KR102096407B1 - A method and an apparatus for decoding a video signal - Google Patents

Abstract

The present invention comprises the steps of obtaining a transform size flag of a current macroblock from a video signal; Checking a number of non-zero transform coefficients at each pixel position in the first transform block for a first transform block corresponding to the transform size flag; Changing a scan order of the first transform block by setting a position of a pixel having the highest number of non-zero transform coefficients in the first transform block as a priority; Determining the number of non-zero transform coefficients at each pixel position in the second transform block, and setting the changed scan order of the first transform block to an initialized scan order of the second transform block; Add the number of non-zero transform coefficients at each pixel position in the first transform block and the number of non-zero transform coefficients at each pixel position in the second transform block, and prioritize the position of the pixel with the largest number. Changing the scan order of the second transform block to? And decoding a transform coefficient arranged according to the changed scan order, wherein the first transform block and the second transform block have a size corresponding to the transform size flag and are included in the current macroblock. It provides a method for decoding a video signal characterized in that.

Description

Decoding method and apparatus for video signal {A METHOD AND AN APPARATUS FOR DECODING A VIDEO SIGNAL}

The present invention relates to a method and apparatus for decoding a video signal.

Compression coding refers to a series of signal processing technologies that transmit digitized information over a communication line or store it in a form suitable for a storage medium. There are objects such as voice, video, and text in the object of compression encoding, and a technique for performing compression encoding on an image is called video video compression. The general feature of the video image is that it has spatial redundancy and temporal redundancy.

If the spatial redundancy and temporal redundancy are not sufficiently removed, the compression rate in encoding a video signal is lowered, and when the spatial redundancy and temporal redundancy are excessively removed, information necessary for decoding the video signal is not generated. There was a problem in that the restoration rate deteriorated because it could not.

The present invention comprises the steps of obtaining a transform size flag of a current macroblock from a video signal; Checking a number of non-zero transform coefficients at each pixel position in the first transform block for a first transform block corresponding to the transform size flag; Changing a scan order of the first transform block by setting a position of a pixel having the highest number of non-zero transform coefficients in the first transform block as a priority; Determining the number of non-zero transform coefficients at each pixel position in the second transform block, and setting the changed scan order of the first transform block to an initialized scan order of the second transform block; Add the number of non-zero transform coefficients at each pixel position in the first transform block and the number of non-zero transform coefficients at each pixel position in the second transform block, and prioritize the position of the pixel with the largest number. Changing the scan order of the second transform block to? And decoding a transform coefficient arranged according to the changed scan order, wherein the first transform block and the second transform block have a size corresponding to the transform size flag and are included in the current macroblock. It provides a method for decoding a video signal characterized in that.

In addition, the present invention further includes the step of obtaining scan order initialization flag information indicating whether to initialize a scan order for each transform block in the current macroblock, wherein the scan order initialization flag information is for each transform block. When instructing to initialize the scan order, it is characterized in that the initialized scan order of the second transform block is set as the changed scan order of the first transform block.

In addition, in the present invention, the conversion size flag information is characterized in that a plurality.

In addition, in the present invention, the conversion size flag information includes first conversion size flag information indicating whether the conversion process is performed in units of 16x16 blocks or smaller block units, and the conversion process in units of 8x8 blocks. It is characterized in that it includes second transform size flag information indicating whether to perform or perform a transform process in units of 4x4 blocks.

Further, in the present invention, when the first conversion size flag information indicates that the conversion process is performed in block units smaller than 16x16 block units, and the second conversion size flag information indicates that the conversion process is performed in 8x8 block units , The first transform block and the second transform block correspond to blocks of 8x8 size.

In addition, the present invention further comprises the step of obtaining transform type flag information indicating a transform type from the video signal, wherein decoding the transform coefficient comprises: discrete cosine transform (DCT) based on the transform type flag information. Or it is characterized by using the Karenen-Lube transform (KLT).

In addition, in the present invention, when the conversion type flag information indicates to use the Karenen-Lube conversion (KLT) and the block type of the current macroblock indicates an intra block, the Karinen-Lube conversion (KLT) is used to And decoding the transform coefficients.

In addition, the present invention, the entropy decoding unit for obtaining a transform size flag of the current macroblock from the video signal; For the first transform block corresponding to the transform size flag, the number of non-zero transform coefficients at each pixel position in the first transform block is checked, and 0 in the first transform block A first scan order changing unit for changing a scan order of the first transform block by setting a position of a pixel having the largest number of transform coefficients as a priority; Determine the number of non-zero transform coefficients at each pixel position in the second transform block, set the changed scan order of the first transform block to the initialized scan order of the second transform block, and within the first transform block The number of non-zero transform coefficients at each pixel position and the number of non-zero transform coefficients at each pixel position in the second transform block are added, and the position of the pixel having the largest number is prioritized. A second scan order changing unit for changing the scan order of the block; And a transform decoding unit for decoding transform coefficients arranged according to the changed scan order, wherein the first transform block and the second transform block have a size corresponding to the transform size flag, and are included in the current macroblock. It provides a video signal decoding apparatus characterized in that.

The present invention can improve coding efficiency by variously defining the size of a unit block used in a conversion process.

In addition, the present invention can improve coding efficiency by defining flag information indicating the size of a unit block used in a conversion process.

In addition, the present invention defines a flag indicating a transform type, and accordingly a discrete cosine transform (hereinafter referred to as 'DCT') or a Karenen-Loeve transform (hereinafter referred to as 'KLT') By using, coding efficiency can be improved.

In addition, the present invention can improve coding efficiency by using DCT or KLT according to the prediction mode of the macroblock.

In addition, the present invention can improve coding efficiency by adaptively setting the size of the unit block used in the conversion process to be smaller than or equal to the size of the residual block.

In addition, the present invention can improve coding efficiency by using a different transform kernel for each transform block of a macroblock coded in intra prediction mode.

In addition, the present invention can improve coding efficiency by updating a scanning order for each transform block of a macroblock.

In addition, the present invention can improve coding efficiency by removing redundancy of information indicating whether a non-zero transform coefficient level is included.

The above inventions can further improve coding efficiency when used in a macroblock-based encoding and decoding process that is larger than a 16x16 macroblock.

1 is an embodiment to which the present invention is applied, and shows a schematic block diagram of a video signal decoding apparatus.
2, 3A and 3B are embodiments to which the present invention is applied, and show a process of determining the size of a unit block used in a conversion process based on flag information.
4 is an embodiment to which the present invention is applied, and is a flowchart illustrating a process of determining a transform method (DCT or KLT) based on a prediction mode of a flag or macroblock indicating a transform type.
5 is an embodiment to which the present invention is applied, and shows a schematic block diagram for explaining a new Mode Dependent Directional Transform that applies a different transform kernel according to the location of a transform block.
6 is an embodiment to which the present invention is applied, and shows a specific example of using a different transform kernel according to the location of the transform block.
7 is an embodiment to which the present invention is applied, and shows a flowchart for explaining a process of rearranging a scan order for each transform block.
8A to 8D are embodiments to which the present invention is applied, and show a specific example for explaining the process of rearranging the scan order for each transform block.

Looking at the configuration of a bit stream of a video signal, a Network Abstraction Layer (NAL) between a video coding layer (VCL) that deals with video encoding processing itself and a subsystem that transmits and stores encoded information is a network abstraction layer. Hierarchy). The output of the encoding process is VCL data and is mapped in NAL units before transmission or storage. Each NAL unit includes compressed video data or RBSP (Raw Byte Sequence Payload), which is data corresponding to header information.

The NAL unit basically consists of two parts: the NAL header and the RBSP. The NAL header includes flag information (nal_ref_idc) indicating whether a slice serving as a reference picture of the NAL unit is included, and an identifier (nal_unit_type) indicating the type of the NAL unit. In the RBSP, the compressed original data is stored, and the RBSP trailing bit is added to the end of the RBSP to express the length of the RBSP in multiples of 8 bits. These types of NAL units include IDR (Instantaneous Decoding Refresh) pictures, SPS (Sequence Parameter Set), PPS (Picture Parameter Set, Picture Parameter Set), SEI (Supplemental Enhancement Information, Supplementary Supplementary Information) And so on.

In addition, in the standard, the target product is restricted to various profiles and levels so that it can be implemented at a reasonable cost, and the decoder must satisfy the constraints specified in the profile and level. In this way, two concepts of profile and level have been defined to indicate the function or parameter of the range of the compressed image to which the decoder can respond. Which profile the bitstream is based on can be identified by a profile identifier (profile_idc). The profile identifier means a flag indicating a profile on which the bitstream is based. For example, in H.264 / AVC, a profile identifier of 66 means based on a baseline profile, 77 means based on a main profile, and 88 means based on an extended profile. The profile identifier may be included in the sequence parameter set.

The sequence parameter set refers to header information including information that spans the entire sequence, such as profile and level. Since the entire compressed video, that is, the sequence must start from the sequence header, the sequence parameter set corresponding to the header information must arrive at the decoder before the data referring to the parameter set. As a result, the sequence parameter set RBSP serves as header information for data resulting from video compression. When a bitstream is input, first, the profile identifier identifies which profile the input bitstream is based on among a plurality of profiles.

1 shows a schematic block diagram of a video signal decoding apparatus to which the present invention is applied.

Referring to FIG. 1, the decoding apparatus is largely an entropy decoding unit 100, an inverse quantization unit 200, an inverse transform unit 300, an intra prediction unit 400, a deblocking filter unit 500, and a decoded picture buffer unit. 600, an inter prediction unit 700, and the like. The inverse transform unit 300 includes a transform size determining unit 310, a scan order initialization unit 320, a scan order changing unit 330, and a transform decoding unit 340.

First, the decoding apparatus parses the received video image in units of NAL. Generally, one or more sequence parameter sets and picture parameter sets are transmitted to the decoder before the slice header and slice data are decoded. At this time, various attribute information may be included in the NAL header region or the extended region of the NAL header.

The parsed bitstream is entropy decoded through the entropy decoding unit 100, and coefficients, motion vectors, and the like of each macroblock are extracted. The inverse quantization unit 200 multiplies the entropy-decoded data by a constant constant to obtain a transform coefficient value, and the inverse transform unit 300 inversely transforms the transform coefficient value to restore a pixel value. At this time, the transform size determiner 310 may obtain transform size related information to determine the transform size. For example, adaptive transform size flag information or transform size flag information may be obtained. This will be described in more detail in FIGS. 2, 3A and 3B. The scan order initialization unit 320 initializes a scan order for inverse quantized transform coefficients, and the scan order changer 330 changes to a scan order suitable for each block based on the number of non-zero transform coefficients. Or update. The transform decoding unit 340 decodes the transform coefficients based on the changed or updated scan order. On the other hand, in the encoding apparatus, the compression effect can be increased as the number of the transform coefficient values is reduced through this conversion process, and the compression effect can be more concentrated by allowing important data in the screen to be intensively distributed in some coefficients through the conversion process. Can be increased. Accordingly, embodiments of the conversion process will be described in the present invention.

Then, the intra prediction unit 400 performs intra prediction from the decoded sample in the current picture using the reconstructed pixel value. The intra prediction unit 400 predicts a current block using pixels of a neighboring block of the current block in the current picture. If more accurate prediction can be performed, not only the image quality can be improved, but also coding efficiency can be improved. There will be.

The deblocking filter unit 500 is applied to each coded macroblock to reduce block distortion. The filter smooths the edges of the block to improve the quality of the decoded frame. The choice of filtering process depends on the boundary strenth and the gradient of the image sample around the boundary. The filtered pictures are output or stored in the decoded picture buffer unit 600 for use as a reference picture.

The decoded picture buffer unit 600 may serve to store or open previously coded pictures in order to perform inter prediction. At this time, in order to store or open the decoded picture buffer unit 600, information indicating a decoding order of each picture (frame_num) and information indicating a picture output order (Picture Order Count, POC) are used. The reference pictures managed in this way may be used by the inter prediction unit 700.

The inter prediction unit 700 compensates for the motion of the current block by using the information transmitted from the entropy decoding unit 100. A motion vector of blocks neighboring the current block is extracted from a video signal, and a motion vector prediction value of the current block is obtained. Motion of the current block is compensated by using the obtained motion vector prediction value and a difference vector extracted from the video signal. Further, such motion compensation may be performed using one reference picture or may be performed using a plurality of pictures.

The decoding apparatus acquires a predicted value generated through an inter prediction or intra prediction process according to the prediction mode information. Then, the current picture may be reconstructed by adding the residual extracted from the video signal and the obtained prediction value. Here, residual refers to a difference between an original signal and a prediction signal, and the signal may be a sequence unit, a picture unit, a slice unit, a macroblock unit, a subblock unit, or a pixel unit, and within the specification The unit in can be interpreted as a unit suitable for each decoding process.

In addition, the present invention can be used for the processing of video signals with more extended resolution. That is, an extended size macroblock can be defined for processing a video signal having an extended resolution. For example, 16x16, 16x8, 8x16 8x8, 8x4, 4x8, and 4x4 macroblocks, as well as macroblocks having sizes of 32x32, 64x32, 32x64, and 64x64 may be used. The size of the macroblock can be adaptively determined according to the resolution of the video signal. For example, if the resolution of the video signal is VGA (640x480, 4: 3) or less, a macroblock of 16x16 size can be used, and if VGA or more is 1080P or less, a macroblock of 32x32 size can be used, and further, 1080P or more and 4kx2k In the following cases, a macroblock having a size of 64x64 may be used.

2, 3A and 3B are embodiments to which the present invention is applied, and show a process of determining the size of a unit block used in a conversion process based on flag information.

Among the conversion techniques used in the encoding device and the decoding device, there are a block-based conversion method and an image-based conversion method. Examples of the block-based transform method include a discrete cosine transform (hereinafter referred to as 'DCT') and a Karen-Loeve transform (hereinafter referred to as 'KLT'). Here, the discrete cosine transform (DCT) decomposes (converts) a signal on a spatial domain into a two-dimensional frequency component. In the block, a pattern having a lower frequency component toward the upper left and a higher frequency component toward the lower right forms a pattern. For example, of the 64 two-dimensional frequency components, only one present in the upper left corner is a component having a frequency of 0 as a direct current (DC) component, and the rest is a low frequency as an alternating current (AC). It consists of 63 components, from components to high frequency components. Performing discrete cosine transform (DCT) is to obtain the size of each of the base components (64 basic pattern components) included in the block of the original video signal, which is a discrete cosine transform coefficient.

In addition, the discrete cosine transform (DCT) is a transform used to simply express the original video signal component, and is completely restored from the frequency component to the original image signal during inverse transform. That is, by changing only the expression method of the image, all information included in the original image, including the duplicated information, is preserved. When the original image is subjected to discrete cosine transform (DCT), unlike the amplitude distribution of the original image, the discrete cosine transform (DCT) coefficients are concentrated at values near 0, so that a high compression effect can be obtained.

On the other hand, Karenen-Lube transform (KLT) processes data by reducing the dimension to the lower dimension (K-dimension) while maintaining the original information as much as possible with respect to data composed of multidimensional (N-dimension) feature vectors. It is one of the methods. That is, by extracting intrinsic components that represent the characteristics of the data well and taking only those values of which the feature component is high, the dimension is reduced and the difference from the original information is minimized. For example, the Karenen-Lube transform (KLT) method may be used according to a prediction mode used when encoding or decoding an intra block, and the coding efficiency may be improved by applying a scanning order method differently according to the prediction mode.

As an embodiment of the present invention, the size of a unit block used in a conversion process (hereinafter, referred to as “conversion size”) may be adaptively set for one macroblock. For example, when processing a video signal having a large resolution, the conversion size may be determined to be suitable for the resolution of the video signal. As a specific example, when a macroblock having a size of 32x32 is used in a video signal having an XGA (eXtended Graphics Array, 1024x768, 4: 3) resolution, the conversion size may be set to 16x16. In this case, the conversion process may be performed using only the 16x16 conversion size for the 32x32 macroblock, or by using at least one conversion size of 32x32, 32x16, 16x32, 16x16, 16x8, 8x16, 8x8, 8x4, 4x8, 4x4 The conversion process can be performed. That is, only one transform size may be used for one macroblock, or a plurality of transform sizes may be used.

In addition, in the present invention, adaptive transform size flag information that can set an appropriate transform size for each block can be defined. The adaptive transform size flag information may be defined as information indicating whether a coding unit is divided into coding units having half sizes in horizontal and vertical directions. Here, the coding unit may mean a picture, slice, macroblock, subblock, transform block, block, pixel, or the like. That is, if the adaptive transform size flag information of the first coding unit is true, the adaptive transform size flag information can be obtained for each second coding unit. At this time, the second coding unit indicates a sub-block included in the first coding unit. In addition, the adaptive transform size flag information may be obtained by the transform unit or the inverse transform unit 300, and may be specifically obtained by the transform size determining unit 310. Further, the adaptive transform size flag information may be obtained by comparing the size of the current transform unit with at least one of a minimum transform size and a maximum transform size. For example, the adaptive transform size flag information may be obtained when the size of the current transform unit is greater than the minimum transform size and less than or equal to the maximum transform size. The size of the transform unit is determined based on the adaptive transform size flag information, and a transform process will be performed accordingly. Alternatively, coding block flag information (coded_block_flag) may be obtained based on the adaptive transform size flag information. Here, the coding block flag information (coded_block_flag) indicates whether the transform unit includes a non-zero transform coefficient level.

For example, if the size of the first coding unit is 128x128, the second coding unit may represent four sub-blocks having a size of 64x64. And, if the adaptive transform size flag information does not exist, the adaptive transform size flag information may be derived more by comparing the transform size of the current block with the minimum transform size. For example, if the transform size of the current block is greater than the minimum transform size, the adaptive transform size flag information will indicate that the coding unit is divided into coding units having half size in the horizontal and vertical directions, and If the transform size is not greater than the minimum transform size, the adaptive transform size flag information will indicate that the adaptive transform size flag information is not split into coding units having a half size in the horizontal and vertical directions. That is, if the transform size of the current block is larger than the minimum transform size, the adaptive transform size flag information is derived as '1', and the adaptive transform size flag information will be checked for each sub-block. And, if the transform size of the current block is not greater than the minimum transform size, the adaptive transform size flag information is derived as '0', so that the transform size of the current block will be applied.

As a specific example, let's say the adaptive transform size flag information is split_transform_unit_flag. Referring to FIG. 2, when the size of the current macroblock (block A) is 128x128, at least one of 128x128, 64x64, 32x32, 16x16, 8x8, and 4x4 transform sizes may be selected according to split_transform_unit_flag. If the split_transform_unit_flag is extracted at the current macroblock (block A) level, and the split_transform_unit_flag represents a value of '0', a 128x128 transform size is applied, and when the split_transform_unit_flag represents a value of '1', the size of 64x64 in the 128x128 current macroblock The split_transform_unit_flag may be obtained for each of the 4 sub-blocks. And, if the split_transform_unit_flag obtained at the level of the sub-block (block B) having a size of 64x64 represents a value of '0', a 64x64 transform size is applied, and when the split_transform_unit_flag represents a value of '1', the size of 32x32 in the 64x64 sub-block is again The split_transform_unit_flag may be obtained for each of four sub-blocks. Similarly, the same process can be applied to the following sub-blocks.

Further, in the present invention, flag information indicating the conversion size can be defined. For example, referring to FIGS. 3A and 3B, one of 16x16, 8x8, and 4x4 transform sizes may be selected according to transform_16x16_size_flag and transform_8x8_size_flag. Here, transform_16x16_size_flag indicates whether a transformation process is performed in 16x16 block units or a smaller block unit, and may be obtained at a macroblock level. And, transform_8x8_size_flag indicates whether the transformation process is performed in units of 8x8 blocks or the transformation process in units of 4x4 blocks, and can be obtained at the sub macroblock level. Referring to FIG. 3A, when transform_16x16_size_flag indicates '1', a transformation process is performed in units of 16x16 blocks, and when '0' is indicated, transform_8x8_size_flag is checked. If transform_8x8_size_flag indicates '1', the transform process is performed in 8x8 block units. If '0' is indicated, a transform process is performed in 4x4 block units.

Similarly, referring to FIG. 3B, when transform_8x8_size_flag represents '0', a transformation process is performed in units of 4x4 blocks, and if '1' is represented, transform_16x16_size_flag is checked. If the transform_16x16_size_flag indicates '1', the transform process is performed in units of 16x16 blocks, and if it indicates '0', the transform process is performed in units of 8x8 blocks.

As another embodiment to which the present invention is applied, when a usable transform size is set in advance according to the size of a macroblock, the decoding apparatus may derive within the decoding apparatus without receiving flag information indicating the transform size. For example, when the size of the macroblock is set to not use the 8x8 transform size when the size of the macroblock is 32x32, if transform_8x8_size_flag represents '0', the transformation process is performed in units of 4x4 blocks, and if transform_8x8_size_flag represents '1', 16x16 The conversion process will be performed in block units. At this time, transform_16x16_size_flag will be derived as '1'. Similarly, when the size of the macroblock is 8x8, if transform_8x8_size_flag represents '0', the transformation process is performed in units of 4x4 blocks, and if transform_8x8_size_flag represents '1', the transformation process will be performed in units of 8x8 blocks. At this time, transform_16x16_size_flag will be derived as '0'.

In another embodiment, if the 8x8 transform size is used only for the intra 8x8 block, only transform_16x16_size_flag is transmitted to identify the intra 8x8 block and the intra 16x16 block, and transform_8x8_size_flag will be derived as '1' in the decoding apparatus. Or, if the 4x4 transform size is used only for the intra 8x8 block, only transform_16x16_size_flag is transmitted to identify the intra 8x8 block and the intra 16x16 block, and transform_8x8_size_flag is derived to '0' E or '1' in the decoding apparatus according to the sub macroblock type. Will be.

As described above, the transform size, macroblock size, flag information indicating the transform size, etc. are only examples, and 32x32, 32x16, 16x32, 16x16, 16x8, 8x16, 8x8, 8x4, 4x8, 4x4 and 32x32 or more All sizes will be applicable. In addition, the combination of two flag information in FIGS. 2, 3A, and 3B corresponds to an embodiment, and may indicate a conversion size by one flag information, and also combines a plurality of flag information to provide various embodiments. You can also set

4 is an embodiment to which the present invention is applied, and is a flowchart illustrating a process of determining a transform method (DCT or KLT) based on a prediction mode of a flag or macroblock indicating a transform type.

2, 3A and 3B, various conversion methods (DCT, KLT, etc.) may be applied when processing a video signal to improve coding efficiency. For example, the encoding apparatus may determine a transformation method to be rate-distortion optimization by applying DCT or KLT to each slice. The determined transform method may be coded and transmitted as a syntax element, and may be defined in a macroblock layer, slice header, or picture parameter set information. As a specific example, the encoding apparatus first applies DCT to the slice, calculates the distortion cost (RD cost), then applies KLT and calculates the distortion cost (RD cost). If the distortion cost (RD cost) of the KLT is smaller than the distortion cost (RD cost) of the DCT, the KLT is selected as a transform method for the slice, and it is coded as a syntax element and transmitted. Here, the syntax element may mean flag information indicating a conversion type.

For example, referring to FIG. 4, first, transformation type flag information indicating a transformation type is acquired (S410), and when the transformation type flag information (transform_type_flag) represents '0', DCT is applied (S420), and the transformation is performed. If the type flag information indicates '1', DCT or KLT may be applied based on the type of the macroblock (S430). If the type of the macroblock indicates an intra block, KLT is applied (S450), and if it is not an intra block, DCT may be applied (S440). The transformed data is subjected to a quantization process (S460). As described above, in the present invention, an appropriate transformation method may be used for each prediction mode based on the prediction mode of the macroblock.

5 is an embodiment to which the present invention is applied, and shows a schematic block diagram for explaining a new Mode Dependent Directional Transform that applies a different transform kernel according to the location of the transform block.

Mode Dependent Directional Transform is a KLT-based transformation method according to the prediction mode used in the coding of intra blocks, and the scanning order is also applied differently for each prediction mode to improve coding efficiency. Speak technology. Referring to FIG. 5, the prediction unit transmits a prediction signal according to each mode, and the transformation unit receives a residual signal indicating a difference between the original signal and the prediction signal, and converts the residual signal to convert the transform coefficients. To create. At this time, the transform unit may apply a separate transform kernel to the residual signal according to each mode. As such, transform coefficients generated according to each prediction mode of the intra block are quantized through a quantization unit. In addition, in the mode-based transform, separate transform kernels for horizontal transform and vertical transform can be used.

6 is an embodiment to which the present invention is applied, and shows a specific example of using a different transform kernel according to the location of the transform block.

As an embodiment of the present invention, the transform size may be smaller than or equal to the size of the prediction block. For example, for 32x32 blocks, the conversion size can be 32x32, 16x16, 8x8, 4x4, for 16x16 blocks, the conversion size can be 16x16, 8x8, 4x4, and for 8x8 blocks, the conversion size is 8x8, 4x4 , And for 4x4 blocks, the transformation size can be only 4x4.

In addition, when 8x8 transform size is applied to a 16x16 block, the same transform kernel can be applied to 4 transform blocks. Alternatively, referring to FIG. 6, different transform kernels may be applied according to the location of each transform block. In this case, you will need 4 conversion kernels.

7 is an embodiment to which the present invention is applied, and shows a flowchart for explaining a process of rearranging a scan order for each transform block.

The scan order means the order for arranging the transform coefficients in a one-dimensional form, and the low frequency coefficient is located in the front and the high frequency coefficient is located in the back. The transform coefficient decoding process may be applied to a 4x4 luminance residual block, an 8x8 luminance residual block, an intra 16x16 prediction mode luminance sample, and a color difference sample. In addition, as described above, if the transform size becomes larger, the target block or sample may be increased accordingly. For example, a transform coefficient decoding process may be applied to a 16x16 luminance residual block, a 32x32 luminance residual block, an intra 32x32 prediction mode luminance sample, and a color difference sample.

As a specific example, in accordance with transform_16x16_size_flag and transform_8x8_size_flag described in FIGS. 2, 3A, and 3B, when the flag information indicates that the transformation process is to be performed in units of 16x16 blocks, a transformation coefficient decoding process is applied to a 16x16 luminance residual block You can. Alternatively, when the flag information indicates that the conversion process is to be performed in units of 8x8 blocks, a conversion coefficient decoding process may be applied to an 8x8 luminance residual block.

In an embodiment of the present invention, when a transform coefficient decoding process is applied to an 8x8 luminance residual block, 64 values in the 8x8 luminance residual block are first set as initial values according to the scan order. Here, the 64 initial values represent conversion coefficients, and the scan order may mean a zigzag scan. In addition, the transform coefficients can be rearranged according to the changed scan order to reduce the number of bits in run-length coding. Here, the changed scan order means that the number of non-zero transform coefficients is calculated and the pixel position having the largest number is set as a priority. The number of non-zero transform coefficients may be calculated and updated every macroblock, or every transform block.

For example, referring to FIG. 7, first, the number of non-zero transform coefficients is counted at each pixel position in the first transform block in the macroblock (S710). In addition, the scan order may be changed by setting a pixel position having the highest number of non-zero transform coefficients as a priority at each pixel position in the first transform block (S720). The number of non-zero transform coefficients may be determined at each pixel position in the second transform block, and the changed scan order of the first transform block may be used as an initialized scan order of the second transform block (S730). Here, the first transform block and the second transform block indicate a divided block in one macroblock, for example, the current macroblock is an 8x8 macroblock, and 4x4 luminance residual blocks in the current macroblock When the transform coefficient decoding process is applied, the upper left 4x4 block is defined as the first transform block, the upper right 4x4 block is the second transform block, the lower left 4x4 block is the third transform block, and the lower right 4x4 block is the fourth transform block. can do. The order of these sub-blocks is not limited to this embodiment. Then, the number of non-zero transform coefficients at each pixel position in the first transform block and the number of non-zero transform coefficients at each pixel position in the second transform block are added, and the pixel position having the largest number is given priority. The scan order of the second transform block may be changed as a priority (S740). As described above, coding efficiency can be improved through the update process of the scan order. A detailed embodiment of this will be described in detail in FIGS. 8A to 8D.

In addition, the update process of the scan order may be performed every macroblock or every transform block or every block. In addition, after the update process of the scan order, initialization of the scan order is performed, flag information indicating this may be defined. For example, if the scan order initialization flag information indicates '1', the scan order is initialized for every macroblock or every transform block, and if it indicates '0', the update process of the scan order is performed again.

8A to 8D are embodiments to which the present invention is applied, and show a specific example for explaining the process of rearranging the scan order for each transform block.

In the present embodiment, when the current macroblock is a 32x32 macroblock and the conversion coefficient decoding process is applied in units of 16x16 blocks in the current macroblock, the upper left 16x16 block is the first block, the upper right 16x16 block is the second block, and the lower block The left 16x16 block can be defined as the third block and the lower right 16x16 block as the fourth block. Here, the first block to the fourth block may mean a transform block. And, the gray block indicates that it contains non-zero transform coefficients, and the number in the gray block indicates the scan order.

Referring to FIG. 8A, when the non-zero transform coefficients are first counted in the first block, it can be seen that non-zero transform coefficients exist in a total of four gray pixels. The initialized scan order follows a zigzag scan order, where the initialized scan order is not limited to a zigzag scan order. When a zigzag scan order is applied based on a gray pixel having a non-zero conversion coefficient, an updated scan order can be obtained as shown in FIG. 8A.

And, referring to FIG. 8B, first, the non-zero transform coefficients in the second block may be counted and summed with the number of non-zero transform coefficients in the first block to finally count the accumulated non-zero transform coefficients. have. In addition, the scan order last updated in the previous first block is used as the initialized scan order of the second block. Likewise, the scan order can be changed based on the pixel position having the largest number of accumulated non-zero transform coefficients.

Then, referring to FIG. 8C, first, the non-zero transform coefficients in the third block are counted, and the number of non-zero transform coefficients accumulated in the second block is finally counted to count the non-zero transform coefficients accumulated in the second block. can do. In addition, the scan order last updated in the previous second block is used as the initialized scan order of the third block. Likewise, the scan order may be changed by setting the pixel position having the largest number among the number of non-zero transform coefficients accumulated as a priority.

Similarly, referring to FIG. 8D, first, a count of non-zero transform coefficients in the fourth block is added, and the number of non-zero transform coefficients accumulated in the third block is finally counted. can do. And, the initialized scan order of the fourth block uses the last updated scan order from the previous third block. Likewise, the scan order may be changed by setting the pixel position having the largest number among the number of non-zero transform coefficients accumulated as a priority.

As described above, coding efficiency can be further improved by updating the scan order for each sub-block. In addition, flag information indicating whether the scan order is re-initialized without initializing the scan order for each slice may be defined in the slice header.

The transform coefficients that are one-dimensionally arranged according to the updated scan order are decoded.

Meanwhile, as described with reference to FIGS. 2, 3A, and 3B, when processing a video signal having a large resolution, the conversion size may be determined to be suitable for the resolution of the video signal. At this time, it is necessary to prevent redundant transmission of information indicating whether a non-zero conversion coefficient level is included. For example, the coding block flag information (coded_block_flag) indicates whether a non-zero transform coefficient level is included, and if the coded block flag information (coded_block_flag) is' 0 ', the non-zero transform coefficient level is not included, and the' 1 'indicates that at least one non-zero transform coefficient level is included. As other flag information, the block pattern information (coded_block_pattern) indicates whether four 8X8 luminance blocks and related color difference blocks include non-zero transform coefficient levels. The block pattern information (coded_block_pattern) is extracted when the prediction mode of the current macroblock in the macroblock layer is not the intra 16X16 prediction mode. Here, the block pattern information is a representative bit flag indicating whether a residual is present, and may indicate whether a residual of 16x16, 8x8 or 4x4 luminance blocks is present. For example, if there is a current macroblock, divide it into 4 equal parts, set the upper left block to the 0th bit, the upper right block to the 1st bit, the lower left block to the 2nd bit, and the lower right block to the 3rd bit. The existence of residual can be expressed. The block pattern information is represented by 6 bits, as described above, the luminance block is represented by 0 to 3rd bit, and the color difference block is represented by 2 bits (AC and DC).

For example, if the size of the current macroblock is 16x16 and the conversion coefficient decoding process is applied in units of 16x16 blocks, the luminance component of the block pattern information (coded_block_pattern) is reduced to 1 bit, and the coding block flag information (coded_block_flag) is transmitted. You may not. As another example, if the size of the current macroblock is 16x16, and a transformation coefficient decoding process is applied in units of 16x8 / 8x16 blocks, the luminance component of the block pattern information (coded_block_pattern) is reduced to 2 bits, and the coding block flag information (coded_block_flag) is It may not be transmitted. As another example, if the size of the current macroblock is 32x32 and the conversion coefficient decoding process is applied in units of 16x16 blocks, the luminance component of the block pattern information (coded_block_pattern) is written in 4 bits, and the coded block flag information (coded_block_flag) is transmitted. You may not.

Meanwhile, the coding block flag information (coded_block_flag) may be obtained based on adaptive transform size flag information. The adaptive transform size flag information has been described in detail in FIG. 2. The adaptive transform size flag information may be obtained by comparing the size of the current transform unit with at least one of a minimum transform size and a maximum transform size. When the adaptive transform size flag information indicates that the coding unit is divided into coding units having half sizes in the horizontal and vertical directions, the size of the transform unit is determined, and a conversion process will be performed accordingly. Alternatively, if the adaptive transform size flag information indicates that the coding unit is not divided into coding units having half sizes in the horizontal and vertical directions, the coding block flag information (coded_block_flag) may be obtained.

As described above, the decoding / encoding device to which the present invention is applied is provided in a multimedia broadcast transmission / reception device such as Digital Multimedia Broadcasting (DMB), and can be used to decode video signals and data signals. In addition, the multimedia broadcast transmission / reception device may include a mobile communication terminal.

In addition, the decoding / encoding method to which the present invention is applied may be produced as a program for execution on a computer and stored in a computer-readable recording medium, and computer-readable multimedia data having a data structure according to the present invention It can be stored in a recording medium. The computer-readable recording medium includes all kinds of storage devices in which data readable by a computer system is stored. Examples of the computer-readable recording medium include ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage device. In addition, the bitstream generated by the encoding method may be stored in a computer readable recording medium or transmitted using a wired / wireless communication network.

The preferred embodiments of the present invention described above have been disclosed for purposes of illustration, and those skilled in the art can improve and change various other embodiments within the technical spirit and the technical scope of the present invention disclosed in the appended claims. , Replacement or addition will be possible.

Claims (9)

A method for decoding a video signal in a decoding device,
Obtaining a prediction value of a current block using inter prediction or intra prediction according to prediction mode information;
Obtaining residual information of the current block by performing a first transform or a second transform; And
Restoring the current block based on the prediction value and the residual information,
The step of obtaining the residual information,
Obtaining conversion type information indicating a conversion type,
Performing the first transform when the transform type information indicates a value of 0 or intra prediction is not used in the current block;
Performing the second transform when the transform type information indicates a value of 1 and intra prediction is used for the current block,
Methods of decoding video signals. The method according to claim 1,
The first transform or the second transform has a transform kernel size of 4x4, 8x8, or 16x16. The method according to claim 1,
The first transform corresponds to a discrete cosine transform (DCT), and the second transform corresponds to a Karhuhen-Loeve Transform (KLT). A decoding device for decoding a video signal, the decoding device comprising:
According to prediction mode information, a prediction value of a current block is obtained using inter prediction or intra prediction, and residual information of the current block is obtained by performing a first transformation or a second transformation, and the prediction value and the residual information are It is configured to restore the current block on the basis,
Acquiring the residual information,
Obtaining conversion type information indicating a conversion type,
Performing the first transform when the transform type information indicates a value of 0 or intra prediction is not used in the current block;
Performing the second transform when the transform type information indicates a value of 1 and intra prediction is used for the current block,
Device for decoding video signals. The method according to claim 4,
The first transform or the second transform has a transform kernel size of 4x4, 8x8, or 16x16, the apparatus for decoding a video signal. The method according to claim 4,
The first transform corresponds to a discrete cosine transform (DCT), and the second transform corresponds to a Karhuhen-Loeve Transform (KLT). A method of encoding a video signal in an encoding device, the method comprising:
Obtaining a prediction value of a current block using inter prediction or intra prediction;
Obtaining residual information of the current block based on the predicted value; And
And encoding the current block by performing a first transform or a second transform based on the residual information,
Encoding the current block,
Obtaining conversion type information indicating a conversion type,
Performing the first transform when the transform type information indicates a value of 0 or intra prediction is not used in the current block;
Performing the second transform when the transform type information indicates a value of 1 and intra prediction is used for the current block,
Video signal encoding method. A computer-readable recording medium storing a video signal encoded by the method of claim 7. An encoding device for encoding a video signal, the encoding device comprising:
The inter prediction or intra prediction is used to obtain the prediction value of the current block, the residual information of the current block is obtained based on the prediction value, and the first transformation or the second transformation is performed based on the residual information to perform the first transformation or the second transformation. Configured to encode the current block,
Encoding the current block,
Obtaining conversion type information indicating a conversion type,
Performing the first transform when the transform type information indicates a value of 0 or intra prediction is not used in the current block;
Performing the second transform when the transform type information indicates a value of 1 and intra prediction is used for the current block,
Device for encoding video signals.

Images