from a plurality of stored reference images in the image storage unit 180 and the extracted prediction block is output.

The intra-prediction unit 150 performs the intra-prediction using the reconstructed pixel values ​​within a current image. The intra-prediction unit 150 receives the current prediction unit to be predictively encoded, selects one of a predetermined number of intra-prediction modes, and performs an intra-prediction. The predetermined number of intra-prediction modes may depend on the size of the current prediction unit. The intra-prediction unit adaptively filters the reference pixels to generate the intra-prediction block. When some of the reference pixels are not available, it is possible to generate the reference pixels at the unavailable positions using one or more available reference pixels.

The entropy encoding unit 140 entropy encodes the quantized coefficients quantized by the quantization unit 130, the intra-prediction information received from the intra-prediction unit 150, the motion information received from the inter-prediction unit 160, and so on. successively.

FIG. 2 is a flow chart illustrating an operation of the scanning unit 131 in accordance with the present invention.

It is determined whether the current quantized coefficient block is divided into a plurality of subsets (S110). The determination is based on a current transform unit size. If the size of the current transform unit is larger than a first reference size, the encoded quantized coefficients are divided into a plurality of subsets. The first reference size can be 4x4 or 8x8. The first reference size can be transmitted to a decoder via a picture header or a picture header.

ES 2 634 819 T3 segment.

When the block of quantized coefficients is not divided into a plurality of subsets, a scan pattern to be applied to the block of quantized coefficients is determined (S120). Step S120 can be performed before step S110 or independently of step S110.

The quantized coefficients of the block of quantized coefficients are scanned according to the determined scan pattern (S130). The scan pattern is adaptively determined according to the prediction mode and the intra-prediction mode. In interprediction mode, only a predetermined scan pattern (eg zigzag scan) can be applied. In the intra-prediction mode, a scan pattern determined according to the intra-prediction mode can be applied. Furthermore, a predetermined number of scan patterns can be selected to scan the coefficients, and the scan pattern information can be transmitted to the decoder. In the intra-prediction mode, a scan pattern determined according to the intra-prediction mode can be applied. For example, a horizontal scan is applied to a vertical intraprediction mode and a predetermined number of intraprediction modes adjacent to the vertical intraprediction mode. A vertical scan is applied to a horizontal intra-prediction mode and a predetermined number of intra-prediction modes adjacent to the horizontal intra-prediction mode. The predetermined number varies according to a number of allowed prediction modes of a prediction unit (or a number of directional intra-prediction modes) or a size of a prediction block. For example, if the number of prediction modes allowed in the current prediction unit is 16, the predetermined number may be two in each of both directions based on the horizontal or vertical intra-prediction mode. If the number of allowed directional prediction modes is 33, the predetermined number can be four in each of both directions based on the horizontal or vertical intra-prediction mode. Meanwhile, zigzag scanning applies to non-directional modes. A non-directional mode can be a direct current (DC) mode or a flat mode.

If it is determined that the block of quantized coefficients is divided into a plurality of subsets, the block of quantized coefficients is divided into a plurality of subsets (S140). The plurality of sub-sets consists of a main sub-set and one or more remaining sub-sets. The main sub-set is located on the upper left side and covers a DC coefficient, and one or more remaining sub-sets cover a different region than the main sub-set.

A scan pattern to be applied to the subsets is determined (S150). The determined scan pattern applies to all subsets. The scan pattern is adaptively determined according to the prediction mode and the intra-prediction mode. Step S150 can be performed before step S110 or independently of step S110.

When the size of the block of quantized coefficients (that is, the size of the transform unit) is larger than a second reference size, the zigzag scan pattern can be applied to the block of quantized coefficients. The second reference size is, for example, 8x8. Therefore, step S150 is performed when the first reference size is smaller than the second reference size.

In interprediction mode, only one predetermined scan pattern (eg zigzag scan) can be applied to each subset. In the intra-prediction mode, the scan pattern is adaptively determined as in step S120.

The quantized coefficients in the subsets can be explored in an inverse direction. In other words, according to the scan pattern, the non-0 quantized coefficients can be scanned and entropy encoded in the reverse direction starting with the last non-0 quantized coefficient in the subsets.

Next, the quantized coefficients of each subset are scanned according to the scan pattern (S160). The quantized coefficients in each subset are scanned in the reverse direction. That is, the quantized transform coefficients are scanned from a last non-zero coefficient to other non-zero coefficients according to the scan pattern, and by entropy coding.

Zigzag scanning can be applied to scan subsets. The sub-sets can be scanned starting with the main sub-set to the remaining sub-sets in a forward direction, or they can be scanned in the reverse direction. A scan pattern to scan the sub-sets can be set the same as a scan pattern to scan the quantized coefficients in the sub-sets.

The moving picture coding apparatus 100 according to the present invention transmits the information capable of indicating a position of the last non-zero quantized coefficient of the transform unit to a decoder. The moving picture encoding apparatus 100 also transmits the information capable of indicating a position of the last non-zero quantized coefficient in each subset to the decoder.

Fig. 3 is a block diagram illustrating a moving picture decoding apparatus in accordance with the present invention.

ES 2 634 819 T3

The moving picture decoding apparatus according to the present invention includes a unit

210 entropy decoding unit, a reverse scan unit 220, an inverse quantization unit 230, an inverse transform unit 240, an intraprediction unit 250, an interprediction unit 260, a processing unit 270, an interprediction unit 280 image storage, an adder 290 and an intra / exchange switch 295.

The entropy decoding unit 210 extracts intra-prediction information, inter-prediction information, and quantized coefficient information from a received bit stream. Entropy decoding unit 210 transmits interprediction information to interprediction unit 260, intraprediction information to intraprediction unit 250 and inverse transform unit 240, and inverse quantized coefficient information to unity. 220 reverse scan.

The reverse scan unit 220 converts the quantized coefficient information into a two-dimensional quantized transform block. One of a plurality of reverse scan patterns is selected for conversion. The reverse scan pattern is selected based on at least one of the prediction mode and the intra-prediction mode. The operation of the reverse scan unit 220 is the same as the reverse operation of the scan unit 131 of FIG. 1. For example, if a size of a current transform unit to be decoded is larger than the first reference size, each subset is scanned in reverse according to the selected reverse scan pattern and an inverse quantized block is generated that has the size of the transform unit using the plurality of inversely scanned subsets.

The inverse quantization unit 230 determines a quantization step size predictor of the current coding unit. The operation for determining the quantization step size predictor is the same as the procedure of the quantization unit 130 of FIG. 1. The inverse quantization unit adds the determined quantization step size predictor and a received residual quantization step size size to generate a quantization step size of the current coding unit. The inverse quantization unit 230 resets the inverse quantized coefficients using a quantization matrix determined by the quantization step size. The quantization matrix varies according to the size of the current block to be reset. The quantization matrix can be selected for a block having the same size based on at least one of a prediction mode and an intra-prediction mode of the current block.

The inverse transform unit 240 inverse transforms the inverse quantized block to restore a residual block. The inverse transform matrix that is applied to the inverse quantized block is adaptively determined according to the prediction mode (intra or inter) and the intra prediction mode. The inverse transform matrix determination procedure is the same as the procedure in transform unit 120 of FIG. 1.

Adder 290 adds the restored residual block restored by inverse transform unit 240 and a prediction block generated by intra-prediction unit 250 or inter-prediction unit 260 to generate a reconstructed image block.

The intra-prediction unit 250 resets the intra-prediction mode of the current block based on the intra-prediction information received from the entropy decoding unit 210, and generates a prediction block in accordance with the restored intra-prediction mode.

The interprediction unit 260 resets the reference image indices and motion vectors based on the interprediction information received from the entropy decoding unit 210, and generates a prediction block using the reference image indices and vectors of motion. When motion compensation is applied with fractional precision, the prediction block is generated using an interpolation filter.

Post-processing unit 270 functions the same as post-processing unit 160 of FIG. 1.

Image storage unit 280 stores the image reconstructed by post-processing by post-processing unit 270

FIG. 4 is a block diagram illustrating the intra-prediction unit 150 of a moving picture coding unit 100 in accordance with the present invention.

Referring to FIG. 4, the intra-prediction unit 150 includes a reference pixel generation unit 151, a reference pixel filter unit 152, a prediction mode determination unit 153, a reference pixel generation unit 154 prediction block, a prediction block filter unit 155, and a prediction mode encoding unit 156.

The reference pixel generation unit 151 determines that it is necessary to generate the reference pixels for the intra-prediction, and generates the reference pixels if necessary to generate the reference pixels.

ES 2 634 819 T3

Figure 5 is a conceptual diagram showing the positions of the reference pixels used for intra prediction in accordance with the present invention. As shown in Fig. 5, the reference pixels consist of the upper reference pixels, the left reference pixels, and a corner reference pixel of the current prediction unit. The top reference pixels of the current prediction unit are pixels (C and D regions) present along twice the width of the current prediction unit and the left reference pixels of the current prediction unit are pixels ( regions A and B) present along twice the height of the current prediction unit.

The reference pixel generation unit 151 determines whether the reference pixels are available or not. If one or more reference pixels are not available, the reference pixel generation unit 151 generates the reference pixels at the unavailable positions using an available reference pixel.

First, a case will be described where all the reference pixels in any of the upper and left regions of a current prediction unit to be encoded are not available.

For example, when the current prediction unit is located at the upper limit of an image or a segment, the upper reference pixels (regions C and D) and the corner reference pixel of the current prediction unit do not exist. When the current prediction unit is located at the left boundary of an image or a segment, the left reference pixels (regions A and B) and the corner reference pixel do not exist. In those cases, the reference pixels are generated by copying the value of an available pixel closest to the unavailable pixel. That is, when the current prediction unit is located at the upper limit of an image or a segment, the upper reference pixels can be generated by copying the uppermost left reference pixel (i.e., a reference pixel located at position highest in region A). When the current prediction unit is located at the left boundary of an image or segment, the left reference pixels can be generated by copying the uppermost reference pixel to the left (i.e., a reference pixel located at the furthest position). to the left of the region of C). The above mentioned procedure is applied by default, but the procedure can vary by sequence, image or segment if necessary.

Next, a case will be described where some of the upper or left reference pixels of a current prediction unit to be encoded are not available. There are two cases where 1) the available reference pixels are present in only one direction relative to the unavailable reference pixels, and 2) the available reference pixels are present in both directions relative to the non-available reference pixels. available.

Case 1) will be described.

For example, when the current block is located on the right edge of an image or a segment or an LCU, the reference pixels that cover the area D are not available. Furthermore, when the current block is located at the lower limit of an image or a segment or an lCu, the reference pixels that cover the B area are not available. In this case, the reference pixels are generated by copying the values ​​of the available pixels closest to the unavailable pixel. Also, the reference pixels are generated using two or more available pixels closest to the unavailable pixel.

Case 2 will be described).

For example, when the current block is located at the upper limit of a segment and the upper left block of the current block is available, the reference pixels that cover area C are not available, but the reference pixels that cover areas A and D are available. When the available reference pixels are therefore present in both directions, an available reference pixel present in the closest position in each direction is selected, and the reference pixels in the unavailable positions are generated using the selected reference pixels. (that is, the highest reference pixel in region A and the leftmost reference pixel in region D).

A value obtained by rounding an average of the two reference pixels (pixels present at the closest positions in the respective directions) can be generated as a reference pixel value. However, when an unavailable reference pixel region is large, there is a high possibility that a pitch difference will occur between an available pixel and a generated pixel, and thus it is useful to generate the reference pixels using the linear interpolation. Specifically, in consideration of a position relative to two available reference pixels, an available reference pixel can be generated at the current position.

Next, a case will be described where all the reference pixels on the upper and left sides of a current prediction unit to be encoded are not available. For example, when a current prediction unit is adjacent to an upper left edge of an image or segment, no reference pixels are available.

In this case, some or all of the reference pixels can be generated using two or more pixels present in the current prediction unit. The number of pixels that are present in the current prediction unit and that are

ES 2 634 819 T3 used to generate the reference pixels can be two or three.

FIG. 6 is a flow chart illustrating a reference pixel generation procedure in accordance with the present invention.

Referring to Fig. 6, the reference pixel generation procedure using two pixels is as follows. An upper left pixel O and one of an upper right pixel □, a lower left pixel Δ and a lower right pixel V of the current prediction unit can be used. When the upper left pixel O and the upper right pixel Π are used. of the current prediction unit, the upper left pixel and the upper right pixel are copied to the corresponding positions on an upper side and the upper right pixel and the copied reference pixels are used to generate the reference pixels that cover area C Reference pixels are generated using average or linear interpolation. The reference pixels covering D are generated by copying the upper right pixel Π, or using a plurality of the generated upper pixels. When the upper left pixel O and the lower left pixel Δ of the current prediction unit are used, the same procedure applies. When the upper left pixel O and the lower right pixel V are used, the lower right pixel V is copied to the corresponding reference pixel position in the horizontal direction and in the vertical direction, and then the residual reference pixels of the same way as described above.

The reference pixel generation procedure using three pixels is as follows. An upper left pixel O, an upper right pixel □, and a lower Left pixel Δ of the current prediction unit can be used. The pixels are copied to the corresponding reference pixel position and then the residual reference pixels are generated using the copied pixels. The residual reference pixels are generated in the same way as described above.

Meanwhile, when using a method as described above, the values ​​of the pixels used to generate the reference pixels are transmitted to the decoder. To minimize the number of bits to transmit, the value of the upper left pixel O, and the difference between the value of the upper left pixel O, and the values ​​of other pixels. The value of the upper left pixel can be a quantized value or be entropy encoded.

When a segment type is intra (I), it is more efficient to generate the reference pixels using two or more pixels.

Another reference pixel generation method will be described when all the reference pixels on the upper and left sides of a current prediction unit to be encoded are not available. This procedure is effective when a segment type is not intra (I).

First, it is determined whether the pixels are present in the same positions as the reference pixels of a current prediction unit in a previously encoded reference image for a current block. When pixels are present, the reference image pixels are copied to generate the reference pixels of the current prediction unit.

When pixels are not present, it is determined whether the pixels are present in the closest positions (1 pixel apart) to refer to the pixels in the current prediction unit. When pixels are present, the pixels are copied and used as the reference pixels for the current prediction unit.

The reference pixel filtering unit 152 adaptively filters the reference pixels of the current prediction unit. A low pass filter is applied to smooth out a variance of the pixel values ​​between the reference pixels. The low pass filter can be a 3-lead filter [1, 2, 1] or a 5-lead filter [1, 2, 4, 2, 1].

The filter can be applied adaptively according to a current block size. If the current block size is equal to or less than a predetermined size, the filter cannot be applied. The default size can be 4x4.

The filter can also be applied adaptively according to a current block size and intra-prediction mode.

If the intra-prediction mode is horizontal mode or vertical mode, the pixels of a prediction block are generated using a reference pixel. Therefore, a filter is not applied in landscape mode and portrait mode. In DC mode, a prediction pixel is generated using the average of the reference pixels. Therefore, a filter is not applied in DC mode because the prediction pixel is not affected by the difference between the reference pixels.

In the 3, 6 or 9 intra-prediction mode that has a direction of 45 ° with respect to the horizontal or vertical direction, a filter is applied regardless of the size of the prediction unit or applied when the current block is greater than a smaller prediction unit. A first filter can be applied to a unit of

ES 2 634 819 T3 having a size smaller than a predetermined size, and a second filter stronger than the first filter can be applied to a prediction unit having a size equal to or greater than the predetermined size. The default size can be 16x16.

In intra-prediction modes other than portrait mode, horizontal mode, DC mode, and intra-prediction mode 3, 6 and 9, a filter can be applied adaptively according to the current prediction unit size and the current prediction unit size. intra prediction mode. However, in flat mode, filtering of the reference pixels can be performed.

Also, the filter cannot be applied to some or all of the reference pixels generated through linear combining.

The prediction block generation unit 153 generates a prediction block corresponding to the intra-prediction mode. The prediction block is generated using the reference pixels or a linear combination of the reference pixels based on the intra-prediction mode. The reference pixels that are used to generate the prediction block can be filtered by the reference pixel filter unit 152.

The prediction block filtering unit 154 adaptively filters the generated prediction block according to the intra-prediction mode to minimize the residual signal between the prediction block and the current block to be encoded. The difference between a reference pixel and a prediction pixel adjacent to the reference pixel varies according to the intra-prediction mode. Therefore, filtering the prediction pixel adjacent to the reference pixel allows the difference to be decreased.

In DC mode, the prediction block consists of the means of the reference pixels, and a pitch difference may occur between the pixels in a prediction block adjacent to the reference pixels. Therefore, the prediction pixels of the upper line and the left line that are adjacent to the reference pixels are filtered using the reference pixels. The upper left prediction pixel adjacent to two reference pixels (the upper reference pixel and the left reference pixel) is filtered by a 3-lead filter. The other prediction pixels (the pixels of the top line and the pixels of the left line in the prediction block) and adjacent to a reference pixel are filtered by the 2-lead filter.

In portrait mode (mode 0), lower-left diagonal mode (mode 6), and intra prediction modes (modes 22, 12, 23, 5, 24, 13, 25) between mode 0 and mode 6 , a prediction block is generated using only the upper reference pixels of the prediction unit. Therefore, the difference between a reference pixel and a prediction pixel adjacent to the reference pixel and of the upper line in the prediction block can be larger than the position of the prediction pixel going down.

Also, in horizontal mode (mode 1), up-right diagonal mode (mode 9), and intra prediction modes (modes 30, 16, 31, 8, 32, 17, 33) between mode 1 and mode 9, a prediction block is generated using only the left reference pixels. Therefore, the difference between a reference pixel and a prediction pixel adjacent to the reference pixel and from the top line in the prediction block can be larger than the position of the prediction pixel going to the right.

Consequently, some prediction pixels in the prediction block may be filtered to compensate for the difference in directional intra-prediction mode other than DC mode.

In mode 6, all or some prediction pixels adjacent to the left reference pixel and the left line in the prediction block are filtered. The pixels of the left line can be, for example, N / 2 pixels in a lower part of the left line. In this case, N is the height of the current prediction unit.

In mode 9, all or some of the prediction pixels adjacent to the top reference pixel and the top line are filtered out in the prediction block. The pixels of the top line can be, for example, M / 2 pixels in a right part of the top line. In this case, M is the width of the current prediction unit.

In a predetermined number of directional intraprediction modes that are closer to mode 6 and exist between mode 0 and mode 6, the same filtering procedure can be used as in mode 6. In this case, the further away is a Mode 6 mode, the number of pixels to be filtered is less than or equal.

The same filtering procedure can be used in directional intra-prediction modes between mode 1 and mode 9.

Meanwhile, the filter is applied adaptively according to the size of the current prediction unit. For example, no filtering is applied to a predetermined size or less according to the intra-prediction mode.

The prediction filter block unit 154 may be integrated into the prediction block generation unit 153. In addition, a prediction block can be generated to produce prediction block filtering effects. In this case, the prediction block is generated using the combination of the operation of

ES 2 634 819 T3 generation and filtering operation.

The intra-prediction mode determining unit 155 determines the intra-prediction mode of a current prediction unit using the reference pixels. The intra-prediction mode determining unit 155 selects an intra-prediction mode in which the amount of coding bits in a residual block is minimized as the intra-prediction mode of the current prediction unit. To generate a residual block, a prediction block is generated according to each intra-prediction mode. The prediction block can be generated using the reference pixels filtered by the reference pixel filter unit or it can be a block filtered by the prediction block filter unit 154.

The prediction block transmission unit 157 transmits the prediction block generated based on the intra-prediction mode by the intra-prediction mode determining unit 155 to the subtractor.

The prediction mode encoding unit 156 encodes the intra-prediction mode of the current prediction unit determined by the intra-prediction mode determining unit 155. The prediction mode coding unit 156 may be integrated into the intra prediction unit 150 or the entropy coding unit 140.

The prediction mode encoding unit 156 encodes the intra-prediction mode of the current prediction unit using an upper intra-prediction mode of the current prediction unit and a left intra-prediction mode of the current prediction unit.

First, the upper intra-prediction mode and the left intra-prediction mode of the current prediction unit are obtained. When there is a plurality of prediction units higher than the current prediction unit, the plurality of previous prediction units is scanned in a predetermined direction (for example, from right to left) to determine the intra-prediction mode of a first unit of prediction available as a superior intra prediction mode. Furthermore, when there is a plurality of left prediction units of the current prediction unit, the plurality of left prediction units is scanned in a predetermined direction (for example, from bottom to top) to determine the intra-prediction mode of a first Prediction unit available as a left intra prediction mode. Alternatively, among a plurality of available prediction units, the intra-prediction mode of an available prediction unit having the lowest prediction mode number can be set as a higher intra-prediction mode.

When the top intra-prediction mode or the left intra-prediction mode is not available, the DC mode (mode 2) can be set as the top intra-prediction mode or as the left intra-prediction mode. The upper intra-prediction mode or the left intra-prediction mode is treated as not available when there is no corresponding prediction unit.

Then the top intra-prediction mode or the left intra-prediction mode becomes one of the predetermined number of intra-prediction modes when the top intra-prediction mode number or the left intra-prediction mode number equals or greater than the number of allowable intra-prediction modes for the current prediction unit. The predetermined number may vary according to a current prediction unit size. For example, when the current prediction unit size is 4x4, the intra prediction mode becomes one of nine modes (mode 0 to mode 8), and when the current prediction unit size is 64x64, the Intra prediction mode is mapped in one of three modes (mode 0 to mode 2). The intra-prediction mode can become one of the allowable intra-prediction modes for the current prediction unit.

Then, if the intra-prediction mode of the current prediction unit is the same as any one of the upper and left intra-prediction modes, a flag is transmitted to the decoder indicating that the intra-prediction mode of the prediction unit current is the same as any one of the top and left intra prediction modes and a flag indicating one of the top and left intra prediction modes. In this case, if the upper and left intra-prediction modes are the same, the flag indicating one of the upper and left intra-prediction modes may be omitted. Also, if only one of the top and left intra-prediction modes is available and the available intra-prediction mode is the same as the intra-prediction mode of the current prediction block, the flag indicating one of the intra-prediction modes can be omitted. top and left prediction.

But, if the intra-prediction mode of the current prediction unit is not the same as either one of the upper and left intra-prediction modes, the intra-prediction mode number of the current prediction unit is compared with the number of intra-prediction mode upper and with the intra-prediction mode number left. The number of cases in which the left or higher intra-prediction mode numbers is no greater than the intra-prediction mode number of the current prediction unit is calculated, and the value obtained by subtracting the number of cases from the mode number The intra-prediction mode number of the current prediction unit is determined as a final intra-prediction mode number of the current prediction unit to be transmitted. In this case, when the left and upper intraprediction mode numbers are identical, the upper and upper intraprediction modes

ES 2 634 819 T3 left are considered as one.

According to whether the upper and left intra-prediction modes are identical or not, a table for entropy coding of the determined final intra-prediction mode is determined.

FIG. 7 is a block diagram illustrating the intra prediction unit 250 of a moving picture decoding apparatus 200 in accordance with the present invention.

The intra-prediction unit 250 according to the present invention includes a prediction mode decoding unit 251, a reference pixel generation unit 252, a reference pixel filter unit 253, a block generation unit 254 prediction, a prediction block filtering unit 255 and a prediction block transmission unit 256.

The prediction mode decoding unit 251 resets the intra prediction mode of a current prediction unit as follows.

First, the prediction mode decoding unit 251 receives the additional information included in the additional information container to generate a prediction block. The additional information includes a prediction flag and the residual prediction information. The prediction flag indicates whether the intra-prediction mode of the current prediction unit is the same as one of the intra-prediction modes of the adjacent prediction units. The residual prediction information includes the information determined by the prediction flag. If the prediction flag is 1, the residual prediction information may include an index of the candidate intra prediction mode. The intra-prediction mode index designates the candidate intra-prediction mode. If the prediction flag is 0, the residual information may include the residual intra prediction mode number.

The candidate intra-prediction modes are obtained from the current prediction unit. The candidate intra-prediction modes are obtained using the intra-prediction modes of the adjacent prediction units. For convenience, a case will be described in which the candidate intra-prediction mode of the current prediction unit is limited to the upper and left intra-prediction modes. When there is a plurality of upper prediction units or a plurality of left prediction units, the intra-prediction mode of the upper or left prediction unit is determined therefrom as described in the operation of the mode encoding unit 156. intra-prediction of the encoding apparatus 100. Furthermore, when the mode number of an available candidate intra-prediction mode is equal to or greater than the number of allowable intra-prediction modes for the current prediction unit, the available candidate intra-prediction mode becomes one of the allowable modes. for the current prediction unit as described in the intra prediction mode coding unit 156.

Next, when the received prediction flag indicates that the current prediction unit has the same intra prediction mode as an adjacent prediction unit, and the candidate prediction mode index exists, a prediction mode indicated by the index is determined. of candidate prediction mode as the intra-prediction mode of the current prediction unit.

If the received prediction flag indicates that the current prediction unit has the same intra prediction mode as an adjacent prediction unit, but there is no candidate prediction mode index and intra prediction mode available from the adjacent prediction unit , the available intra-prediction mode is reset to the intra-prediction mode of the current prediction unit.

If the received prediction flag indicates that the current prediction unit does not have the same intra prediction mode as an adjacent prediction unit, a residual received intra prediction mode value is compared to the intra prediction mode numbers of the modes. available intra-prediction candidates to reset the intra-prediction mode of the current prediction unit.

The reference pixel generation unit 252 generates the reference pixels using the same procedure as described in the reference pixel generation unit 151 of the encoding apparatus 100. However, the reference pixel generation unit 252 is different from the reference pixel generator 151 of the encoding apparatus 100 in that the reference pixels are adaptively generated according to the intra-prediction mode reset by the decoder 251. prediction mode. That is, the reference pixel generation unit 252 can generate the reference pixels only when the reference pixels used to generate a prediction block and determined by the intra-prediction mode are not available.

The reference pixel filtering unit 253 adaptively filters the reference pixels based on the intra-prediction mode restored by the prediction mode decoding unit 251 and a prediction block size. The filtering condition and a filter are the same as those of the reference pixel filtering unit 152 of the encoding apparatus 100.

ES 2 634 819 T3

The prediction block generation unit 254 generates a prediction block using the reference pixels in accordance with the intra-prediction mode reset by the prediction mode decoding unit 251.

The prediction block filtering unit 255 adaptively filters the prediction block according to the intra-prediction mode restored by the prediction mode decoding unit 251. The filtering operation is the same as that of the prediction block filtering unit 154 of the encoding apparatus 100.

Prediction block transmission unit 256 transmits the prediction block received from prediction block generator 254 or prediction block filter unit 255 to adder 290.