WO2019190284A1 - 비디오 압축을 위한 변환 커널의 저복잡도 연산을 수행하는 방법 및 장치 - Google Patents

비디오 압축을 위한 변환 커널의 저복잡도 연산을 수행하는 방법 및 장치 Download PDF

Info

Publication number
WO2019190284A1
WO2019190284A1 PCT/KR2019/003743 KR2019003743W WO2019190284A1 WO 2019190284 A1 WO2019190284 A1 WO 2019190284A1 KR 2019003743 W KR2019003743 W KR 2019003743W WO 2019190284 A1 WO2019190284 A1 WO 2019190284A1
Authority
WO
WIPO (PCT)
Prior art keywords
transform
inverse
present
horizontal
index
Prior art date
Application number
PCT/KR2019/003743
Other languages
English (en)
French (fr)
Inventor
구문모
살레히파메흐디
김승환
임재현
Original Assignee
엘지전자 주식회사
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 엘지전자 주식회사 filed Critical 엘지전자 주식회사
Priority to US17/042,722 priority Critical patent/US20210021871A1/en
Publication of WO2019190284A1 publication Critical patent/WO2019190284A1/ko

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/60Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
    • H04N19/61Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding in combination with predictive coding
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/70Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by syntax aspects related to video coding, e.g. related to compression standards
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/12Selection from among a plurality of transforms or standards, e.g. selection between discrete cosine transform [DCT] and sub-band transform or selection between H.263 and H.264
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/134Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
    • H04N19/146Data rate or code amount at the encoder output
    • H04N19/147Data rate or code amount at the encoder output according to rate distortion criteria
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/134Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
    • H04N19/157Assigned coding mode, i.e. the coding mode being predefined or preselected to be further used for selection of another element or parameter
    • H04N19/159Prediction type, e.g. intra-frame, inter-frame or bidirectional frame prediction
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/169Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
    • H04N19/17Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
    • H04N19/176Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a block, e.g. a macroblock
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/42Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation
    • H04N19/439Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation using cascaded computational arrangements for performing a single operation, e.g. filtering
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/60Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
    • H04N19/625Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding using discrete cosine transform [DCT]

Definitions

  • the present invention relates to a method and an apparatus for processing a video signal, and more particularly to memory usage and computational complexity of a discrete sine transform-4 (DST4) and a discrete cosine transform-4 (DCT4) in a transform kernel for video compression. To reduce technology.
  • DST4 discrete sine transform-4
  • DCT4 discrete cosine transform-4
  • Next generation video content will be characterized by high spatial resolution, high frame rate and high dimensionality of scene representation. Processing such content will result in a tremendous increase in terms of memory storage, memory access rate, and processing power.
  • This paper proposes a low complexity algorithm for transform kernel for video compression.
  • the present invention provides a DST4 (Discrete Sine Transform-) among transform kernels for video compression. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • DCT4 Discrete Cosine Transform-4
  • the present invention intends to propose an encoder / decoder structure for reflecting a new transform design.
  • the present invention provides a method of reducing complexity and improving coding efficiency through a new transform design.
  • DST4 Discrete Sine Transform-4
  • DCT4 Discrete Cosine
  • Transform-4) is provided as a forward DCT2.
  • the present invention provides a method of performing DST4 and DCT4 as inverse DCT2.
  • the present invention provides a method of applying DST4 and DCT4 to a transform configuration group to which MTS (Multiple Transform Selection) is applied.
  • MTS Multiple Transform Selection
  • DST4 Discrete Sine Transform-4
  • DCT4 Discrete Cosine
  • the present invention can perform more efficient coding by applying DST4 and DCT4 to a transform configuration group to which MTS (Multiple Transform Selection) is applied.
  • MTS Multiple Transform Selection
  • the new low complexity algorithm can reduce computational complexity and improve coding efficiency. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • FIG. 1 is an embodiment to which the present invention is applied and shows a schematic block diagram of an encoder in which encoding of a video signal is performed.
  • FIG. 2 is a schematic block diagram of a decoder in which decoding of a video signal is performed according to an embodiment to which the present invention is applied.
  • Figure 3 is an embodiment to which the present invention can be applied,
  • Figure 3A is QT (QuadTree, referred to as' QT ')
  • Figure 3B is BT (Binary Tree, referred to as' BT)
  • Figure 3C is TT 3D is a diagram for describing block division structures by an AT (Asymmetric Tree, hereinafter called 'AT').
  • FIG. 4 is a schematic block diagram of a transform and quantization unit 120/130 and an inverse quantization and inverse transform unit 140/150 in an encoder according to an embodiment to which the present invention is applied.
  • FIG. 5 is a schematic block diagram of an inverse quantization and inverse transform unit 220/230 in a decoder according to an embodiment to which the present invention is applied.
  • FIG. 6 is a table illustrating a transform configuration group to which an MTS (Multiple Transform Selecti) is applied as an embodiment to which the present invention is applied.
  • MTS Multiple Transform Selecti
  • FIG. 7 is an embodiment to which the present invention is applied and is a flowchart illustrating an encoding process in which MTS (Multiple Transform Selection) is performed.
  • MTS Multiple Transform Selection
  • FIG. 8 is a flowchart illustrating a decoding process in which multiple transform selection (MTS) is performed, according to an embodiment to which the present invention is applied.
  • MTS multiple transform selection
  • FIG. 9 is a flowchart illustrating a process of encoding an MTS flag and an MTS index according to an embodiment to which the present invention is applied.
  • FIG. 10 illustrates an embodiment to which the present invention is applied and includes an MTS flag and an MTS index. 2019/190284 1 »(1 ⁇ 1 ⁇ 2019/003743
  • FIG. 4 is a flowchart to explain the decoding process of applying a horizontal transform or a vertical transform to a row or a column based on the above.
  • FIG. 11 is a schematic block diagram of an inverse transform unit in a decoder according to an embodiment to which the present invention is applied.
  • FIG. 12 is a block diagram for performing inverse transformation based on transformation-related parameters according to an embodiment to which the present invention is applied.
  • FIG. 13 is an embodiment to which the present invention is applied and shows a flowchart of performing inverse transformation based on transformation related parameters.
  • DST4 Discrete Sine Transform-4
  • DCT4 Discrete Cosine Transform-4
  • FIG. 15 illustrates a decoding flowchart of performing Discrete Sine Transform-4 (DST4) and Discrete Cosine Transform-4 (DCT4) with a forward DCT2 or an inverse DCT2 as an embodiment to which the present invention is applied. .
  • DST4 Discrete Sine Transform-4
  • DCT4 Discrete Cosine Transform-4
  • FIG. 6 is an embodiment to which the present invention is applied and shows diagonal elements for a pair of transform block size (N) and right shift amount () when DST4 and DCT4 are performed as forward DCT2.
  • FIG. 17 illustrates sets of DCT2 kernel coefficients that may be applied to DST4 or DCT4 as an embodiment to which the present invention is applied.
  • FIG. 18 illustrates a forward DCT2 matrix generated from a set of DCT2 kernel coefficients that may be applied to DST4 or DCT4 as an embodiment to which the present invention is applied.
  • 20 shows code execution of an output stage for DCT4 as an embodiment to which the present invention is applied.
  • FIG. 21 shows the configuration of parameter sets and multiplication coefficients for 0814 and 1X74 when 0814 and 1: 14 are performed in a forward 0012 as an embodiment to which the present invention is applied.
  • FIG. 22 shows code execution of a preprocessing step for 0014 as an embodiment to which the present invention is applied.
  • FIG. 23 shows code execution of a preprocessing step for 0814 as an embodiment to which the present invention is applied.
  • FIG. 25 shows the configuration of parameter sets and multiplication coefficients for 14 and 1X74 when executed with 0814 and 00X41-reverse 0012 as an embodiment to which the present invention is applied.
  • FIG. 26 illustrates eight mappings to intra prediction residuals according to an embodiment to which the present invention is applied.
  • FIG. 27 is a diagram illustrating 1 18 mapping of inter prediction residuals according to an embodiment to which the present invention is applied.
  • a method of reconstructing a video signal based on low complexity conversion comprising: obtaining a transform index of a current block from the video signal, wherein the transform index is composed of a combination of 0814 and / or DC ⁇ 4.
  • any one of a plurality of transform combinations Deriving a transform combination corresponding to the transform index, wherein the transform combination comprises a horizontal transform and a vertical transform, wherein the horizontal transform and the vertical transform correspond to either 0814 or 0014; Performing inverse transform in a vertical direction with respect to the current block using the 0314; Performing inverse transform in a horizontal direction with respect to the current block using the 1 ⁇ ⁇ 4; And reconstructing the video signal using the inversely transformed current block.
  • the 0814 and / or the 1X714 is characterized in that the execution using the forward 0072 or the reverse direction.
  • 0814 and / or ⁇ ) ( 4) is characterized by applying a post-processing matrix ⁇ and a pre-processing matrix to the forward DCT2 or the reverse 1X72.
  • the transformation combination (horizontal transformation, vertical transformation), ⁇ ⁇ 4, 0814), ( 0 ⁇ 4, 0814), (0814, 0014) and (1) ⁇ the 4, 1) (4) It is characterized by including.
  • the transform combination corresponds to the transform indices 0, 1, 2, and 3.
  • the transform combination corresponds to the transform index 3, 2, 1, 0.
  • the present invention provides an apparatus for reconstructing a video signal based on low complexity conversion execution, the apparatus comprising: a parser for obtaining a transform index of a current block from the video signal, wherein the transform index is a combination of 0814 and / or DC ⁇ 4 Corresponds to any one of the configured plurality of transform combinations; A transform unit for deriving a transform combination corresponding to the transform index, performing a reverse transformation in the vertical direction with respect to the current block using 4, and performing a reverse transformation in the horizontal direction with respect to the current block using 1 ⁇ 74, Wherein the transform combination comprises a horizontal transform and a vertical transform, wherein the horizontal transform and the vertical transform correspond to either 0814 or DCT4; And a reconstruction unit for reconstructing the video signal using the inversely transformed current block.
  • terms used in the present invention may be replaced for more appropriate interpretation when there are general terms selected to describe the invention or other terms having similar meanings.
  • signals, data, samples, pictures, frames, blocks, etc. may be appropriately replaced and interpreted in each coding process.
  • partitioning, decomposition, splitting, and division may be appropriately replaced and interpreted in each coding process.
  • MTS Multiple Transform Selection
  • AMT Adaptive Multiple Transform
  • EMT Explicit Multiple Transform
  • mtsjdx may be expressed as AMTJdx, EMT_idx, tu_mts_idx, AMT_TU_idx, EMT_TU_idx, transform index, or transform combination index, and the present invention. It is not limited to this expression. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • FIG. 1 is a schematic block diagram of an encoder in which encoding of a video signal is performed according to an embodiment to which the present invention is applied.
  • the encoder 100 includes an image splitter 1 10, a transform unit 120, a quantization unit 130, an inverse quantization unit 140, an inverse transform unit 150, a filtering unit 160, It may include a decoded picture buffer (DPB) 170, an inter predictor 180, an intra predictor 185, and an entropy encoder 190.
  • DPB decoded picture buffer
  • the image divider 1 10 may divide an input image (or a picture or a frame) input to the encoder 100 into one or more processing units.
  • the processing unit may be a Coding Tree Unit (CTU), a Coding Unit (CU), a Prediction Unit (PU), or a Transform Unit (TU).
  • CTU Coding Tree Unit
  • CU Coding Unit
  • PU Prediction Unit
  • TU Transform Unit
  • the terms are only used for the convenience of description of the present invention, the present invention is not limited to the definition of the terms.
  • the term coding unit is used as a unit used in a process of encoding or decoding a video signal, but the present invention is not limited thereto and may be appropriately interpreted according to the present disclosure.
  • the encoder 100 may generate a residual signal by subtracting a prediction signal output from the inter predictor 180 or the intra predictor 185 from the input image signal, and generate the residual signal. Is transmitted to the converter 120.
  • the transform unit 120 may generate a transform coefficient by applying a transform technique to the residual signal.
  • the conversion process consists of a quadtree square block, a binarytree structure, a ternary tree, or a non-symbol. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the transform unit 120 may perform a transform based on a plurality of transforms (or transform combinations), and this transform scheme may be referred to as MTS (Multiple Transform Selection).
  • MTS Multiple Transform Selection
  • the MTS may be called an AMT (Adaptive Multiple Transform) or an EMT (Enhanced Multiple Transform).
  • the MTS may refer to a transform scheme performed based on a transform (or transform combinations) adaptively selected from a plurality of transforms (or transform combinations).
  • the plurality of transforms may include the transforms (or transform combinations) described with reference to FIGS. 6 and 26 to 27.
  • the transform or transform type may be expressed as, for example, DCT-Type 2, DCT-II, DCT2, or DCT2.
  • the converter 120 may perform the following embodiments.
  • the present invention provides a method of performing Discrete Sine Transform-4 (DST4) and Discrete Cosine Transform-4 (DCT4) as forward DCT2.
  • DST4 Discrete Sine Transform-4
  • DCT4 Discrete Cosine Transform-4
  • the present invention provides a method for performing DST4 and DCT4 with inverse DCT2.
  • the present invention provides a method of applying DST4 and DCT4 to a transform configuration group to which MTS (Multiple Transform Selection) is applied.
  • MTS Multiple Transform Selection
  • the quantization unit 130 may quantize the transform coefficients and transmit them to the entropy encoding unit 190, and the entropy encoding unit 190 may entropy-code the quantized signal and output the bitstream.
  • the converter 120 and the quantization unit 130 are described as separate functional units, the present invention is not limited thereto and may be combined into one functional unit. In the case of the inverse quantization unit 140 and the inverse transform unit 150, the same may be combined into one functional unit.
  • the quantized signal output from the quantization unit 130 may be used, for example, to generate a signal.
  • the quantized signal may recover the residual signal by applying inverse quantization and inverse transformation through inverse quantization unit 140 and inverse transform unit 150 in a loop.
  • a reconstructed signal may be generated by adding the reconstructed residual signal to a prediction signal output from the inter predictor 180 or the intra predictor 185.
  • deterioration of the block boundary may occur due to the quantization error generated in the compression process as described above. This phenomenon is called blocking artifacts, which is one of the important factors in evaluating image quality.
  • a filtering process may be performed. Through this filtering process, the image quality can be improved by removing the blocking degradation and reducing the error of the current picture.
  • the filtering unit 160 applies filtering to the reconstruction signal and outputs it to the reproduction apparatus or transmits the decoded picture buffer to the decoded picture buffer 170.
  • the filtered signal transmitted to the decoded picture buffer 170 may be used as the reference picture in the inter predictor 180. like this, 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the decoded picture buffer 170 may store the filtered picture for use as a reference picture in the inter prediction unit 180.
  • the inter prediction unit 180 performs temporal prediction and / or spatial prediction to remove temporal redundancy and / or spatial redundancy with reference to a reconstructed picture.
  • the reference picture used to perform the prediction is a transformed signal that has been quantized and dequantized in units of blocks at the time of encoding / decoding in the previous time, blocking artifacts or ringing artifacts may exist. have.
  • the inter prediction unit 180 may interpolate the signals between pixels in sub-pixel units by applying a lowpass filter to solve the performance degradation due to discontinuity or quantization of such signals.
  • the subpixel refers to a virtual pixel generated by applying an interpolation filter
  • the integer pixel refers to an actual pixel existing in the reconstructed picture.
  • the interpolation method linear interpolation, bi linear interpolation, and Wiener filter may be applied.
  • the interpolation filter may be applied to the reconstructed picture to improve the accuracy of the example.
  • the inter prediction unit 180 generates an interpolation pixel by applying an interpolation filter to an integer pixel, and uses an interpolated block composed of interpolated pixels as a prediction block. Yes you can.
  • the intra prediction unit 185 is the block of the current encoding 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the intra prediction unit 185 may perform the following process to perform intra prediction. First, a reference sample necessary for generating a prediction signal may be prepared. Then, a prediction signal may be generated using the prepared reference sample. Then, the prediction mode is encoded. In this case, the reference sample may be prepared through reference sample padding and / or reference sample filtering. Since the reference sample has been predicted and reconstructed, there may be a quantization error. Accordingly, the reference sample filtering process may be performed for each prediction mode used for intra prediction to reduce such an error.
  • a prediction signal generated through the inter predictor 180 or the intra predictor 185 may be used to generate a reconstruction signal or to generate a residual signal.
  • 2 is a schematic block diagram of a decoder in which decoding of a video signal is performed as an embodiment to which the present invention is applied.
  • the decoder 200 includes a parser (not shown), an entropy decoder 210, an inverse quantizer 220, an inverse transformer 230, a filter 240, and a decoded picture buffer (DPB). It may include a decoded picture buffer unit) 250, an inter predictor 260, and an intra predictor 265.
  • the reconstructed video signal output through the decoder 200 may be reproduced through the reproducing apparatus.
  • the decoder 200 may receive a signal output from the encoder 100 of FIG. 1, and the received signal may be entropy decoded through the entropy decoding unit 210. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the inverse quantization unit 220 obtains a transform coefficient from the entropy decoded signal using the quantization step size information.
  • the inverse transform unit 230 inversely transforms the transform coefficient to obtain a residual signal.
  • the present invention provides a method of configuring a transform combination for each transform configuration group divided by at least one of a prediction mode, a block size, or a block shape.
  • 230 may perform an inverse transform based on the transform combination constructed by the present invention.
  • the embodiments described herein may be applied.
  • the inverse transform unit 230 may perform the following embodiments.
  • the present invention provides a method of performing Discrete Sine Transform-4 (DST4) and Discrete Cosine Transform-4 (DCT4) as forward DCT2.
  • DST4 Discrete Sine Transform-4
  • DCT4 Discrete Cosine Transform-4
  • the present invention provides a method of performing DST4 and DCT4 as inverse DCT2.
  • the present invention provides a method of applying DST4 and DCT4 to a transform configuration group to which MTS (Multiple Transform Selection) is applied.
  • MTS Multiple Transform Selection
  • a method of reconstructing a video signal based on low complexity conversion execution comprising: obtaining a transform index of a current block from the video signal, wherein the transform index comprises a plurality of combinations of DST4 and / or DCT4; Corresponds to any one of the transform combinations; Deriving a transform combination corresponding to the transform index, wherein the transform combination is comprised of a horizontal transform and a vertical transform, wherein the horizontal transform and the vertical transform are the DST4 or the 2019/190284 1 »(: 1/10 ⁇ 019/003743
  • the 14 and / or 1) (the 14 is executed using the forward £> (the 2 or the reverse 1) 0:72).
  • the 0814 and / or 1X714 is applied to the post-processing matrix and pre-processing matrix in the forward 0012 or the reverse 1) 12
  • the inverse transform of 14 is applied for each column when the vertical transform is 14, and the inverse transform of 1X ⁇ 4 is applied for each row when the horizontal transform is 1X ⁇ 4.
  • the transform combination (horizontal transform, vertical transform) includes ⁇ 4, 0 ⁇ 4), (0014, 0814), (0814, 0014), and (0 4, 1) 14 do.
  • the transform combination corresponds to the transform indices 0, 1, 2, and 3. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the transform combination corresponds to the transform index 3, 2, 1, 0.
  • the inverse quantization unit 220 and the inverse transform unit 230 are described as separate functional units, the present invention is not limited thereto and may be combined into one functional unit.
  • a reconstructed signal is generated by adding the obtained residual signal to a prediction signal output from the inter predictor 260 or the intra predictor 265.
  • the filtering unit 240 applies filtering to the reconstructed signal and outputs the filtering to the reproducing apparatus or transmits it to the decoded picture buffer unit 250.
  • the filtered signal transmitted to the decoded picture buffer unit 250 may be used as the reference picture in the inter predictor 260.
  • FIG. 3A is a QT (QuadTree, hereinafter referred to as QT)
  • FIG. 3B is a Binary Tree (BT)
  • FIG. 3C is a TT (Ternary Tree).
  • 3D is a diagram for describing block division structures by an AT (Asymmetric Tree, hereinafter called 'AT').
  • one block may be divided on a QT (QuadTree) basis.
  • QT QuadTree
  • one sub-block divided by QT uses QT. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • a leaf block that is no longer QT split may be split by at least one of a binary tree (BT), a primary tree (TT), and an asymmetric tree (AT).
  • BT may have two types of divisions: horizontal BT (2NxN, 2NxN) and vertical BT (Nx2N, Nx2N).
  • the TT may have two types of divisions, horizontal TT (2Nxl / 2N, 2NxN, 2Nxl / 2N) and vertical TT (l / 2Nx2N, Nx2N, l / 2Nx2N).
  • AT is horizontal-up AT (2Nxl / 2N, 2Nx3 / 2N), horizontal-down AT (2Nx3 / 2N, 2Nxl / 2N), vertical-left AT (l / 2Nx2N, 3 / 2Nx2N), vertical-right AT (3 / 2Nx2N, l / 2Nx2N) can be divided into four types. Each BT, TT, or AT may be further recursively divided using BT, TT, and AT.
  • Block A may be divided into four sub-blocks (AO, Al, A2, A3) by QT.
  • the sub-block time may be divided into four sub-blocks B0, B1, B2, and B3 by Ah.
  • 3B shows an example of BT partitioning.
  • B2 which is no longer divided by the group, can be divided into vertical BT (CO, Cl) or horizontal BT (DO, Dl).
  • Block B3 which is no longer split by QT, may be split into vertical TT (CO, Cl, C2) or horizontal TT (DO, Dl, D2). Like the block Cl, each subblock may be further recursively divided into the form of horizontal TT (E0, El, E2) or vertical TT (F0, F1, F2).
  • Block B3 which is no longer split by Ah, can be split into vertical AT (CO, Cl) or horizontal AT (DO, Dl). 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • Each block can be further recursively divided into the form of horizontal AT (E0, El) or vertical TT (F0, F1).
  • BT, TT, AT partitions can be used together to divide.
  • a subblock divided by BT may be divided by TT or AT.
  • the sub-block divided by TT can be divided by BT or AT.
  • a subblock divided by AT may be divided by BT or TT.
  • each subblock may be divided into vertical BTs, or after vertical BT partitioning, each subblock may be split into horizontal BTs. Are different, but the final split is the same.
  • FIG. 4 is a schematic block diagram of a transform and quantization unit 120/130 and an inverse quantization and inverse transform unit 140/150 in an encoder, and FIG. 5. Shows a schematic block diagram of the inverse quantization and inverse transform units 220/230 in the decoder.
  • the transform and quantization unit 120/130 is a primary transform unit (primary). 2019/190284 1 »(: 1/10 ⁇ 019/003743
  • the inverse quantization and inverse transform unit 140/150 may include an inverse quantization unit 140, an inverse secondary transform unit 151, and an inverse primary transform unit 152. Can be.
  • the inverse quantization unit 220/230 includes an inverse quantization unit 220, an inverse secondary transform unit 231, and an inverse primary transform unit. 232 may include.
  • the transformation may be performed through a plurality of steps when performing the transformation.
  • two stages of a primary transform and a secondary transform may be applied as shown in FIG. 4, or more transformation steps may be used according to an algorithm.
  • the primary transform may be referred to as a core transform.
  • the primary transform unit 121 may apply a primary transform to the residual signal, wherein the primary transform is set in the table at the encoder and / or the decoder. Can be
  • DCT2 Discrete Cosine Transform type 2
  • DST7 Discrete Sine Transform-type 7
  • DST7 may be applied to a 4x4 block in the intra prediction mode.
  • the secondary transform unit 122 may apply a secondary transform on the primary transformed signal, where the secondary transform is an encoder and / or a decoder. Can be defined as in the table.
  • the secondary transform may be conditionally applied to a non-separable secondary transform (hereinafter, referred to as 'NSST').
  • 'NSST' non-separable secondary transform
  • the NSST is applied only to an intra prediction block and may have a transform set applicable to each prediction mode group.
  • the prediction mode group may be set based on symmetry with respect to the prediction direction. For example, since the prediction mode 52 and the prediction mode 16 are symmetric with respect to the prediction mode 34 (diagonal direction), the same transform set may be applied by forming one group. In this case, when the transform for the prediction mode 52 is applied, the input data is transposed and then applied, since the prediction set 16 and the transform set are the same.
  • each plane since there is no symmetry of the direction, each plane has its own transform set, and the transform set may be composed of two transforms. For the remaining directional mode, three transforms may be configured per transform set.
  • the NSST may not be applied to the entire first transformed block but may be applied only to a top-left 8 ⁇ 8 region. For example, if the block size is 8x8 or more, 8x8 NSST is applied, and if the block size is less than 8x8, 4x4 NSST7> is applied.
  • 4x4 NSST may be applied. () 2019/190284 1 »(: 1/10 ⁇ 019/003743
  • the quantization unit 130 may perform quantization on the quadratic transformed signal.
  • the inverse quantization and inverse transform unit 140/150 performs the above-described process in reverse, and redundant description thereof will be omitted.
  • 5 shows a schematic block diagram of inverse quantization and inverse transform units 220/230 in a decoder.
  • the inverse quantization unit 220/230 includes an inverse quantization unit 220, an inverse secondary transform unit 231, and an inverse primary transform unit. 232 may include.
  • the inverse quantization unit 220 obtains a transform coefficient from the entropy decoded signal using the quantization step size information.
  • the inverse secondary transform unit 231 performs inverse secondary transform on the transform coefficients.
  • the inverse secondary transform indicates an inverse transform of the secondary transform described with reference to FIG. 4.
  • the inverse primary transform unit 232 performs an inverse first transform on an inverse secondary transformed signal (or block) and obtains a residual signal.
  • the inverse first transform represents an inverse transform of the primary transform described with reference to FIG. 4.
  • the present invention provides a method of configuring a transform combination for each transform configuration group divided by at least one of a prediction mode, a block size, or a block shape.
  • An inverse primary transform unit 232 may be used for transform combinations constructed by the present invention. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • FIG. 6 is a table illustrating a transform configuration group to which Multiple Transform Selection (MTS) is applied as an embodiment to which the present invention is applied.
  • MTS Multiple Transform Selection
  • the j th transform combination candidate for the transform setting group Gi is represented by a pair as shown in Equation 1 below.
  • H (Gi, j) indicates a horizontal transform for the j th candidate
  • V (Gi, j) indicates a vertical transform for the j th candidate.
  • H (G 3 , 2) DST7
  • V (G 3J 2)-DCT8 can be expressed.
  • the value assigned to H (Gi, j) or V (Gi, j) may be a nominal value to distinguish between transformations, as in the example above, or may be an index value indicating the transformation, or It may be a 2D matrix for the transformation.
  • 2D matrix values for DCT and DST may be expressed as in Equations 2 to 3 below.
  • DCT type 2 DCT type 8: Cf v
  • DST type 7 S N V "
  • DST type 4 S N ! V
  • transform configuration groups may be determined based on a prediction mode, and the number of groups may be six (G0 to G5).
  • G0 to G4 correspond to a case where intra prediction is applied
  • G5 represents transform combinations (or transform sets or transform combination sets) applied to a residual block generated by inter prediction.
  • One transform combination is a horizontal transform (or row transform) applied to the rows of the corresponding 2D block and a vertical transform (or column) applied to the columns. It can consist of a column (column transform).
  • all of the transform configuration groups may have four transform combination candidates.
  • the four transform combination candidates may be selected or determined through a transform combination index of 0 to 3, and may encode and transmit the transform combination index from an encoder to a decoder.
  • the residual data (or residual signal) obtained through intra prediction may have different statistical characteristics according to the intra prediction mode.
  • a plurality of transform combinations may be applied to each transform setting group divided in each intra prediction mode column.
  • the plurality of transformation combinations may be composed of four (row direction transformation, column direction transformation) combinations.
  • DST-7 and DCT-5 can be applied in both the row (horizontal) direction and the column (vertical) direction, so a total of four combinations are possible.
  • transform combination index may be referred to as an MTS index and may be expressed as mts_idx.
  • the transformation can be adaptively performed by defining the MTS flag for each coding unit.
  • the MTS flag is 0, DCT2 is applied to both the row direction and the column direction. If the MTS flag is 1, one of four combinations may be selected or determined through the MTS index.
  • the MTS flag when the MTS flag is 1, if the number of non-zero transform coefficients for one transform unit is not greater than a threshold value, the DST ⁇ for both the row direction and the column direction is not applied without applying the transform kernels of FIG. 6. 7 can be applied.
  • the threshold may be set to 2, which may be set differently based on the block size or the size of the transform unit. This also applies to other embodiments of the specification. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the amount of additional information transmission may be reduced by applying DST-7 without parsing an MTS index.
  • the MTS index when the MTS flag is 1, when the number of non-zero transform coefficients for one transform unit is greater than the threshold, the MTS index may be parsed, and the horizontal transform and the vertical transform may be determined based on the MTS index.
  • the MTS may be applied only when the width and height of the transform unit are both 32 or less.
  • FIG. 6 may be preset through off-line training.
  • the MTS index may be defined as one index that can simultaneously indicate a combination of a horizontal transform and a vertical transform.
  • the MTS index may separately define a horizontal transform index and a vertical transform index.
  • the MTS flag or the MTS index may be defined at at least one level of a sequence, picture, slice, block, coding unit, transform unit, or prediction unit.
  • the MTS flag or the MTS index may be defined at at least one level of a sequence parameter set (SPS) or a conversion unit.
  • SPS sequence parameter set
  • 7 is an embodiment to which the present invention is applied and is a flowchart illustrating an encoding process in which MTS (Multiple Transform Selection) is performed. 2019/190284 1 »(: 1/10 ⁇ 019/003743
  • it may consist of a mixture of separable and non-separable transforms.
  • a non-separable transform 0 the selection of transforms by row / column or selection by horizontal / vertical direction is unnecessary and a separable transform is required.
  • the transformation combinations of FIG. 6 above can be used only when.
  • the schemes proposed in this specification may be applied regardless of a primary transform or a secondary transform. That is, there is no restriction that it should be applied to either one, and both can be applied.
  • the primary transform may mean a transform for transforming a residual block first
  • the secondary transform is applied to a block generated as a result of the primary transform. This may mean a transformation for applying a transformation to the.
  • the encoder may determine a transform setting group corresponding to the current block (S710).
  • the conversion setting group may mean the conversion setting group of FIG. 6, but the present invention is not limited thereto and may be configured with other conversion combinations.
  • the encoder may perform transform on candidate transform combinations available in the transform configuration group (S720).
  • the encoder has the lowest RD (Rate Distortion) cost. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • a combination of transformations may be determined or selected (S730).
  • the encoder may encode a transform combination index corresponding to the selected transform combination (S740).
  • S740 Multiple Transform Selection
  • 8 is a flowchart illustrating a decoding process in which MTS (Multiple Transform Selection) is performed according to an embodiment to which the present invention is applied.
  • the decoder may determine a transform setting group for the current block (S810).
  • the decoder may parse (or obtain) a transform combination index from a video signal, where the transform combination index may correspond to any one of a plurality of transform combinations in the transform configuration group (S820).
  • the transform configuration group may include Discrete Sine Transform type 7 (DST7) and Discrete Cosine Transform type 8 (DCT8).
  • the transform combination index may be referred to as an MTS index.
  • the transform setting group may be set based on at least one of a prediction mode, a block size, or a block shape of the current block.
  • the decoder may derive a transform combination corresponding to the transform combination index (S830).
  • the transform combination may include a horizontal transform and a vertical transform, and may include at least one of the DST-7 and the DCT-8.
  • the transform combination may mean the transform combination described with reference to FIG. 6, but the present invention is not limited thereto. That is, the configuration by other conversion combinations according to another embodiment of the present specification is also possible.
  • the decoder may perform an inverse transform on the current block based on the transform combination (SM0).
  • the transformation combination converts rows (horizontal) and columns (vertical) 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the present invention is not limited thereto, and in the case of applying the reverse or non-separated transform, the non-separated transform may be applied immediately.
  • the inverse transform of 0 -7 or 0 may be applied for each row after applying the column for each column.
  • the vertical transformation or the horizontal transformation a different transformation may be applied to each row and / or each column.
  • the transform combination index may be obtained based on a yaw flag indicating whether 1 18 is performed. That is, the transform combination index is assigned to the MTS flag. May be obtained if performed.
  • the decoder may determine whether the number of non-zero transform coefficients is greater than a threshold. In this case, the transform combination index may be obtained when the number of non-zero transform coefficients is greater than a threshold.
  • the MTS flag or the An index may be defined at at least one level of a sequence, picture, slice, block, coding unit, transform unit, or prediction unit.
  • the inverse transform may be applied only when the width and height of the transform unit are both 32 or less.
  • step 8810 may be omitted and preset in the encoder and / or decoder. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • FIG. 9 is a flowchart illustrating a process of encoding an MTS flag and an MTS index according to an embodiment to which the present invention is applied.
  • the encoder may determine whether MTS (Multiple Transform Selection) is applied to the current block (S910).
  • MTS Multiple Transform Selection
  • the encoder may determine an MTS index based on at least one of a prediction mode, a horizontal transform, and a vertical transform of the current block (S930).
  • the MTS index means an index indicating any one of a plurality of transform combinations for each intra prediction mode, and the MTS index may be transmitted for each transform unit.
  • the encoder can encode the MTS index (S940).
  • FIG. 10 is a flowchart illustrating a decoding process of applying a horizontal transform or a vertical transform to a row or a column based on an MTS flag and an MTS index as an embodiment to which the present invention is applied.
  • the decoder may parse the MTS flag from the bitstream (S1010).
  • the MTS flag is set to MTS (Multiple Transform Selection) for the current block. ⁇ ) 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the decoder may determine whether MTS (Multiple Transform Selection) is applied to the current block based on the MTS flag (S1020). For example, it may be checked whether the MTS flag is 1 or not.
  • MTS Multiple Transform Selection
  • the decoder may check whether the number of non-zero transform coefficients is greater than (or greater than) a threshold value (S1030). For example, the threshold may be set to 2, which may be set differently based on the block size or the size of the transform unit.
  • the decoder may parse an MTS index (S1040).
  • the MTS index means an index indicating any one of a plurality of transform combinations for each intra prediction mode or inter prediction mode, and the MTS index may be transmitted for each transform unit.
  • the MTS index may mean an index indicating any one transform combination defined in a preset transform combination table, and the preset transform combination table may mean the FIG. 6, but the present invention is limited thereto. It doesn't work.
  • the decoder may derive or determine a horizontal transform and a vertical transform based on at least one of the MTS index and the prediction mode (S1050).
  • the decoder may derive a transform combination corresponding to the MTS index.
  • the decoder may derive or determine a horizontal transform and a vertical transform corresponding to the MTS index.
  • the decoder may apply a predetermined vertical inverse for each column (060).
  • the vertical inverse transform may be an inverse transform of 17.
  • the decoder may apply a predetermined horizontal inverse transformation for each row 1070.
  • the horizontal inverse transform may be an inverse transform of 17. That is, when the number of non-zero transform coefficients is not greater than a threshold value, a transform kernel preset by an encoder or a decoder may be used.
  • a conversion kernel that is widely used may not be defined in the conversion combination table as shown in FIG. 6.
  • the decoder may apply a predetermined tube vertical inverse transform for each column 1080).
  • the vertical inverse transform may be an inverse transform of 1 ) 12.
  • the decoder may apply a predetermined horizontal inverse transformation for each row 1090.
  • the horizontal inverse transform may be an inverse transform of 0012.
  • a conversion kernel preset at the encoder or the decoder may be used.
  • a conversion kernel that is widely used may not be defined in the conversion combination table as shown in FIG. 6.
  • FIG. 1 is a schematic block diagram of an inverse transform unit in a decoder according to an embodiment to which the present invention is applied.
  • the decoder to which the present invention is applied includes a second inverse transform application determining unit (or an element for determining whether to apply the second inverse transform) (1 1 10), a second inverse transform determining unit (or an element for determining the second inverse transform) ( 1 120), the second inverse transform unit (or to perform the second inverse transform 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the second inverse transform application determining unit 1 1 10 may determine whether to apply the second inverse transform.
  • the secondary inverse transform may be a Non-Separable Secondary Transform (hereinafter referred to as NSST) or Reduced Secondary Transform (hereinafter referred to as RST).
  • NSST Non-Separable Secondary Transform
  • RST Reduced Secondary Transform
  • the second inverse transform application determiner 1 1 10 may determine whether to apply the second inverse transform based on the second transform flag received from the encoder.
  • the second inverse transform determining unit 1 1 10 may determine whether to apply the second inverse transform based on the transform coefficient of the residual block.
  • the second inverse transform determiner 1 120 may determine the second inverse transform. In this case, the second inverse transform determiner 1 120 may determine the second inverse transform applied to the current block based on the NSST (or RST) transform set specified according to the intra prediction mode.
  • the secondary transform determination method may be determined based on the primary transform determination method.
  • Various combinations of primary and secondary transforms may be determined according to the intra prediction mode.
  • the second inverse transform determiner 1 120 may determine an area to which the second inverse transform is applied based on the size of the current block.
  • the second inverse transform unit 1 130 may perform a second inverse transform on the residual quantized residual block by using the determined second inverse transform.
  • the first inverse transform unit 1140 may perform a first inverse transform on the residual block that is second inverse transformed.
  • the primary transform may be referred to as a primary transform or a core transform.
  • the first inverse transform unit 1140 may use the above-described MTS to perform the first order. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the first inverse transform unit 1140 may determine whether MTS is applied to the current block.
  • the MTS when the MTS is applied to the current block (that is, The first inverse transform unit 1 140 based on the intra prediction mode of the current block.
  • Candidates can be configured.
  • the 1 ⁇ 8 candidate may consist of a combination of 0814 and / or DCT4 ⁇ ⁇ , or may include a combination of 0817 and / or DCTSS].
  • the MTS candidate may include at least one of the embodiments of FIG. 6 and the embodiments of FIGS. 26 and 27 to be described later.
  • the first inverse transform unit 1140 is configured Particular candidates
  • the 11 _ ⁇ indicating the MTS may be used to determine the primary transform applied to the current block.
  • FIG. 12 is a block diagram for performing inverse transformation based on transformation related parameters in an embodiment to which the present invention is applied.
  • the decoder 200 to which the present invention is applied includes an element 1210 for acquiring sequence parameters, a multi-conversion selection flag (element 1220 for acquiring a power factor), and a multi-conversion selection index ( ⁇ ⁇ output 0 ⁇ ⁇ 0 rule. Element 1230 to obtain and element 1240 to derive the transform kernel.
  • the element 1210 obtaining the sequence parameter may acquire sps_mts_intra_enabled_flag or sps_mts_intcr_enabled_flag. here whether tu_mts_flag exists in the residual coding syntax of the intra coding unit, and sps_mts_inter_enabled_flag indicates whether tu_mts_flag exists in the residual coding syntax of the inter coding unit.
  • the description of FIG. 12 may be applied.
  • tu_mts_flag indicates whether multi-transform selection is applied to the residual sample of the luma transform block. As a specific example, the description of FIG. 12 may be applied.
  • mts_idx indicates which transform kernel is applied to luma residual samples along the horizontal and / or vertical direction of the current transform block. For example, at least one of the embodiments of FIG. 6 and the embodiments of FIGS. 26 and / or 27 described later may be applied.
  • An element 1240 for deriving a transform kernel may derive a transform kernel corresponding to mtsjdx.
  • the decoder 200 may perform inverse transformation based on the transform kernel. 2019/190284 1 »(: 1/10 ⁇ 019/003743
  • 13 is an embodiment to which the present invention is applied and shows a flowchart of performing inverse transformation based on transformation related parameters.
  • the decoder to which the present invention is applied may acquire sps_mts_intra_enabled_flag or sps_mts_inter_enabled_flag (S 1310).
  • sps_mts_intra_enabled_flag indicates whether tu_mts_flag exists in the residual coding syntax of the intra coding unit.
  • sps__mts__intra_enabled_flag 0 tu_mts__flag is not present in the residual coding syntax of the intra coding unit
  • sps_mts_intra_enabled_flag 0 tu_mts_flag is present in the residual coding syntax of the intra coding unit.
  • sps_mts_i nter_enab 1 ed_fl ag indicates whether tu__mts_flag exists in the residual coding syntax of the inter coding unit.
  • At least one of the embodiments of the present document may be applied to the tix_mtsJlag.
  • 0 indicates which transform kernel is applied to luma residual samples along the horizontal and / or vertical direction of the current transform block.
  • At least one of the document embodiments may be applied.
  • at least one of the embodiments of FIG. 6 and the embodiments of FIGS. 26 and / or 27 described later may be applied.
  • a corresponding translation kernel can be derived 1340.
  • the 11__1 conversion kernel corresponding to urine
  • transform kernels may be applied to the horizontal transform and the vertical transform.
  • present invention is not limited thereto, and the same transform kernel may be applied to the horizontal transform and the vertical transform.
  • the decoder may perform inverse transformation based on the transform kernel.
  • DST4 Discrete Sine Transform-4
  • DCT4 Discrete Cosine Transform-4
  • the encoder may determine (or select) a horizontal transform and / or a vertical transform based on at least one of a prediction mode, a block shape, and / or a block size of the current block (S1410).
  • the candidate for horizontal transformation and / or vertical transformation may include at least one of the embodiments of FIG. 6 and the embodiments of FIGS. 26 and / or 27 to be described later.
  • the encoder may determine an optimal horizontal transformation and / or an optimal vertical transformation through RD optimization (Rate Distortion optirr ⁇ zation).
  • the optimal horizontal transform and / or the optimal vertical transform may correspond to one of a plurality of transform combinations, and the plurality of transform combinations may be defined by a transform index.
  • the encoder may signal a transform index corresponding to the optimal horizontal transform and / or the optimal vertical transform (S1420).
  • a transform index corresponding to the optimal horizontal transform and / or the optimal vertical transform S1420
  • other embodiments described herein may be applied to the conversion index.
  • it may include at least one of the embodiments of FIG. 6 and the embodiments of FIGS. 26 and 27 to be described later.
  • a horizontal transform index for the optimal horizontal transform and a vertical transform index for the optimal vertical transform may be signaled independently.
  • the encoder may perform a forward transformation in the horizontal direction with respect to the current block by using the optimal horizontal transformation (S1430).
  • the current block may mean a transform block. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the encoder may perform a forward transform in a vertical direction with respect to the current block by using the optimal vertical transform.
  • the vertical transformation is performed after the horizontal transformation, but the present invention is not limited thereto. That is, the vertical transformation may be performed first, and then the horizontal transformation may be performed.
  • the forward direction in the horizontal forward direction conversion of the step S1430 is not limited thereto.
  • DCT4 can be applied. Or vice versa.
  • the combination of the horizontal transform and the vertical transform may include at least one of the embodiments of FIG. 6 and the embodiments of FIGS. 26 and 27 to be described later.
  • the encoder may generate a transform coefficient block by performing quantization on the current block (S 1450).
  • the encoder may perform entropy encoding on the transform coefficient block to generate a bitstream.
  • FIG. 15 shows a decoding flowchart of performing 3 ⁇ 4 DST4 (Discrete Sine Transform-4) and DCT4 (Discrete Cosine Transfomi-4) as a forward DCT2 or a reverse DCT2 as an embodiment to which the present invention is applied. .
  • the decoder may obtain a transform index from the bitstream (S1510).
  • a transform index from the bitstream (S1510).
  • other embodiments described herein may be applied to the conversion index.
  • it may include at least one of the embodiments of FIG. 6 and the embodiments of FIGS. 26 and 27 to be described later. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the decoder may induce a horizontal transform and a vertical transform corresponding to the transform index 1520.
  • the candidate for the horizontal transformation and / or the vertical transformation may include at least one of the embodiments of FIG. 6 and the embodiments of FIGS. 26 and / or 27 to be described later.
  • steps 81510 and 81520 are an embodiment, and the present invention is not limited thereto.
  • the decoder may induce a horizontal transform and a vertical transform based on at least one of a prediction mode, a block shape, and / or a block size of the current block.
  • the transform index may include a horizontal transform index corresponding to a horizontal transform and a vertical transform index corresponding to a vertical transform.
  • the decoder may entropy decode the bitstream to obtain a transform coefficient block and perform inverse quantization on the transform coefficient block (81530).
  • the decoder may perform inverse transformation in the vertical direction on the inverse quantized transform coefficient block by using the vertical transformation.
  • the decoder may perform inverse transformation in the horizontal direction using the horizontal transformation (81550).
  • the horizontal transformation is applied after the vertical transformation is applied, but the present invention is not limited thereto. That is, the horizontal transformation may be applied first, and then the vertical transformation may be applied.
  • the reverse direction 14 is applied in the vertical reverse direction conversion of the step 81540, and the reverse ⁇ (the 4) may be applied in the horizontal direction reverse conversion of the 81440 step, or vice versa.
  • the combination of the horizontal transform and the vertical transform is shown in FIG. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the decoder generates a residual block through step S1550, and the reconstructed block is generated by adding the residual block and the prediction block.
  • FIG. 16 shows diagonal elements for a pair of transform block size (N) and right shift amount (SO) when DST4 and DCT4 are performed in forward DCT2 as an embodiment to which the present invention is applied.
  • the present invention proposes a method of reducing memory usage and computational complexity for Discrete Sine Transform-4 (DST4) and Discrete Cosine Transform-4 (DCT4) among transform kernels for video compression.
  • DST4 Discrete Sine Transform-4
  • DCT4 Discrete Cosine Transform-4
  • the present invention provides a method of performing Discrete Sine Transform-4 (DST4) and Discrete Cosine Transform-4 (DCT4) to forward DCT2.
  • DST4 Discrete Sine Transform-4
  • DCT4 Discrete Cosine Transform-4
  • a method of performing DST4 and DCT4 to inverse DCT2 is provided.
  • a method of applying DST4 and DCT4 to a transform configuration group to which MTS (Multiple Transform Selection) is applied is provided.
  • MTS Multiple Transform Selection
  • Equations 4 and 5 generate inverse transform matrices of DST4 and DCT4, respectively. And these transposes represent a forward transformation matrix.
  • the present invention changes the input order or the output order through a pre-processing stage or a post-processing stage and signs.
  • DCT4 DST4
  • the present invention can easily derive one from another without additional calculations when performing 0 4 or 00141 ⁇ .
  • £> (# 4 is £) (# 2) can be expressed as follows.
  • 1X ⁇ 4 can be designed based on the post-processing matrix 71, the preprocessing matrix, and 1X72. here, 2019/190284 1 »(: 1/10 ⁇ 019/003743
  • DCT2 can reduce the number of coefficients to be stored and is well known as a transformation for fast implementation based on symmetry between coefficients in DCT2 matrix.
  • Equation 9 The inverse of the post-processing matrix ⁇ and the preprocessing matrix ⁇ can be expressed by Equation 9 below.
  • A- ] , M ⁇ ] are composed of simpler multiply than (), it is possible to fast implementation of DCT4 with low complexity. And A- 'results in fewer additions or subtractions than and v, but the coefficients in MJ- have a wider range than M N. Accordingly, the present invention can design a conversion type based on Equations 9 and 10 in consideration of a tradeoff between complexity and performance.
  • Equation 7 the present invention can execute low complexity DST4 by reusing a fast implementation of DCT2. This is shown in the following equations (11) and (12).
  • Equation 11 When Equation 11 is used to execute DST4, the input vector of length N must first be scaled by M N J n ). Similarly, when Equation 8 is used for the execution of DCT4, the input vector of length N must first be scaled by (M n ).
  • Diagonal elements in M N are floating point numbers, which must be scaled appropriately for use in fixed-point or integer multiplications. Integerized (M N J n ) and
  • diag ( ⁇ ) means converting the argument matrix into an associative vector constituting the diagonal elements in the argument matrix.
  • Acid ((%, and) of the same (ss ⁇ ) can easily be derived from the FIG. 16 by changing the order of the elements in each vector. For example, [251,213,142,50] may be changed to [50,142,213,251].
  • the present invention may set 5 / differently for each. For example, it may be set to 7 for a 4x4 transform and 5; to 8 for an 8x8 transform.
  • Equation 13 Denotes the amount of left shift to scale by 2 " 1 , and the" skin line "operator performs the appropriate 1 > 01111 (111. ⁇ and () are diagonal matrices and the 1 * elements of the input vector X ( ⁇ 11 ( (3 ⁇ ) is multiplied by [ ⁇ ,, Multiplication result and diagonal of the input vector X
  • Equation 14 The matrices can be expressed as Equation 14 below.
  • Fig. 14 shows the multiplication result.
  • X must be scaled down later. Downscaling of ⁇ is performed before applying 1X72, or 2019/190284 1 »(: 1/10 ⁇ 019/003743
  • the scaled down (the down-scaled one), x may be determined based on Equation 15 below.
  • 3 ⁇ 4 may be the same value as described above. However, the present invention is not limited thereto, and 3 ⁇ 4 may have a different value from the above.
  • Equation 15 any type of scaling and rounding may be used, and (1) and (2) of Equation 15 may be used in one embodiment. That is, as shown in Equation 15, (1), (2) or other functions may be applied to find a station. 17 and 18 are embodiments to which the present invention is applied, and FIG. 17 shows sets of DCT2 kernel coefficients that can be applied to DST4 or DCT4, and FIG. 18 shows a forward DCT2 matrix generated from a set of DCT2 kernel coefficients.
  • One embodiment of the present invention may use the same DCT2 kernel coefficients as HEVC. It is necessary to maintain 31 different coefficients of DCT2, which is facilitated by symmetries among all DCT2 kernel coefficients of all sizes up to 32x32, used by symmetry between all DCT2 kernel coefficients of up to 32x32.
  • the present invention may add only one set of DCT2 kernel coefficients, which are 31 coefficients using homogeneous symmetry. That is, if up to 2 n x 2 n DCT2 is supported, the present invention only needs (2 n -l) different coefficients.
  • This additional set may have higher or lower accuracy than before. If the dynamic range of x does not exceed the range supported by the existing DCT2 design, the present invention will reuse the same routine as DCT2 without extending the bit length of internal variables. And reuse the legacy design of DCT2.
  • Each coefficient of FIG. 17 may be further adjusted to improve orthogonality between basis vectors, the norm of each basis vector is close to 1, and the Frobenius norm error can be reduced from a floating-point accurate DCT2 kernel.
  • the coefficient set is (a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p, q, r, s, t, u ⁇ w, x, y, z, A, B, C, D, E)
  • the forward DCT2 generated from the coefficient set can be set as shown in FIG.
  • each DCT2 coefficient set (each row in FIG. 18) is represented by (a, b ? C, d, e, f, g, h, i, j, k, l, m, n, o, p , q, r, s, t, u, v, w, x, y, z, A, B, C, D, E).
  • This is 31 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the output of the DCT2 transform must be post-processed via the matrix A N (or DNA N ) of DCT4 (or DST4).
  • the DCT2 output vector as the input vector may be rounded to a value for accuracy adjustment to store as variables of defined bit lengths. Assuming that the DCT2 output vector before scaling and rounding is y, a rounded one y can be determined from the following equation (16).
  • Equation 17 it is assumed that the present invention applies a right shift equal to the left shift of in order not to change the overall scaling, but need not be. If, when applying the right shifted by 3 ⁇ 4 ( ⁇ S 4) instead of S 4, the present invention is All j) must be scaled up. Considering all the shifts of the previous equations and the expected result scaling after DCT4 (or DST4) calculation (5V, where positive value means right shift), the present invention provides all scaling bit shift values ( Equation 18 with all the scaling bit shift values) can be set. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • Equation 18 Sc represents a left shift amount due to DCT2 integer multiplication, which may be a non-integer value as shown in FIG. 17. So represents the right shift amount that yields the final output of DCT4 (or DST4). Some portions of Equation 18 may be zero. For example, (5 /-&), 3 ⁇ 4 or (Ss-S 4 ) can be zero. 19 and 20 illustrate embodiments to which the present invention is applied. FIG. 19 illustrates code execution of an output stage for DST4, and FIG. 20 illustrates code execution of an output stage for DCT4.
  • an embodiment of the present invention can provide an example of final step code execution for DST4 corresponding to multiplication of the public as shown in FIG.
  • another embodiment of the present invention may provide an example of final step code execution for DCT4 corresponding to multiplication of as shown in FIG.
  • the cutoff in FIG. 19 represents the effective number of coefficients in the vector X.
  • the cutoff may be N.
  • steps S 1910 and S1920 may be merged into one calculation process as shown in Equation 19 below.
  • steps S2010 and S2020 may be merged into one calculation process as shown in Equation 20 below.
  • Clip3 represents an operation of clipping the argument value to both ends (clipMinimum, clipMaximum).
  • a N (or (D N A n )) ⁇ ⁇
  • Each row has a common pattern with its previous row, and the present invention is directed to appropriate sign sign reversal. You can reuse a result of the previous row.
  • Such a pattern may be utilized through the variables z and prev in FIGS. 19 and 20.
  • the variable z, prev reduces the multiplication calculation of A N (or ( ⁇ ⁇ )).
  • FIG. 21 illustrates an embodiment to which the present invention is applied and shows configuration of parameter sets and multiplication coefficients for DST4 and DCT4 when DST4 and DCT4 are performed in forward DCT2. 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • each transform of a different size can be configured separately. In other words, each transform of a different size may have its own parameter set and multiplication coefficients.
  • the multiplication factor value is (8, 8, 0, 8, 8, identical to HEVC for all block sizes). Can be).
  • the configuration of the parameter set of DCT4 is (Si, S 2 , S 3 , S4, Ss, So)
  • the multiplication factor value is (8, 8, 0, 8, 8, identical to HEVC for all block sizes). Can be).
  • each block size may have respective multiplication coefficient values described in FIG. 21.
  • FIG. 22 illustrates code execution of a preprocessing step for DCT4
  • FIG. 23 illustrates code execution of a preprocessing step for DST4.
  • Example 3 Alternative implementation of DST4 and DCT4 with inverse DCT2
  • the present invention provides a method of executing DCT4 and DST4 through Equations 10 and 12, respectively.
  • A, (A N -'J n ), M- ⁇ and (D N MJ-) may be used instead of A n , (D N A n ), M n , and ⁇ M N J n , each of which Requires less computation compared to DCT2.
  • Reverse DCT2 is applied instead of forward DCT2 in Equations 10 and 12 above. 2019/190284 1 »(: 1/10 ⁇ 019/003743
  • Equations 9 and 12 only one element is multiplied by V2 in AJ- and ( ⁇ 1 ⁇ ).
  • a 'and (A ⁇ ] J n ) may be approximated by integer multiplication by the right shift.
  • FIG. 22 An example of code execution in the preprocessing step of DCT4 in Equation 10 is shown in FIG. 22, which corresponds to the multiplication of A ⁇ ′.
  • FIG. 23 An example of code execution in the preprocessing step of DST4 in Equation 12 is shown in FIG. 23, which corresponds to (multiplication of three times).
  • N indicates the length of the transform basis vector as well as the length of the input vector x.
  • the present invention can scale down the reverse 1X ⁇ 2 output to take advantage of shorter bit length variables. If the backward 1X ⁇ 2 output vector is ⁇ and the W element is written, the scaled output vector can be obtained according to Equation 21 below. As in equation (15), other forms of scaling and rounding are applicable to equation (21).
  • Equations 10 and 12 the post-processing steps correspond to and ( ⁇ ⁇ >), respectively.
  • the associated diagonal coefficients can be scaled up for fixed point or integer multiplication. This scale up may be performed with appropriate left shifts as shown in Equation 22 below.
  • Diagonal elements for a pair of transform block sizer and right shift amount 4) when executed at 56. Examples of diagonal elements of can be seen in various combinations of Figures 24 and VII above.
  • S 4 may be set differently for each transform size.
  • (N, S4) is (32, 9)
  • large numbers such as '10431,' can be decomposed as shown in Equation 23 below, which is suitable for multiplication of a shorter bit length operator part. This can be applied when a large number of multiplications appear.
  • Non-zero elements are only available on diagonal lines in M and ⁇ D N M ⁇ ), and the associated matrix multiplication is element-wise multiplication, as in Equation 24 It can be performed by).
  • FIG. 25 illustrates a configuration of parameter sets and multiplication coefficients for DST4 and DCT4 when DST4 and DCT4 are performed in reverse DCT2 as an embodiment to which the present invention is applied.
  • each transform of a different size can be configured separately. In other words, each transform of a different size may have its own parameter set and multiplication coefficients.
  • the multiplication factor value for all block sizes is (8, 8, 0, 8, 8, identical to HEVC).
  • each block size may have respective multiplication coefficient values described in FIG. 25.
  • the execution of the reverse DST4 [DCT4] is the same as the forward DST4 [DCT4].
  • 26 and 27 illustrate embodiments to which the present invention is applied, FIG. 26 shows MTS mapping for intra prediction residual, and FIG. 27 shows inter prediction residual.
  • Embodiment 4 Possible Multiple Transform Selection (MTS) mapping with DST4 and DCT4
  • DCT4 and DST4 may be used to generate MTS mapping.
  • DST7 and DCT8 may be replaced by DCT4 and DST4.
  • DCT4 and DST4 may be used to generate the MTS.
  • Tables 13 and 14 show MTS examples for intra predicted residual and inter predicted residual, respectively.
  • mapping is also possible by other combinations of DST4, DCT4, DCT2, and the like.
  • mapping for the inter predicted residual consisting of 1 ⁇ 78/0817 may be retained and replaced only for the intra predicted residual.
  • a combination of the above embodiments is possible.
  • a content streaming system to which the present invention is applied may largely include an encoding server, a streaming server, a web server, a media storage, a user device, and a multimedia input device.
  • the encoding server compresses content input from multimedia input devices such as a smartphone, a camera, a camcorder, etc. into digital data to generate a bitstream and transmit the bitstream to the streaming server.
  • multimedia input devices such as smart phones, cameras, camcorders, etc. directly generate a bitstream
  • the encoding server may be omitted.
  • the bitstream may be generated by an encoding method or a bitstream generation method to which the present invention is applied, and the streaming server may temporarily store the bitstream in the process of transmitting or receiving the bitstream.
  • the streaming server transmits the multimedia data to the user device based on the user's request through the web server, and the web server serves as a medium for informing the user of what service.
  • the web server delivers it to a streaming server, and the streaming server transmits multimedia data to the user.
  • the content streaming system may include a separate control server.
  • the control server controls a command / response between devices in the content streaming system.
  • the streaming server may receive content from a media store and / or an encoding server. For example, when the content is received from the encoding server, the content may be received in real time. In this case, in order to provide a smooth streaming service, the streaming server may store the bitstream for a predetermined time.
  • Examples of the user device include a mobile phone, a smart phone, a laptop computer, a digital broadcasting terminal, a personal digital assistant (PDA), a portable multimedia player (PMP), a navigation device, a slate PC, Tablet PCs, ultrabooks, wearable devices such as watchwatches, glass glasses, head mounted displays, digital TVs, desktops Computer, digital signage, and the like.
  • PDA personal digital assistant
  • PMP portable multimedia player
  • a navigation device a slate PC
  • Tablet PCs Tablet PCs
  • ultrabooks wearable devices such as watchwatches, glass glasses, head mounted displays, digital TVs, desktops Computer, digital signage, and the like.
  • Each server in the content streaming system may be operated as a distributed server, in which case data received from each server may be distributed.
  • the embodiments described herein may be implemented and performed on a processor, microprocessor, controller, or chip.
  • the functional units shown in each drawing may be implemented and performed on a computer, a processor, a microprocessor, a controller, or a chip.
  • the decoder and encoder to which the present invention is applied are multimedia broadcast transmission and reception 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • OTT video (Over the top video) device may include a game console, a Blu-ray player, an Internet-connected TV, a home theater system, a smartphone, a tablet PC, a digital video recorder (DVR), and the like.
  • the processing method to which the present invention is applied can be produced in the form of a program executed by a computer, and stored in a computer-readable recording medium.
  • Multimedia data having a data structure according to the present invention can also be stored in a computer-readable recording medium.
  • the computer readable recording medium includes all kinds of storage devices and distributed storage devices in which computer readable data is stored.
  • the computer-readable recording medium may be, for example, a Blu-ray disc (BD), a universal serial bus (USB), a ROM, a PROM, an EPROM, an EEPROM, a RAM, a CD-ROM, a magnetic tape, a floppy disk, and an optical disc. It may include a data storage device.
  • the computer-readable recording medium also includes media embodied in the form of a carrier wave (for example, transmission over the Internet).
  • the bitstream generated by the encoding method may be stored in a computer-readable recording medium or transmitted through a wired or wireless communication network.
  • an embodiment of the present invention is a computer program by the program code 2019/190284 1 »(: 1 ⁇ 1 ⁇ 2019/003743
  • the program code may be implemented in a product, and the program code may be executed on a computer by an embodiment of the present invention.
  • the program code may be stored on a carrier readable by a computer.

Abstract

본 발명은, 저복잡도 변환 실행에 기초하여 비디오 신호를 복원하는 방법에 있어서, 상기 비디오 신호로부터 현재 블록의 변환 인덱스를 획득하는 단계, 여기서 상기 변환 인덱스는 DST4 및/또는 DCT4의 조합으로 구성된 복수개의 변환 조합들 중 어느 하나에 대응됨; 상기 변환 인덱스에 대응되는 변환 조합을 유도하는 단계, 여기서 상기 변환 조합은 수평 변환 및 수직 변환으로 구성되고, 상기 수평 변환 및 상기 수직 변환은 상기 DST4 또는 상기 DCT4 중 어느 하나에 대응됨; 상기 DST4를 이용하여, 상기 현재 블록에 대해 수직 방향으로 역방향 변환을 수행하는 단계; 상기 DCT4를 이용하여, 상기 현재 블록에 대해 수평 방향으로 역방향 변환을 수행하는 단계; 및 상기 역방향 변환된 현재 블록을 이용하여 상기 비디오 신호를 복원하는 단계를 포함하는 것을 특징으로 하는 방법을 제공한다.

Description

2019/190284 1»(:1^1{2019/003743
1
【명세서】
【발명와 명칭】
비디오 압축을 위한 변환 커널의 저복잡도 연산을 수행하는 방법 및 장치 【기술분야】
본 발명은 비디오 신호를 처리하는 방법 및 장치에 관한 것이며, 보다 구체적으로 비디오 압축을 위한 변환 커널 중 DST4(Discrete Sine Transform-4) 및 DCT4(Discrete Cosine Transform-4)에 대한 메모리 사용 및 연산 복잡도를 감소시키는 기술에 관한 것이다.
【배경기술】
차세대 비디오 컨텐츠는 고해상도 (high spatial resolution), 고프레임율 (high frame rate) 및 영상 표현의 고차원화 (high dimensionality of scene representation)라는 특징을 갖거 l·될 것이다. 그러한 컨텐츠를 처리하기 위해서는 메모리 저장 (memory storage), 메모리 액세스율 (memory access rate) 및 처리 전력 (processing power) 측면에서 엄청난 증가를 가져올 것이다.
따라서, 차세대 비디오 컨텐츠를 보다 효율적으로 처리하기 위한 새로운 코딩 툴을 디자인할 필요가 있다. 특히, 변환 (transform)을 적용할 때 부호화 효율과 복잡도 측면에서 훨씬 효율적인 변환을 설계할 필요가 있다.
【발명의 상세한 설명】
【기술적 과제】
본 발땅은 비디오 압축을 위한 변환 커널에 대한 저복잡도의 연산 알고리즘을 제안하고자 한다.
본 발명은 비디오 압축을 위한 변환 커널 중 DST4(Discrete Sine Transform- 2019/190284 1»(:1^1{2019/003743
2
4) 및 DCT4(Discrete Cosine Transform-4)에 대한 메모리 사용 및 연산 복잡도를 감소시키는 방법을 제안하고자 한다.
본 발명은 새로운 변환 디자인을 반영하기 위한 인코더/디코더 구조를 제안하·고자 한다.
【기술적 해결방법】
본 발명은, 새로운 변환 디자인을 통해 복잡도를 감소시키고 코딩 효율을 향상시키는 방법을 제공한다.
본 발명은, DST4(Discrete Sine Transform-4) 및 DCT4(Discrete Cosine
Transform-4)를 순방향 DCT2 (forward DCT2)로 수행하는 방법을 제공한다.
본 발명은, DST4 및 DCT4를 역방향 DCT2 (inverse DCT2)로 수행하는 방법을 제공한다.
본 발명은, DST4 및 DCT4를 MTS(Multiple Transform Selection)가 적용되는 변환 설정 그룹 (transform configuration group)에 적용하는 방법을 제공한다.
【발명의 효과】
본 발명은, DST4(Discrete Sine Transform-4) 및 DCT4(Discrete Cosine
Transform-4)를 순방향 DCT2 (forward DCT2) 또는 역방향 DCT2 (inverse DCT2)로 수행하는 방법을 제공함으로써, 메모리 사용 및 연산 복잡도를 감소시킬 수 있다. 또한, 본 발명은, DST4 및 DCT4를 MTS(Multiple Transform Selection)가 적용되는 변환 설정 그룹 (transform configuration group)에 적용함으로써, 보다 효율적인 코딩을 수행할 수 있다.
이와 같이, 새로운 저복잡도 연산 알고리즘을 통해 연산 복잡도를 감소시키고 코딩 효율을 향상시킬 수 있다. 2019/190284 1»(:1^1{2019/003743
3
【도면의 간단한 설명】
도 1은 본 발명이 적용되는 실시예로서, 비디오 신호의 인코딩이 수행되는 인코더의 개략적인 블록도를 나타낸다.
도 2는 본 발명이 적용되는 실시예로서, 비디오 신호의 디코딩이 수행되는 디코더의 개략적인 블록도를 나타낸다.
도 3은 본 발명이 적용될 수 있는 실시예들로서, 도 3A는 QT(QuadTree, 이하 ‘QT’라 함), 도 3B는 BT(Binary Tree, 이하 ‘BT라 함), 도 3C는 TT(Ternary Tree, 이하 ‘TT’라 함) 도 3D는 AT(Asymmetric Tree, 이하 ‘AT,라 함)에 의한 블록 분할 구조들을 설명하기 위한 도면이다.
도 4는 본 발명이 적용되는 실시예로서, 인코더 내의 변환 및 양자화부 (120/130), 역양자화 및 역변환부 (140/150)의 개략적인 블록도를 나타낸다.
도 5는 본 발명이 적용되는 실시예로서, 디코더 내의 역양자화 및 역변환부 (220/230)의 개략적인 블록도를 나타낸다.
도 6은 본 발명이 적용되는 실시예로서, MTS(Multiple Transform Selecti에)가 적용되는 변환 설정 그룹 (transform configuration group)을 나타내는 표이다.
도 7은 본 발명이 적용되는 실시예로서, MTS(Multiple Transform Selection)가 수행되는 인코딩 과정을 나타내는 흐름도이다.
도 8은 본 발명이 적용되는 실시예로서, MTS(Multiple Transform Selection)가 수행되는 디코딩 과정을 나타내는 흐름도이다.
도 9는 본 발명이 적용되는 실시예로서, MTS 플래그 및 MTS 인덱스를 인코딩하는 과정을 설명하기 위한 흐름도이다.
도 10은 본 발명이 적용되는 실시예로서, MTS 플래그 및 MTS 인덱스에 2019/190284 1»(그1^1{2019/003743
4 기초하여 수평 변환 또는 수직 변환을 행 또는 열에 적용하는 디코딩 과정을 설명하기 위한흐름도이다.
도 11은 본 발명이 적용되는 실시예로서, 디코더 내 역변환부의 개략적인 블록도를 나타낸다.
도 12는 본 발명이 적용되는실시예로서, 변환 관련 파라미터에 기초하여 역변환을 수행하는 블록도를 나타낸다.
도 13은 본 발명이 적용되는 실시예로서, 변환 관련 파라미터에 기초하여 역변환을 수행하는 흐름도를 나타낸다.
도 14는 본 발명이 적용되는 실시예로서, DST4(Discrete Sine Transform-4) 및 DCT4(Discrete Cosine Transform-4)를 순방향 DCT2 (forward DCT2) 또는 역방향 DCT2 (inverse DCT2)로 수행하는 인코딩 흐름도를 나타낸다.
도 15는 본 발명이 적용되는 실시예로서, DST4(Discrete Sine Transform-4) 및 DCT4(Discrete Cosine Transform-4)를 순방향 DCT2 (forward DCT2) 또는 역방향 DCT2 (inverse DCT2)로 수행하는 디코딩 흐름도를 나타낸다.
도 】6은 본 발명이 적용되는 실시예로서, DST4 및 DCT4를 순방향 DCT2 로 수행시 변환 블록 사이즈 (N) 및 우측 쉬프트 양 ( )의 쌍에 대한 대각 요소들을 나타낸다.
도 17은 본 발명이 적용되는 실시예로서, DST4 또는 DCT4에 적용될 수 있는 DCT2 커널 계수들의 세트들을 나타낸다.
도 18은 본 발명이 적용되는 실시예로서, DST4 또는 DCT4에 적용될 수 있는 DCT2 커널 계수의 세트로부터 생성된 순방향 DCT2 행렬을 나타낸다.
도 19는 본 발명이 적용되는 실시예로서 , DST4 에 대한 출력 단계의 코드 2019/190284 1»(:1^1{2019/003743
5 실행을 나타낸다.
도 20은 본 발명이 적용되는 실시예로서, DCT4 에 대한 출력 단계의 코드 실행을 나타낸다.
도 21은 본 발명이 적용되는 실시예로서, 0814 및 1 :14를 순방향 0012 로 수행시 0814 및 1X74 에 대한 파라미터 세트 및 곱셈 계수들의 구성을 나타낸다.
도 22는 본 발명이 적용되는 실시예로서, 0014 에 대한 전처리 단계의 코드 실행을 나타낸다.
도 23은 본 발명이 적용되는 실시예로서, 0814 에 대한 전처리 단계의 코드 실행을 나타낸다.
도 24는 본 발명이 적용되는 실시예로서, 0814 및 1X714를 역방향 0012 로 수행시 변환 블록 사이즈어) 및 우측 쉬프트 양 4)의 쌍에 대한 대각 요소들을 나타낸다.
도 25는 본 발명이 적용되는 실시예로서, 0814 및 00X41- 역방향 0012 로 수행시 14 및 1X74 에 대한 파라미터 세트 및 곱셈 계수들의 구성을 나타낸다.
도 26은 본 발명이 적용되는 실시예로서, 인트라 예측 레지듀얼에 대한 8 매핑을 나타낸다.
도 27은 본 발명이 적용되는 실시예로서, 인터 예측 레지듀얼에 대한 1 18 매핑을 나타낸다.
도 28은 본 발명이 적용되는 실시예로서, 컨텐츠 스트리밍 시스템 구조도를 나타낸다. 2019/190284 1»(:1^1{2019/003743
6
【발명의 실시를 위한 최선의 형태】
본 발명은, 저복잡도 변환 실행에 기초하여 비디오 신호를 복원하는 방법에 있어서, 상기 비디오 신호로부터 현재 블록의 변환 인덱스를 획득하는 단계, 여기서 상기 변환 인덱스는 0814 및/또는 DC^4 의 조합으로 구성된 복수개의 변환 조합들 중 어느 하나에 대응됨; 상기 변환 인덱스에 대응되는 변환 조합을 유도하는 단계, 여기서 상기 변환 조합은 수평 변환 및 수직 변환으로 구성되고, 상기 수평 변환 및 상기 수직 변환은 상기 0814 또는 상기 0014 중 어느 하나에 대응됨; 상기 0314를 이용하여, 상기 현재 블록에 대해 수직 방향으로 역방향 변환을 수행하는 단계; 상기 1X^4를 이용하여, 상기 현재 블록에 대해 수평 방향으로 역방향 변환을 수행하는 단계; 및 상기 역방향 변환된 현재 블록을 이용하여 상기 비디오 신호를 복원하는 단계를 포함하는 것을 특징으로 하는 방법을 제공한다.
본 발명에서, 상기 0814 및/또는 상기 1X714는 순방향 0072 또는 역방향 이용하여 실행되는 것을 특징으로 한다.
본 발명에서, 상기 0814 및/또는 상기 å)( 4는 상기 순방향 DCT2 또는 상기 역방향 1X72에 후처리 행렬 ^ 및 선처리 행렬 을 적용하는 것을 특징으로 한다. 여기서, 2003
Figure imgf000008_0002
Figure imgf000008_0001
Figure imgf000008_0003
사이즈를 나타냄. 2019/190284 1»(:1^1{2019/003743
7 본 발명에서, 상기 수직 변환이 상기 14일 때 상기 14의 역변환을 열마다 적용하고, 상기 수평 변환이 å)0:14일 때 상기 1)(714의 역변환을 행마다 적용하는 것을 특징으로 한다.
본 발명에서, 상기 변환 조합 (수평 변환, 수직 변환)은, 쨘計4, 0814), (0^4, 0814), (0814, 0014) 및 (1)<그4, 1)( 4)를 포함하는 것을 특징으로 한다. 본 발명에서, 상기 현재 블록이 인트라 예측된 레지듀얼일 때, 상기 변환 조합은 상기 변환 인덱스 0,1,2, 3에 대응되는 것을 특징으로 한다.
본 발명에서, 상기 현재 블록이 인터 예측된 레지듀얼일 때, 상기 변환 조합은 상기 변환 인덱스 3, 2, 1,0에 대응되는 것을 특징으로 한다.
본 발명은, 저복잡도 변환 실행에 기초하여 비디오 신호를 복원하는 장치에 있어서, 상기 비디오 신호로부터 현재 블록의 변환 인덱스를 획득하는 파싱부, 여기서 상기 변환 인덱스는 0814 및/또는 DC^4 의 조합으로 구성된 복수개의 변환 조합들 중 어느 하나에 대응됨; 상기 변환 인텍스에 대응되는 변환 조합을 유도하고, 4를 이용하여 상기 현재 블록에 대해 수직 방향으로 역방향 변환을 수행하고, 1X74를 이용하여 상기 현재 블록에 대해 수평 방향으로 역방향 변환을 수행하는 변환부, 여기서 상기 변환 조합은 수평 변환 및 수직 변환으로 구성되고, 상기 수평 변환 및 상기 수직 변환은 상기 0814 또는 상기 DCT4 중 어느 하나에 대응됨; 상기 역방향 변환된 현재 블록을 이용하여 상기 비디오 신호를 복원하는 복원부를 포함하는 것을 특징으로 하는 장치를 제공한다.
【발명의 실시를 위한 형태】
이하, 첨부된 도면을 참조하여 본 발명의 실시예의 구성과 그 작용을 설명하며, 도면에 의해서 설명되는 본 발명의 구성과 작용은 하나의 실시예로서 2019/190284 1»(:1^1{2019/003743
8 설명되는 것이며, 이것에 의해서 본 발명의 기술적 사상과 그 핵심 구성 및 작용이 제한되지는 않는다.
아울러, 본 발명에서 사용되는 용어는 가능한 한 현재 널리 사용되는 일반적인 용어를 선택하였으나, 특정한 경우는 출원인이 임의로 선정한 용어를 사용하여 설명한다. 그러한 경우에는 해당 부분의 상세 설명에서 그 의미를 명확히 기재하므로, 본 발명의 설명에서 사용된 용어의 명칭만으로 단순 해석되어서는 안 될 것이며 그 해당 용어의 의미까지 파악하여 해석되어야 함을 밝혀두고자 한다.
또한, 본 발명에서 사용되는 용어들은 발명을 설명하기 위해 선택된 일반적인 용어들이나, 유사한 의미를 갖는 다른 용어가 있는 경우 보다 적절한 해석을 위해 대체 가능할 것이다. 예를 들어, 신호, 데이터, 샘플, 픽쳐, 프레임, 블록 등의 경우 각 코딩 과정에서 적절하게 대체되어 해석될 수 있을 것이다. 또한, 파티셔닝 (partitioning), 분해 (decomposition), 스플리팅 (splitting) 및 분할 (division) 등의 경우에도 각 코딩 과정에서 적절하게 대체되어 해석될 수 있을 것이다. _
본 문서에서 MTS(Multiple Transform Selection, 이하 ‘MTS’라 함)라 함은 적어도 2개 이상의 변환 타입을 아용하여 변환을 수행하는 방법을 의미할 수 있다. 이는 AMT(Adaptive Multiple Transform) 또는 EMT (Explicit Multiple Transform)로도 표현될 수 있으며, 마찬가지로, mtsjdx 도 AMTJdx, EMT_idx, tu_mts_idx, AMT_TU_idx, EMT_TU_idx, 변환 인덱스 또는 변환 조합 인덱스 등과 같이 표현될 수 있으며, 본 발명은 이러한 표현에 한정되지 않는다. 2019/190284 1»(:1^1{2019/003743
9 도 1은 본 발명이 적용되는 실시예로서, 비디오 신호의 인코딩이 수행되는 인코더의 개략적인 블록도를 나타낸다.
도 1을 참조하면, 인코더 (100)는 영상 분할부 (1 10), 변환부 (120), 양자화부 (130), 역양자화부 (140), 역변환부 (150), 필터링부 (160), 복호 픽쳐 버퍼 (DPB: Decoded Picture Buffer)(170), 인터 예측부 (180), 인트라 예측부 (185) 및 엔트로피 인코딩부 (190)를 포함하여 구성될 수 있다.
영상 분할부 (1 10)는 인코더 (100)에 입력된 입력 영상 (Input image) (또는, 픽쳐, 프레임)를 하나 이상의 처리 유닛으로 분할할 수 있다. 예를 들어, 상기 처리 유닛은 코딩 트리 유닛 (CTU: Coding Tree Unit), 코딩 유닛 (CU: Coding Unit), 예측 유닛 (PU: Prediction Unit) 또는 변환 유닛 (TU: Transform Unit)일 수 있다.
다만, 상기 용어들은 본 발명에 대한 설명의 편의를 위해 사용할 뿐이며, 본 발명은 해당 용어의 정의에 한정되지 않는다. 또한, 본 명세서에서는 설명의 편의를 위해, 비디오 신호를 인코딩 또는 디코딩하는 과정에사 이용되는 단위로써 코딩 유닛이라는 용어를 사용하지만, 본 발명은 그에 한정되지 않으며 발명 내용에 따라 적절하게 해석 가능할 것이다.
인코더 (100)는 입력 영상 신호에서 인터 예측부 (180) 또는 인트라 예측부 (185)로부터 출력된 예측 신호 (prediction signal)를 감산하여 잔여 신호 (residual signal)를 생성할 수 있고, 생성된 잔여 신호는 변환부 (120)로 전송된다.
변환부 (120)는 잔여 신호에 변환 기법을 적용하여 변환 계수 (transform coefficient)를 생성할 수 있다. 변환 과정은 쿼드트리 (quadtree) 구조의 정사각형 블록, 바이너리트리 (binarytree) 구조, 삼진트리 (ternary) 구조 또는 비대징 2019/190284 1»(:1^1{2019/003743
10 트리 (asymmetric) 구조에 의해 분할된 블록 (정사각형 또는 직사각형)에 적용될 수 있다.
상기 변환부 (120)는 복수개의 변환 (또는 변환 조합들)에 기초하여 변환을 수행할 수 있으며, 이러한 변환 방식을 MTS(Multiple Transform Selection) 라 부를 수 있다. 상기 MTS는 AMT( Adaptive Multiple Transform) 또는 EMT(Enhanced Multiple Transform)로 부를 수도 있다.
상기 MTS (또는 AMT, EMT)는 복수개의 변환 (또는 변환 조합들)로부터 적응적으로 선택되는 변환 (또는 변환 조합들)에 기초하여 수행되는 변환 방식을 의미할 수 있다.
상기 복수개의 변환 (또는 변환 조합들)은 본 명세서의 도 6 및 도 26 내지 도 27에서 설명하는 변환 (또는 변환 조합들)을 포함할 수 있다. 본 명세서에서, 상기 변환 또는 변환 타입은, 예를 들어 DCT-Type 2, DCT-II, DCT2, DCT2 와 같이 표기할 수 있다.
상기 변환부 (120)는 다음의 실시예들을 수행할 수 있다.
본 발명은, DST4(Discrete Sine Transform-4) 및 DCT4(Discrete Cosine Transform-4)를 순방향 DCT2 (forward DCT2)로 수행하는 방법을 제공한다.
본 발명은, DST4 및 DCT4를 역방향 DCT2 (inverse DCT2)로 수행하는 방법을 제공한다.
본 발명은, DST4 및 DCT4를 MTS(Multiple Transform Selection)가 적용되는 변환 설정 그룹 (transform configuration group)에 적용하는 방법을 제공한다.
이에 대한 구체적인 실시예들은 본 명세서에서 보다 상세히 설명하도록 한다. 2019/190284 1»(:1^1{2019/003743
1 1 양자화부 (130)는 변환 계수를 양자화하여 엔트로피 인코딩부 (190)로 전송하고, 엔트로피 인코딩부 (190)는 양자화된 신호 (quantized signal)를 엔트로피 코딩하여 비트스트림으로 출력할 수 있다.
상기 변환부 (120)와 상기 양자화부 (130)는 별개의 기능 유닛으로 설명되지만, 본 발명은 이에 한정되지 않으며, 하나의 기능 유닛으로 결합될 수 있다. 상기 역양자화부 (140)와 상기 역변환부 (150)의 경우에도, 마찬가지로 하나의 기능 유닛으로 결합될 수 있다.
양자화부 (130)로부터 출력된 양자화된 신호 (quantized signal)는 예즉 신호를 생성하기 위해 이용될 수 있다. 예를 들어, 양자화된 신호 (quantized signal)는 루프 내의 역양자화부 (140) 및 역변환부 (150)를 통해 역양자화 및 역변환을 적용함으로써 잔여 신호를 복원할 수 있다. 복원된 잔여 신호를 인터 예측부 (180) 또는 인트라 예측부 (185)로부터 출력된 예측 신호 (prediction signal)에 더함으로써 복원 신호 (reconstructed signal)가 생성될 수 있다.
한편, 위와 같은 압축 과정에서 발생한 양자화 에러에 의해 블록 경계가 보이는 열화가 발생될 수 있다. 이러한 현상을 블록킹 열화 (blocking artifacts)라고 하며, 이는 화질을 평가하는 중요한 요소 중의 하나이다. 이러한 열화를 줄이기 위해 필터링 과정을 수행할 수 있다. 이러한 필터링 과정을 통해 블록킹 열화를 제거함과 동시에 현재 픽쳐에 대한 오차를 줄임으로써 화질을 향상시킬 수 있게 된다.
필터링부 (160)는 복원 신호에 필터링을 적용하여 이를 재생 장치로 출력하거나 복호 픽쳐 버퍼 (170)에 전송한다. 복호 픽쳐 버퍼 (170)에 전송된 필터링된 신호는 인터 예측부 (180)에서 참조 픽쳐로 사용될 수 있다. 이처럼, 2019/190284 1»(:1^1{2019/003743
12 필터링된 픽쳐를 화면간 예측 모드에서 참조 픽쳐로 이용함으로써 화질 뿐만 아니라부호화 효율도 향상시킬 수 있다.
복호 픽쳐 버퍼 (170)는 필터링된 픽쳐를 인터 예측부 (180)에서의 참조 픽쳐로사용하기 위해 저장할수 있다.
인터 예즉부 (180)는 복원 픽쳐 (reconstructed picture)를 참조하여 시간적 중복성 및/또는 공간적 중복성을 제거하기 위해 시간적 예측 및/또는 공간적 예측을 수행한다. 여기서, 예측을 수행하기 위해 이용되는 참조 픽쳐는 이전 시간에 부호화/복호화 시 블록 단위로 양자화와 역양자화를 거친 변환된 신호이기 때문에, 블로킹 아티팩트 (blocking artifact)나 링잉 아티팩트 (ringing artifact)가존재할수 있다.
따라서, 인터 예측부 (180)는 이러한 신호의 불연속이나 양자화로 인한 성능 저하를 해결하기 위해, 로우패스 필터 (lowpass filter)를 적용함으로써 픽셀들 사이의 신호를 서브 픽셀 단위로 보간할 수 있다. 여기서, 서브 픽셀은 보간 필터를 적용하여 생성된 가상의 화소를 의미하고, 정수 픽셀은 복원된 픽쳐에 존재하는 실제 화소를 의미한다. 보간 방법으로는 선형 보간, 양선형 보간 (bi linear interpolation), 위너 필터 (wiener filter)등이 적용될 수 있다.
보간 필터는 복원 픽쳐 (reconstructed picture)에 적용되어 예즉의 정밀도를 향상시킬 수 있다. 예를 들어, 인터 예측부 (180)는 정수 픽셀에 보간 필터를 적용하여 보간 픽셀을 생성하고, 보간 픽셀들 (interpolated pixels)로 구성된 보간 블록 (interpolated block)을 예즉 블록 (prediction block)으로 사용하여 예즉을 수행할 수 있다.
한편, 인트라 예측부 (185)는 현재 부호화를 진행하려고 하는 블록의 2019/190284 1»(:1^1{2019/003743
13 주변에 있는 샘플들을 참조하여 현재 블록을 예즉할 수 있다. 상기 인트라 예측부 (185)는 인트라 예측을 수행하기 위해 다음과 같은 과정을 수행할 수 있다. 먼저, 예측 신호를 생성하기 위해 필요한 참조 샘플을 준비할 수 있다. 그리고, 준비된 참조 샘플을 이용하여 예측 신호를 생성할 수 있다. 이후, 예측 모드를 부호화하게 된다. 이때, 참조 샘플은 참조 샘플 패딩 및/또는 참조 샘플 필터링을 통해 준비될 수 있다. 참조 샘플은 예측 및 복원 과정을 거쳤기 때문에 양자화 에러가 존재할 수 있다. 따라서, 이러한 에러를 줄이기 위해 인트라 예측에 이용되는 각 예측 모드에 대해 참조 샘플 필터링 과정이 수행될 수 있다.
상기 인터 예측부 (180) 또는 상기 인트라 예측부 (185)를 통해 생성된 예측 신호 (prediction signal)는 복원 신호를 생성하기 위해 이용되거나 잔여 신호를 생성하기 위해 이용될 수 있다. 도 2는 본 발명이 적용되는 실시예로서, 비디오 신호의 디코딩이 수행되는 디코더의 개략적인 블록도를 나타낸다.
도 2를 참조하면, 디코더 (200)는 파싱부 (미도시), 엔트로피 디코딩부 (210), 역양자화부 (220), 역변환부 (230), 필터링부 (240), 복호 픽쳐 버퍼 (DPB: Decoded Picture Buffer Unit)(250), 인터 예측부 (260) 및 인트라 예측부 (265)를 포함하여 구성될 수 있다.
그리고, 디코더 (200)를 통해 출력된 복원 영상 신호 (reconstructed video signal)는 재생 장치를 통해 재생될 수 있다.
디코더 (200)는 도 1의 인코더 (100)로부터 출력된 신호을 수신할 수 있고, 수신된 신호는 엔트로피 디코딩부 (210)를 통해 엔트로피 디코딩될 수 있다, 2019/190284 1»(:1^1{2019/003743
14 역양자화부 (220)에서는 양자화 스텝 사이즈 정보를 이용하여 엔트로피 디코딩된 신호로부터 변환 계수 (transform coefficient)를 획득한다.
역변환부 (230)에서는 변환 계수를 역변환하여 잔여 신호 (residual signal)를 획득하게 된다.
여기서, 본 발명은 예측 모드, 블록 크기 또는 블록 모양 (block shape) 중 적어도 하나에 의해 구분되는 변환 설정 그룹 (transform configuration group) 별로 변환 조합 (transform combination)을 구성하는 방법을 제공하며, 상기 역변환부 (230)는 본 발명에 의해 구성된 변환 조합에 기초하여 역변환을 수행할 수 있다. 또한, 본 명세서에서 설명한 실시예들이 적용될 수 있다.
상기 역변환부 (230)는 다음의 실시예들을 수행할 수 있다.
본 발명은, DST4(Discrete Sine Transform-4) 및 DCT4(Discrete Cosine Transform-4)를 순방향 DCT2 (forward DCT2)로 수행하는 방법을 제공한다.
본 발명은, DST4 및 DCT4를 역방향 DCT2 (inverse DCT2)로 수행하는 방법을 제공한다.
본 발명은, DST4 및 DCT4를 MTS(Multiple Transform Selection)가 적용되는 변환설정 그룹 (transform configuration group)에 적용하는 방법을 제공한다.
본 발명은, 저복잡도 변환 실행에 기초하여 비디오 신호를 복원하는 방법에 있어서, 상기 비디오 신호로부터 현재 블록의 변환 인덱스를 획득하는 단계, 여기서 상기 변환 인덱스는 DST4 및/또는 DCT4 의 조합으로 구성된 복수개의 변환 조합들 중 어느 하나에 대응됨; 상기 변환 인덱스에 대응되는 변환 조합을 유도하는 단계, 여기서 상기 변환 조합은 수평 변환 및 수직 변환으로 구성되고, 상기 수평 변환 및 상기 수직 변환은 상기 DST4 또는 상기 2019/190284 1»(:1/10公019/003743
1 5
0014 중 어느 하나에 대응됨; 상기 14를 이용하여, 상기 현재 블록에 대해 수직 방향으로 역방향 변환을 수행하는 단계; 상기 £>( 4를 이용하여, 상기 현재 블록에 대해 수평 방향으로 역방향 변환을 수행하는 단계; 및 상기 역방향 변환된 현재 블록을 이용하여 상기 비디오 신호를 복원하는 단계를 포함하는 것을 특징으로 하는 방법을 제공한다.
본 발명에서, 상기 14 및/또는 상기 1)(그14는 순방향 £>(그2 또는 역방향 1)0:72를 이용하여 실행되는 것을 특징으로 한다.
본 발명에서, 상기 0814 및/또는 상기 1X714는 상기 순방향 0012 또는 상기 역방향 1) 12에 후처리 행렬 및 선처리 행렬 서 을 적용하는 것을
Figure imgf000017_0001
사이즈를 나타냄.
본 발명에서, 상기 수직 변환이 상기 14일 때 상기 14의 역변환을 열마다 적용하고, 상기 수평 변환이 1X^4일 때 상기 1X^4의 역변환을 행마다 적용하는 것을 특징으로 한다.
본 발명에서, 상기 변환 조합 (수평 변환, 수직 변환)은, 必 4, 0^4), (0014, 0814),(0814, 0014) 및 (0 4, 1) 14)를 포함하는 것을 특징으로 한다. 본 발명에서, 상기 현재 블록이 인트라 예측된 레지듀얼일 때, 상기 변환 조합은 상기 변환 인덱스 0,1,2, 3에 대응되는 것을 특징으로 한다. 2019/190284 1»(:1^1{2019/003743
16 본 발명에서, 상기 현재 블록이 인터 예측된 레지듀얼일 때, 상기 변환 조합은 상기 변환 인덱스 3, 2, 1,0에 대응되는 것을 특징으로 한다.
상기 역양자화부 (220)와 상기 역변환부 (230)는 별개의 기능 유닛으로 설명되지만, 본 발명은 이에 한정되지 않으며, 하나의 기능 유닛으로 결합될 수 있다.
획득된 잔여 신호를 인터 예측부 (260) 또는 인트라 예측부 (265)로부터 출력된 예측 신호 (prediction signal)에 더함으로써 복원 신호 (reconstructed signal)가 생성된다.
필터링부 (240)는 복원 신호 (reconstructed signal)에 필터링을 적용하여 이를 재생 장치로 출력하거나 복호 픽쳐 버퍼부 (250)에 전송한다. 복호 픽쳐 버퍼부 (250)에 전송된 필터링된 신호는 인터 예측부 (260)에서 참조 픽쳐로 사용될 수 있다.
.명세서에서, 인코더 (100)의 변환부 (120) 및 각 기능 유닛들에서 설명된 실시예들은 각각 디코더의 역변환부 (230) 및 대응되는 기능 유닛들에도 동일하게 적용될 수 있다. 도 3은 본 발명이 적용될 수 있는 실시예들로서, 도 3A는 QT(QuadTree, 이하 ‘QT라 함), 도 3B는 BT(Binary Tree, 이하 ‘BT,라 함), 도 3C는 TT(Ternary Tree, 이하 ‘TT’라 함) 도 3D는 AT(Asymmetric Tree, 이하 ‘AT’라 함)에 의한 블록 분할 구조들을 설명하기 위한 도면이다.
비디오 코딩에서 하나의 블록은 QT(QuadTree) 기반으로 분할될 수 있다. 또한, QT에 의해서 분할된 하나의 서브 블록 (sub block)은 QT를 사용하여 2019/190284 1»(:1^1{2019/003743
17 재귀적으로 더 분할될 수 있다. 더 이상 QT 분할되지 않는 리프 블록 (leaf block)은 BT(B inary Tree), TT(Temary Tree) 또는 AT (Asymmetric Tree) 중 적어도 하나의 방식에 의해서 분할될 수 있다. BT는 horizontal BT (2NxN, 2NxN)과 vertical BT (Nx2N, Nx2N)의 두 가지 형태의 분할을 가질 수 있다. TT는 horizontal TT (2Nxl/2N, 2NxN, 2Nxl/2N)와 vertical TT (l/2Nx2N, Nx2N, l/2Nx2N)의 두 가지 형태의 분할을 가질 수 있다. AT는 horizontal-up AT (2Nxl/2N, 2Nx3/2N), horizontal- down AT (2Nx3/2N, 2Nxl/2N), vertical-left AT (l /2Nx2N, 3/2Nx2N), vertical-right AT (3/2Nx2N, l/2Nx2N) 의 네 가지 형태의 분할을 가질 수 있다. 각각의 BT, TT, AT는 BT,TT, AT를 사용하여 재귀적으로 더 분할될 수 있다.
상기 도 3A는 QT 분할의 예를 보여준다. 블록 A는 QT에 의해서 4개의 서브 블록 (AO, Al , A2, A3)으로 분할될 수 있다. 서브 블록 시은 다시 아에 의해서 4개의 서브 블록 (B0,B 1,B2,B3)로 분할될 수 있다.
상기 도 3B는 BT 분할의 예를 보여준다. (기에 의해서 더 이상 분할되지 않는 블록 B3은 vertical BT (CO, Cl) 또는 horizontal BT (DO, Dl)으로 분할될 수 있다. 블록 C0와 같이 각각의 서브 블록은 horizontal BT (E0, El)또는 vertical BT (F0,F1)의 형태와 같이 재귀적으로 더 분할될 수 있다.
상기 3C는 TT 분할의 예를 보여준다. QT에 의해서 더 이상 분할되지 않는 블록 B3은 vertical TT (CO, Cl , C2) 또는 horizontal TT (DO, Dl, D2)으로 분할될 수 있다. 블록 Cl와 같이 각각의 서브 블록은 horizontal TT (E0, El , E2) 또는 vertical TT (F0,F1, F2)의 형태와 같이 재귀적으로 더 분할될 수 있다.
상가 3D는 AT 분할의 예를 보여준다. 아에 의해서 더 이상 분할되지 않는 블록 B3은 vertical AT (CO, Cl) 또는 horizontal AT (DO, Dl)으로 분할 될 수 2019/190284 1»(:1^1{2019/003743
18 있다.블록 C1와 같이 각각의 서브블록은 horizontal AT (E0, El)또는 vertical TT (F0, F1)의 형태와같이 재귀적으로 더 분할 될 수 있다. 한편, BT, TT, AT분할은 함께 사용하여 분할이 가능하다. 예를들어, BT에 의해 분할된 서브 블록은 TT또는 AT에 의한 분할이 가능하다. 또한, TT에 의해 분할된 서브 블록은 BT 또는 AT에 의한 분할이 가능하다. AT에 의해 분할된 서브블록은 BT또는 TT에 의한 분할이 가능하다. 예를 들어, horizontal BT분할 이후, 각각의 서브 블록이 vertical BT로 분할될 수 있고, 또는 vertical BT 분할 이후,각각의 서브블록이 horizontal BT로분할될 수도 있다.상기 두 종류의 분할 방법은분할순서는 다르지만 최종적으로 분할되는 모양은동일하다.
또한, 블록이 분할되면 블록을 탐색하는 순서를 다양하게 정의할 수 있다. 일반적으로, 좌측에서 우측으로, 상단에서 하단으로 탐색을 수행하며, 블록을 탐색한다는 것은 각 분할된 서브 블록의 추가적인 블록 분할 여부를 결정하는 순서를 의미하거나, 블록이 더 이상 분할되지 않을 경우 각 서브 블록의 부호화 순서를 의미하거나, 또는 서브 블록에서 다른 이웃 블록의 정보를 참조할 때의 탐색 순서를 의미할수 있다. 도 4 및 도 5는 본 발명이 적용되는 실시예들로서, 도 4는 인코더 내의 변환 및 양자화부 (120/130), 역양자화 및 역변환부 (140/150)의 개략적인 블록도를 나타내고, 도 5는 디코더 내의 역양자화 및 역변환부 (220/230)의 개략적인 블록도를 나타낸다.
상기 도 4를 살펴보면, 변환 및 양자화부 (120/130)는 1차 변환부 (primary 2019/190284 1»(:1/10公019/003743
19 transform unit)(121), 2차 변환부 (secondary transform unit)(122) 및 양자화부 (130)을 포함할 수 있다. 역양자화 및 역변환부 (140/150)는 역양자화부 (140), 역 2차 변환부 (inverse secondary transform unit)(151) 및 역 1차 변환부 (inverse primary transform unit)( 152)를 포함할 수 있다.
상기 도 5를 살펴보면, 역양자화 및 역변환부 (220/230)는 역양자화부 (220), 역 2차 변환부 (inverse secondary transform unit)(231) 및 역 1차 변환부 (inverse primary transform unit)(232)를 포함할 수 있다.
본 발명에서, 변환을 수행할 때 복수의 단계를 통해 변환을 수행할 수 있다. 예를 들어, 도 4에서와 같이 1차 변환 (primary transform), 2차 변환 (secondary transform)의 2단계를 적용할 수 있고, 또는 알고리즘에 따라 그 이상의 변환 단계가 이용될 수도 있다. 여기서, 상기 1차 변환 (primary transform)은 주요 변환 (core transform)이라고 부를 수도 있다.
상기 1차 변환부 (primary transform unit)(121)는 잔여 신호에 대해 1차 변환 (primary transform)을 적용할 수 있으며, 여기서 상기 1차 변환 (primary transform)은 인코더 및/또는 디코더에서 테이블로 기정의될 수 있다.
상기 1차 변환 (primary transform)의 경우, Discrete Cosine Transform type 2(이하, "DCT2,라 함)가 적용될 수 있다. 또는, 특정한 경우에 한정하여 Discrete Sine Transform-type 7(이하, ‘DST7’ 이라 함)이 적용될 수도 있다. 예를 들어, 인트라 예측 모드에서 4x4 블록에 대해 DST7이 적용될 수 있다.
또한, 상기 1차 변환 (primary transform)의 경우, MTS(Multiple Transform Selection)의 여러 변환들 (DST 7, DCT 8, DST 1, DCT 5) 의 조합들이 적용될 수도 있다. 예를 들어, 도 6이 적용될 수있다 2019/190284 1»(:1^1{2019/003743
20 상기 2차 변환부 (secondary transform unit)(122)는 1차 변환된 신호에 대해 2차 변환 (secondary transform)을 적용할 수 있으며, 여기서 상기 2차 변환 (secondary transform)은 인코더 및/또는 디코더에서 테이블로 기정의될 수 있다.
일실시예로, 상기 2차 변환 (secondary transform)은 Non-Separable Secondary Transform (이하, ‘NSST,라 함)이 조건적으로 적용될 수 있다. 예를 들어, 상기 NSST는 인트라 예측 블록일 경우에만 적용되며, 예측 모드 그룹마다 적용 가능한 변환 셋을 가질 수 있다.
여기서, 상기 예측 모드 그룹은 예측 방향에 대한 대칭성에 기초하여 설정될 수 있다. 예를 들어, 예측 모드 52와 예측 모드 16은 예측 모드 34 (대각 방향)를 기준으로 대칭이므로, 하나의 그룹을 형성하여 동일한 변환 셋 (transform set)이 적용될 수 있다. 이때, 상기 예측 모드 52에 대한 변환을 적용할 때 입력 데이터를 트랜스포즈 (仕 anspose)한 후 적용하며, 이는 예측 모드 16과 변환 셋이 동일하기 때문이다.
한편, 플래너 모드 (Planar mode)와 DC 모드 (DC mode)의 경우 방향에 대한 대칭성이 존재하지 않으므로 각자의 변환 셋을 가지며, 해당 변환 셋은 2개의 변환들로 구성될 수 있다. 나머지 방향성 모드에 대해서는 변환 셋마다 3개의 변환들로 구성될 수 있다.
다른 실시예로, 상기 NSST는 1차 변환된 블록 전체에 대해 적용되는 것이 아니라 좌상측 (top-left) 8x8 영역에 대해서만 적용될 수 있다. 예를 들어, 블록의 크기가 8x8 이상인 경우는 8x8 NSST가 적용되며, 8x8 미만인 경우는 4x4 NSST7> 적용되는데, 이때 4x4 블록들로 나눈 후 각기 4x4 NSST가 적용된다.
다른 실시예로, 4xN/Nx4 (N ñ= 16)인 경우에도 4x4 NSST가 적용될 수 있다. () 2019/190284 1»(:1/10公019/003743
21 상기 양자화부 (130)는 2차 변환된 신호에 대해 양자화를 수행할 수 있다. 상기 역양자화 및 역변환부 (140/150)는 앞서 설명된 과정을 역으로 수행하며, 중복되는 설명은 생략하도록 한다. 도 5는 디코더 내의 역양자화 및 역변환부 (220/230)의 개략적인 블록도를 나타낸다.
상기 도 5를 살펴보면, 역양자화 및 역변환부 (220/230)는 역양자화부 (220), 역 2차 변환부 (inverse secondary transform unit)(231) 및 역 1차 변환부 (inverse primary transform unit)(232)를 포함할 수 있다.
상기 역양자화부 (220)에서는 양자화 스텝 사이즈 정보를 이용하여 엔트로피 디코딩된 신호로부터 변환 계수 (transform coefficient)를 획득한다.
상기 역 2차 변환부 (inverse secondary transform unit)(231)에서는 상기 변환 계수에 대해 역 2차 변환을 수행한다. 여기서, 상기 역 2차 변환은 상기 도 4에서 설명한상기 2차 변환 (secondary transform)의 역변환을 나타낸다.
상기 역 1차 변환부 (inverse primary transform unit)(232)는 역 2차 변환된 신호 (또는 블록)에 대해 역 1차 변환을 수행하고, 잔여 신호 (residual signal)를 획득하게 된다. 여기서, 상기 역 1차 변환은 상기 도 4에서 설명한 상기 1차 변환 (primary transform)의 역변환을 나타낸다.
본 발명은 예측 모드, 블록 크기 또는 블록 모양 (block shape) 중 적어도 하나에 의해 구분되는 변환 설정 그룹 (transform configuration group) 별로 변환 조합 (transform combination)을 구성하는 방법을 제공하며, 상기 역 1차 변환부 (inverse primary transform unit)(232)는 본 발명에 의해 구성된 변환 조합에 2019/190284 1»(:1^1{2019/003743
22 기초하여 역변환을 수행할 수 있다. 또한, 본 명세서에서 설명한 실시예들이 적용될 수 있다. 도 6은본 발명이 적용되는 실시예로서, MTS(Multiple Transform Selection)가 적용되는 변환설정 그룹 (transform configuration group)을 나타내는 표이다.
MTSfMultiple Transform Selection)가적용되는 변환설정 그룹
본 명세서에서는, 변환 설정 그룹 Gi에 대한 j번째 변환 조합후보를 다음 수학식 1과 같은쌍으로 표시한다.
【수학식 1 ]
(H(GiJ), V(GiJ))
여기서: H(Gi, j)는 j번째 후보에 대한 수평 변환 (horizontal transform)을 가리키며, V(Gi, j)는 j번째 후보에 대한 수직 변환 (vertical transform)을 가리킨다. 예를들어, 도 6에서 H(G3, 2) = DST7, V(G3J 2) - DCT8 과 같이 표기할수 있다. 문맥에 따라 H(Gi, j)또는 V(Gi, j)에 할당되는 값은상기 예시에서와 같이 변환을 구분하기 위한 표시상의 값 (nominal value)일 수도 있고 해당 변환을 가리키는 인덱스 값일 수도 있으며 해당 변환에 대한 2D 행렬 (2 Dimensional matrix)일 수도 있다.
또한, 본 명세서에서, DCT와 DST에 대한 2D 행렬 값을 다음 수학식 2 내지 3과 같이 표기할수 있다.
【수학식 2]
DCT type 2: , DCT type 8: Cfv
【수학식 3】 2019/190284 1»(:1^1{2019/003743
23
DST type 7: SN V" , DST type 4: SN !V
여기서, DST 또는 DCT인지 여부는 S 또는 C로 나타내고, type 숫자는 로마숫자의 형태로 위첨자로 표기하며, 하부첨자의 N는 NxN 변환임을 가리킨다. 또한, 상기 Cl 와 SN JV 와 같은 2D 행렬은 열 벡터 (column vector)들이 변환 기저 (transform basis)를 이루는 것을 가정한다. 상기 도 6을 살펴보면, 변환 설정 그룹 (transform configuration group)들은 예측 모드에 기초하여 결정되고 그룹의 개수는 총 6개 (G0 - G5)일 수 있다. 그리고, G0 - G4는 인트라 예측이 적용되는 경우에 해당되고 G5는 인터 예측에 의해 생성된 잔차 (residual) 블록에 적용되는 변환 조합들 (또는 변환 셋, 변환 조합 셋)을 나타낸다.
하나의 변환 조합은 해당 2D 블록의 행 (row)들에 적용되는 수평 변환 (horizontal transform) (또는 행 변환 (row transform)) 과 열 (column)들에 적용되는 수직 변환 (vertical transform) (또는 열 변환 (column transform)) 으로 이루어질 수 있다.
여기서, 모든 변환 설정 그룹들은 각 4개의 변환 조합 후보들을 가질 수 있다. 상기 4개의 변환 조합 후보들은 0〜 3의 변환 조합 인덱스를 통해 선택 또는 결정될 수 있으며, 인코더에서 디코더로 상기 변환 조합 인덱스를 인코딩하여 전송할수 있다.
일실시예로, 인트라 예측을 통해 획득된 잔차 데이터 (또는 잔차 신호)는 인트라 예측 모드에 따라 그 통계적 특성이 각기 다를 수 있다. 따라서, 상기 도
6에서와 같이 인트라 예측 모드 별로 일반적인 코사인 변환이 아닌 다른 2019/190284 1»(:1^1{2019/003743
24 변환들을 적용할 수 있다.
상기 도 6을 살펴보면, 35개의 인트라 예측 모드를 사용하는 경우와 67개의 인트라 예측 모드를 사용하는 경우를 나타낸다. 각 인트라 예측 모드 열에서 구분된 변환 설정 그룹 별로 복수개의 변환 조합들이 적용될 수 있다. 예를 들어, 상기 복수개의 변환 조합들은 4개의 (행 방향 변환, 열 방향 변환) 조합으로 구성될 수 있다. 구체적 예로, 그룹 0에서는 행 (수평) 방향과 열 (수직) 방향모두 DST-7과 DCT-5가 적용될 수 있어 총 4개의 조합이 가능하다.
각 인트라 예측 모드에 대해 총 4개의 변환 커널 조합이 적용될 수 있으므로, 그 중 하나를 선택하기 위한 변환 조합 인덱스가 변환 단위 (transform unit)마다 전송될 수 있다. 본 명세서에서, 상기 변환 조합 인덱스는 MTS 인덱스 (MTS index)라 부를 수 있으며, mts_idx로 표현될 수 있다.
또한, 상기 도 6에서 제시한 변환 커널들 이외에도, 잔차 신호 (residual signal)의 특성상 행 방향과 열 방향 모두에 대해 DCT2가 최적인 경우가 발생할 수 있다. 따라서, 코딩 유닛 (Coding Unit)마다 MTS 플래그를 정의함으로써 적응적으로 변환을 수행할 수 있다. 여기서, MTS 플래그가 0이면 행 방향과 열 방향 모두에 대해 DCT2를 적용하고, MTS 플래그가 1이면 MTS 인덱스를 통해 4개의 조합 중 하나를 선택 또는 결정할 수 있다.
일실시예로, MTS 플래그가 1인 경우 하나의 변환 단위에 대해 0이 아닌 변환 계수의 개수가 임계값보다 크지 않으면 상기 도 6의 변환 커널들을 적용하지 않고 행 방향과 열 방향에 대해 모두 DST-7을 적용할 수 있다. 예를 들어, 상기 임계값은 2로 설정될 수 있으며, 이는 블록 크기 또는 변환 유닛의 크기에 기초하여 다르게 설정될 수 있다. 이는 명세서 다른 실시예에서도 적용 2019/190284 1»(:1^1{2019/003743
25 가능하다.
일실시예로, 변환 계수 값들을 먼저 파싱하여 0이 아닌 변환 계수의 개수가 임계값보다 크지 않으면, MTS 인덱스를 파싱하지 않고 DST-7을 적용함으로써 부가 정보 전송량을 줄일 수 있다.
일실시예로, MTS 플래그가 1인 경우 하나의 변환 단위에 대해 0이 아닌 변환 계수의 개수가 임계값보다 크면 MTS 인덱스를 파싱하고, MTS 인텍스에 기초하여 수평 변환과 수직 변환을 결정할 수 있다.
일실시예로, MTS는 변환 단위의 너비와 높이가 모두 32 이하일 경우에만 적용될 수 있다.
일실시예로, 상기 도 6은 오프라인 트레이닝 (off-line training)을 통해 기설정될 수 있다.
일실시예로, 상기 MTS 인덱스는 수평 변환과 수직 변환의 조합을 동시에 가리킬 수 있는 하나의 인덱스로 정의될 수 있다. 또는, 상기 MTS 인덱스는 수평 변환 인덱스와 수직 변환 인덱스를 별도로 정의할 수도 있다.
일실시예로, 상기 MTS 플래그 또는 상기 MTS 인덱스는 시퀀스, 픽쳐, 슬라이스, 블록, 코딩 유닛, 변환 유닛, 또는 예측 유닛 중 적어도 하나의 레벨에서 정의될 수 있다. 예를 들어, 상기 MTS 플래그 또는 상기 MTS 인덱스는 SPS(Sequence Parameter Set) 또는 변환 유닛 중 적어도 하나의 레벨에서 정의될 수 있다. 도 7은 본 발명이 적용되는 실시예로서, MTS(Multiple Transform Selection)가 수행되는 인코딩 과정을 나타내는 흐름도이다. 2019/190284 1»(:1/10公019/003743
26 본 명세서에서는 기본적으로 수평 방향과 수직 방향에 대해 분리해서 변환을 적용하는 실시예를 설명하고 있으나, 변환 조합은 비분리 변환 (non- separable transform)들로도 구성이 가능하다.
또는, 분리가능한 변환 (separable transform)들과 비분리 변환 (non-separable transform)들의 혼합으로도 구성될 수 있다. 이 경우, 비분리 변환 (non-separable transform)0] 이용되면 행/열 (row/column)별 변환 선택이나 수평/수직 (horizontal/vertical) 방향별 선택은 불필요해지며, 분리가능한 변환 (separable transform) °1 선택되는 경우에만 상기 도 6의 변환 조합들이 이용될 수 있다.
또한, 본 명세서에서 제안하는 방식들은 1차 변환 (primary transform)이나 2차 변환 (secondary transform)에 관계 없이 적용할 수 있다. 즉, 둘 중 어느 한쪽에만 적용되어야 한다는 제약이 없으며, 둘다 적용될 수 있다. 여기서, 상기 1차 변환 (primary transform)은 잔차 (residual) 블록을 맨 처음에 변환하기 위한 변환을 의미할 수 있고, 상기 2차 변환 (secondary transform)은 상기 1차 변환의 결과로 생성된 블록에 대해 변환을 적용하기 위한 변환을 의미할 수 있다.
먼저, 인코더는 현재 블록에 대응되는 변환 설정 그룹을 결정할 수 있다 (S710). 여기서, 상기 변환 설정 그룹은 상기 도 6의 변환 설정 그룹을 의미할 수 있으나, 본 발명은 이에 한정되지 않으며 다른 변환 조합들로 구성될 수도 있다,
상기 인코더는 상기 변환 설정 그룹 내 이용가능한 후보 변환 조합들에 대해 변환을 수행할 수 있다 (S720).
상기 변환수행 결과, 상기 인코더는 RD(Rate Distortion) 비용이 가장 작은 2019/190284 1»(:1^1{2019/003743
27 변환 조합을 결정 또는 선택할 수 있다 (S730).
상기 인코더는 상기 선택된 변환 조합에 대응되는 변환 조합 인덱스를 인코딩할 수 있다 (S740). 도 8은 본 발명이 적용되는 실시예로서, MTS(Multiple Transform Selection)가 수행되는 디코딩 과정을 나타내는 흐름도이다.
먼저, 디코더는 현재 블록을 위한 변환 설정 그룹을 결정할 수 있다 (S810). 상기 디코더는 비디오 신호로부터 변환 조합 인덱스를 파싱 (또는 획득)할 수 있으며, 여기서 상기 변환 조합 인덱스는 상기 변환 설정 그룹 내 복수개의 변환 조합들 중 어느 하나에 대응될 수 있다 (S820). 예를 들어, 상기 변환 설정 그룹은 DST7(Discrete Sine Transform type 7) 및 DCT8(Discrete Cosine Transform type 8)을 포함할 수 있다. 상기 변환 조합 인덱스는 MTS 인덱스로 부를 수 있다. 일실시예로, 상기 변환 설정 그룹은 현재 블록의 예측 모드, 블록 크기 또는 블록 모양 중 적어도 하나에 기초하여 설정될 수 있다.
상기 디코더는 상기 변환 조합 인덱스에 대응되는 변환 조합을 유도할 수 있다 (S830). 여기서, 상기 변환 조합은 수평 변환 및 수직 변환으로 구성되고, 상기 DST-7 또는 DCT-8 중 적어도 하나를 포함할 수 있다.
또한, 상기 변환 조합은 상기 도 6에서 설명한 변환 조합을 의미할 수 있으나, 본 발명은 이에 한정되지 않는다. 즉, 본 명세서 내 다른 실시예에 따른 다른 변환조합에 의한 구성도 가능하다.
상기 디코더는 상기 변환 조합에 기초하여 상기 현재 블록에 대해 역변환을 수행할 수 있다 (SM0). 상기 변환 조합이 행 (수평) 변환과 열 (수직) 2019/190284 1»(:1^1{2019/003743
28 변환으로 구성된 경우, 행(수평) 변환을 먼저 적용한 후 열(수직) 변환을 적용할 수 있다. 다만, 본 발명은 이에 한정되지 않으며, 반대로 적용되거나, 또는 비분리 변환들로 구성된 경우에는 바로 비분리 변환이 적용될 수 있다.
일실시예로, 상기 수직 변환 또는 상기 수평 변환이 상기 081-7 또는 인 경우, 상기 0 -7의 역변환 또는 0(그-8의 역변환을 열마다 적용하고 난 이후 행마다 적용할 수 있다.
일실시예로, 상기 수직 변환 또는 상기 수평 변환은, 각 행마다 및/또는 각 열마다 다른 변환이 적용될 수 있다.
일실시예로, 상기 변환 조합 인덱스는 1 18가 수행되는지 여부를 나타내는 요 플래그에 기초하여 획득될 수 있다. 즉, 상기 변환 조합 인덱스는 상기 MTS 플래그에
Figure imgf000030_0001
수행되는 경우에 획득될 수 있다.
일실시예로, 상기 디코더는 0이 아닌 변환 계수의 개수가 임계값보다 큰지 여부를 확인할 수 있다. 이때, 상기 변환 조합 인덱스는 상기 0이 아닌 변환 계수의 개수가 임계값보다 큰 경우에 획득될 수 있다.
일실시예로, 상기 MTS 플래그 또는 상기
Figure imgf000030_0002
인덱스는 시퀀스, 픽쳐, 슬라이스, 블록, 코딩 유닛, 변환 유닛, 또는 예측 유닛 중 적어도 하나의 레벨에서 정의될 수 있다.
일실시예로, 상기 역변환은 변환 단위의 너비와 높이가 모두 32 이하일 경우에만 적용될 수 있다.
한편, 다른 일실시예로, 변환 설정 그룹을 결정하는 과정과 변환 조합 인덱스를 파싱하는 과정은 동시에 수행될 수 있다. 또는 상기 8810 단계는 인코더 및/또는 디코더에서 기설정되어 생략될 수 있다. 2019/190284 1»(:1^1{2019/003743
29
도 9는 본 발명이 적용되는 실시예로서, MTS 플래그 및 MTS 인덱스를 인코딩하는 과정을 설명하기 위한흐름도이다.
인코더는현재 블록에 대해 MTS(Multiple Transform Selection)가 적용되는지 여부를 결정할수 있다 (S910).
만약, MTS(Multiple Transform Selection)가 적용되는 경우, 상기 인코더는 MTS플래그 = 1 로 인코딩할수 있다 (S920).
그리고,상기 인코더는상기 현재 블록의 예측모드,수평 변환, 수직 변환 중 적어도 하나에 기초하여 MTS 인덱스를 결정할 수 있다 (S930). 여기서, MTS 인덱스는 각 인트라 예측 모드에 대해 복수개의 변환 조합들 중 어느 하나를 가리키는 인덱스를 의미하며, 상기 MTS 인덱스는 변환 단위 (transform unit)마다 전송될 수 있다.
상기 MTS 인덱스가 결정되면, 상기 인코더는 상기 MTS 인덱스를 인코딩할수 있다 (S940).
한편, 상기 MTS(Multiple Transform Selection)가 적용되지 않는 경우, 상기 인코더는 MTS 플래그 = 0으로 인코딩할수 있다 (S950). 도 10은 본 발명이 적용되는 실시예로서, MTS 플래그 및 MTS 인덱스에 기초하여 수평 변환 또는 수직 변환을 행 또는 열에 적용하는 디코딩 과정을 설명하기 위한흐름도이다.
디코더는 비트스트림으로부터 MTS 플래그를 파싱할 수 있다 (S1010). 여기서, MTS 플래그는 현재 블록에 대해 MTS(Multiple Transform Selection)가 \\ ) 2019/190284 1»(:1^1{2019/003743
30 적용되는지 여부를 나타낼 수 있다.
상기 디코더는 상기 MTS 플래그에 기초하여 현재 블록에 대해 MTS(Multiple Transform Selection)가 적용되는지 여부를 확인할 수 있다 (S1020). 예를 들어, 상기 MTS 플래그가 1인지 여부를 확인할 수 있다.
만약, 상기 MTS 플래그가 1인 경우, 상기 디코더는 0이 아닌 변환 계수의 개수가 임계값보다 큰지 (또는 이상인지) 여부를 확인할 수 있다 (S1030). 예를 들어, 상기 임계값은 2 로 설정될 수 있으며, 이는 블록 크기 또는 변환 유닛의 크기에 기초하여 다르게 설정될 수 있다.
상기 0이 아닌 변환 계수의 개수가 임계값보다 큰 경우, 상기 디코더는 MTS 인덱스를 파싱할 수 있다 (S1040). 여기서, 상기 MTS 인덱스는 각 인트라 예측 모드 또는 인터 예측 모드에 대해 복수개의 변환 조합들 중 어느 하나를 가리키는 인덱스를 의미하며, 상기 MTS 인덱스는 변환 단위 (transform unit)마다 전송될 수 있다. 또는, 상기 MTS 인덱스는 기설정된 변환 조합 테이블에 정의된 어느 하나의 변환 조합을 가리키는 인덱스를 의미할 수 있으며, 여기서 상기 기설정된 변환 조합 테이블은 상기 도 6을 의미할 수 있으나, 본 발명은 이에 한정되지 않는다.
상기 디코더는 상기 MTS 인덱스 또는 예측 모드 중 적어도 하나에 기초하여 수평 변환 및 수직 변환을 유도 또는 결정할 수 있다 (S1050).
또는, 상기 디코더는 상기 MTS 인덱스에 대응되는 변환 조합을 유도할 수 있다. 예를 들어, 상기 디코더는 상기 MTS 인덱스에 대응되는 수평 변환 및 수직 변환을 유도 또는 결정할수 있다.
한편, 상기 0이 아닌 변환 계수의 개수가 임계값보다 크지 않은 경우, 2019/190284 1»(:1^1{2019/003743
31 상기 디코더는 기설정된 수직 역변환을 열마다 적용할 수 있다 ( 060). 예를 들어, 상기 수직 역변환은 17의 역변환일 수 있다.
그리고, 상기 디코더는 기설정된 수평 역변환을 행마다 적용할 수 있다 1070). 예를 들어, 상기 수평 역변환은 17의 역변환일 수 있다. 즉, 상기 0이 아닌 변환 계수의 개수가 임계값보다 크지 않은 경우에는 인코더 또는 디코더에서 기설정된 변환 커널이 이용될 수 있다. 예를 들어, 상기 도 6과 같은 변환 조합 테이블에서 정의된 것이 아니라, 많이 이용되는 변환 커널이 이용될 수 있다.
한편, 상기 1 18 플래그가 0인 경우, 상기 디코더는 기설정관 수직 역변환을 열마다 적용할 수 있다 1080). 예를 들어 , 상기 수직 역변환은 1) 12의 역변환일 수 있다.
그리고, 상기 디코더는 기설정된 수평 역변환을 행마다 적용할 수 있다 1090). 예를 들어 , 상기 수평 역변환은 0012의 역변환일 수 있다.
Figure imgf000033_0001
플래그가 0인 경우에는 인코더 또는 디코더에서 기설정된 변환 커널이 이용될 수 있다. 예를 들어, 상기 도 6과 같은 변환 조합 테이블에서 정의된 것이 아니라, 많이 이용되는 변환 커널이 이용될 수 있다. 도 1 1은 본 발명이 적용되는 실시예로서, 디코더 내 역변환부의 개략적인 블록도를 나타낸다.
본 발명이 적용되는 디코더는, 2차 역변환 적용 여부 결정부 (또는 2차 역변환의 적용 여부를 결정하는 요소) (1 1 10), 2차 역변환 결정부 (또는 2차 역변환을 결정하는 요소) (1 120), 2차 역변환부 (또는 2차 역변환을 수행하는 2019/190284 1»(:1^1{2019/003743
32 요소) (1 130), 1차 역변환부 (또는 1차 역변환을 수행하는 요소) (1 140)를 포함할 수 있다.
상기 2차 역변환 적용 여부 결정부 (1 1 10)는 2차 역변환의 적용 여부를 결정할 수 있다. 예를 들어, 2차 역변환은 Non-Separable Secondary Transform (이하, NSST) 또는 Reduced Secondary Transform (이하, RST)일 수 있다. 일 예로, 상기 2차 역변환 적용 여부 결정부 (1 1 10)는 인코더로부터 수신한 2차 변환 플래그에 기초하여 2차 역변환의 적용 여부를 결정할 수 있다. 다른 일 예로, 상기 2차 역변환 적용 여부 결정부 (1 1 10)는 레지듀얼 블록의 변환 계수에 기초하여 2차 역변환의 적용 여부를 결정할 수도 있다.
상기 2차 역변환 결정부 (1 120)는 2차 역변환을 결정할 수 있다. 이때, 상기 2차 역변환 결정부 (1 120)는 인트라 예측 모드에 따라 지정된 NSST (또는 RST) 변환 셋에 기초하여 현재 블록에 적용되는 2차 역변환을 결정할 수 있다.
또한, 일 실시예로서, 1차 변환 결정 방법에 기초하여 2차 변환 결정 방법이 결정될 수 있다. 인트라 예측 모드에 따라 1차 변환과 2차 변환의 다양한 여러 조합이 결정될 수 있다.
또한, 일 예로, 상기 2차 역변환 결정부 (1 120)는 현재 블록의 크기에 기초하여 2차 역변환이 적용되는 영역을 결정할 수도 있다.
상기 2차 역변환부 (1 130)는 결정된 2차 역변환을 이용하여 역양자화된 레지듀얼 블록에 대하여 2차 역변환을 수행할 수 있다.
상기 1차 역변환부 (1 140)는 2차 역변환된 레지듀얼 블록에 대하여 1차 역변환을 수행할 수 있다. 1차 변환은 primary transform, core transform으로 지칭될 수 있다. 실시예로서, 상기 1차 역변환부 (1 140)는 전술한 MTS를 이용하여 1차 2019/190284 1»(:1^1{2019/003743
33 변환을 수행할 수 있다. 또한, 일 예로, 상기 1차 역변환부 (1140)는 현재 블록에 MTS 가 적용되는지 여부를 결정할수 있다.
일 예로, 현재 블록에 MTS가 적용되는 경우 (즉,
Figure imgf000035_0001
1차 역변환부 (1 140)는 현재 블록의 인트라 예측 모드에 기초하여
Figure imgf000035_0002
후보를 구성할 수 있다. 예를 들어, 상기 1^8 후보는 0814 및/또는 DCT4^\ 조합으로 구성되거나, 0817 및/또는 DCTSS] 조합을 포함할 수 있다. 또는 상기 MTS 후보는 상기 도 6의 실시예들, 이후 설명하게 될 도 26 및 도 27의 실시예들 중 적어도 하나를 포함할 수 있다.
그리고, 상기 1차 역변환부 (1140)는 구성된
Figure imgf000035_0003
후보들 중에서 특정
MTS를 지시하는 11 _ \를 이용하여 현재 블록에 적용되는 1차 변환을 결정할 수 있다.
상기 실시예는 각각 이용될 수 있으나, 본 발명은 이에 한정되지 않으며, 위 실시예 및 본 명세서의 다른 실시예와 결합하여 이용될 수 있다. 도 12는 본 발명이 적용되는 실시예로서, 변환 관련 파라미터에 기초하여 역변환을 수행하는 블록도를 나타낸다.
본 발명이 적용되는 디코더 (200)는, 시퀀스 파라미터를 획득하는 요소 (1210), 다변환 선택 플래그 ( 汗 크幻를 획득하는 요소 (1220), 다변환 선택 인덱스 (^ᄄ 出0^}0룰 획득하는 요소 (1230) 및 변환 커널을 유도하는 요소 (1240)를 포함할 수 있다.
시퀀스 파라미터를 획득하는 요소 (1210)는 sps_mts_intra_enabled_flag 또는 sps_mts_intcr_enabled_flag 를 획득할 수 있다. 여기서
Figure imgf000035_0004
tu_mts_flag 가 인트라 코딩 유닛의 레지듀얼 코딩 신택스에 존재하는지 여부를 나타내고, sps_mts_inter_enabled_flag 는 tu_mts_flag 가 인터 코딩 유닛의 레지듀얼 코딩 신택스에 존재하는지 여부를 나타낸다. 구체적인 예는 상기 도 12의 설명이 적용될 수 있다.
다변환 선택 플래그 (MTS flag)를 획득하는 요소 (1220)는 sps_mts_intra_enabled_flag 또는 sps_mts_inter_enabled_flag에 기초하여 tu_mts_flag 를 획득할 수 있다. 예를 들어, sps_mts_intra_enabled_flag = 1 또는 sps_mts_inter_enabled_flag = 1 일때, 상가다변환 선택 플래그 (MTS flag)를 획득하는 요소 (1220)는 tu_mts_flag를 획득할 수 있다. 여기서, tu_mts_flag 는 다변환 선택이 루마 변환 블록의 레지듀얼 샘플에 적용되는지 여부를 나타낸다. 구체적인 예는 상기 도 12의 설명이 적용될 수 있다.
다변환 선택 인덱스 (MTS index)를 획득하는 요소 (1230)는 tu_mts_flag 에 기초하여 mts_idx를 획득할 수 있다. 예를 들어, tu_mts_flag = 1 일때, 상기 다변환 선택 인덱스 (MTS index)를 획득하는 요소 (1230)는 mts_idx 를 획득할 수 있다. 여기서, mts_idx 는 어떤 변환 커널이 현재 변환 블록의 수평 및/또는 수직 방향에 따른 루마 레지듀얼 샘플들에 적용되는지를 나타낸다. 예를 들어, 상기 도 6의 실시예들, 이후 설명하게 될 도 26 및/또는 도 27의 실시예들 중 적어도 하나가 적용될 수 있다.
변환 커널을 유도하는 요소 (1240)는 mtsjdx 에 대응되는 변환 커널을 유도할 수 있다.
그리고, 상기 디코더 (200)는 상기 변환 커널에 기초하여 역변환을 수행할 수 있다. 2019/190284 1»(:1/10公019/003743
35 상기 실시예는 각각 이용될 수 있으나, 본 발명은 이에 한정되지 않으며, 위 실시예 및 본 명세서의 다른 실시예와 결합하여 이용될 수 있다. 도 13은 본 발명이 적용되는 실시예로서, 변환 관련 파라미터에 기초하여 역변환을 수행하는 흐름도를 나타낸다.
본 발명이 적용되는 디코더는, sps__mts_intra_enabled_flag 또는 sps_mts_inter_enabled_flag 를 획득할 수 있다 (S 1310). 여기서, sps_mts_intra_enabled_flag 는 tu_mts_flag 가 인트라 코딩 유닛의 레지듀얼 코딩 신택스에 존재하는지 여부를 나타낸다. 예를 들어, sps__mts__intra_enabled_flag = 0 이면, tu_mts__flag 가 인트라 코딩 유닛의 레지듀얼 코딩 신택스에 존재하지 않고, sps_mts_intra_enabled_flag = 0 이면, tu_mts_flag 가 인트라 코딩 유닛의 레지듀얼 코딩 신택스에 존재한다. 그리고, sps_mts_i nter_enab 1 ed_fl ag 는 tu__mts_flag 가 인터 코딩 유닛의 레지듀얼 코딩 신택스에 존재하는지 여부를 나타낸다. 예를 들어, sps_mts_inter_enabled_flag = 0 이면, tu_mts_flag 가 인터 코딩 유닛의 레지듀얼 코딩 신택스에 존재하지 않고, sps_mts_inter_enabled_flag = 0 이면, tu_mts_flag 가 인터 코딩 유닛의 레지듀얼 코딩 신택스에 존재한다.
상기 디코더는 sps_mts_intra_enabled_flag 또는 sps_mts_inter_enabled_flag에 기초하여 tu_mts_flag 를 획득할 수 있다 (S1320). 예를 들어, sps__mts_intra_enabled_flag = 1 또는 sps_mts_inter_enabled_flag = 1 일때, 상기 디코더는 tu__mts_flag 를 획득할 수 있다. 여기서, tu_mts_flag 는 다변환 선택 (multiple transform selection, 이하 ‘MTS’라 함)이 루마 변환 블록의 레지듀얼 샘플에 적용되는지 여부를 나타낸다. 예를 들어, tu_mts_flag = 0 이면 MTS가 루마 \¥0 2019/190284 1(1710{2019/003743
36 변환 블록의 레지듀얼 샘플에 적용되지
Figure imgf000038_0001
1 이면 MTS가 루마 변환 블록의 레지듀얼 샘플에 적용된다.
다른 예로, 상기 tix_mtsJlag 에 대해 본 문서의 실시예들 중 적어도 하나가 적용될 수 있다.
상기 디코더는
Figure imgf000038_0002
기초하여 11伯_1(뇨 를 획득할 수 있다 1330). 예를 들어, _11伯_ 3§ = 1 일때, 상기 디코더는
Figure imgf000038_0003
획득할 수 있다. 여기서, 0 는 어떤 변환 커널이 현재 변환 블록의 수평 및/또는 수직 방향에 따른 루마 레지듀얼 샘플들에 적용되는지를 나타낸다.
예를 들어, ᄂ 에 대해, 본 문서 실시예들 중 적어도 하나가 적용될 수 있다. 구체적 예로, 상기 도 6의 실시예들, 이후 설명하게 될 도 26 및/또는 도 27의 실시예들 중 적어도 하나가 적용될 수 있다.
상기
Figure imgf000038_0004
대응되는 변환 커널을 유도할 수 있다 1340). 예를 들어, 상기 11 _1(뇨에 대응되는 변환 커널은 수평 변환 및 수직 변환으로 구분되어 정의될 수 있다.
다른 예로, 상기 수평 변환 및 수직 변환은 서로 다른 변환 커널이 적용될 수 있다. 다만, 본 발명은 이에 한정되지 않으며, 상기 수평 변환 및 수직 변환은 동일한 변환 커널이 적용될 수도 있다.
그리고, 상기 디코더는 상기 변환 커널에 기초하여 역변환을 수행할 수 있다 1350).
상기 실시예는 각각 이용될 수 있으나, 본 발명은 이에 한정되지 않으며, 위 실시예 및 본 명세서의 다른 실시예와 결합하여 이용될 수 있다. 2019/190284 1»(:1^1{2019/003743
37 도 14는 본 발명이 적용되는 실시예로서, DST4(Discrete Sine Transform-4) 및 DCT4(Discrete Cosine Transform-4)를 순방향 DCT2 (forward DCT2) 또는 역방향 DCT2 (inverse DCT2)로 수행하는 인코딩 흐름도를 나타낸다.
인코더는 현재 블록의 예측 모드, 블록 모양 및/또는 블록 크기 중 적어도 하나에 기초하여 수평 변환 및/또는 수직 변환을 결정 (또는 선택)할 수 있다 (S1410). 이때, 수평 변환 및/또는 수직 변환의 후보는 상기 도 6의 실시예들, 이후 설명하게 될 도 26 및/또는 도 27의 실시예들 중 적어도 하나를 포함할 수 있다.
상기 인코더는 RD 최적화 (Rate Distortion optirr^zation)를 통해 최적의 수평 변환 및/또는 최적의 수직 변환을 결정할 수 있다. 상기 최적의 수평 변환 및/또는 상기 최적의 수직 변환은 복수개의 변환 조합 중 하나에 대응될 수 있고, 상기 복수개의 변환 조합은 변환 인덱스에 의해 정의될 수 있다.
상기 인코더는 상기 최적의 수평 변환 및/또는 상기 최적의 수직 변환에 대응되는 변환 인덱스를 시그널링할 수 있다 (S1420). 여기서, 상기 변환 인덱스는 본 명세서에서 설명한 다른 실시예들이 적용될 수 있다. 예를 들어, 상기 도 6의 실시예들, 이후 설명하게 될 도 26 및 도 27의 실시예들 중 적어도 하나를 포함할 수 있다.
다른 예로, 상기 최적의 수평 변환에 대한 수평 변환 인덱스와 상기 최적의 수직 변환에 대한 수직 변환 인덱스를 독립적으로 시그널링할 수도 있다. 상기 인코더는, 상기 최적의 수평 변환을 이용하여 상기 현재 블록에 대해 수평 방향으로 순방향 변환을 수행할 수 있다 (S1430). 여기서, 상기 현재 블록은 변환 블록을 의미할 수 있다. 2019/190284 1»(:1^1{2019/003743
38 그리고, 상기 인코더는, 상기 최적의 수직 변환을 이용하여 상기 현재 블록에 대해 수직 방향으로 순방향 변환을 수행할 수 있다 (S1440). 본 실시예에서는, 수평 변환을 수행한 후 수직 변환을 수행하지만, 본 발명은 이에 한정되지 않는다. 즉, 수직 변환을 먼저 수행한 후 수평 변환을 수행할 수 있다. 일실시예로, 상기 S1430 단계의 수평 방향 순방향 변환에서 순방향
DST4가 적용되고, 상기 S1440 단계의 수직 방향 순방향 변환에서 순방향
DCT4가 적용될 수 있다. 또는 반대의 경우도 가능하다.
일실시예로, 상기 수평 변환과 상기 수직 변환의 조합는 상기 도 6의 실시예들, 이후 설명하게 될 도 26 및 도 27의 실시예들 중 적어도 하나를 포함할 수 있다.
한편, 상기 인코더는, 상기 현재 블록에 대해 양자화를 수행함으로써, 변환 계수 블록을 생성할 수 있다 (S 1450).
상기 인코더는, 상기 변환 계수 블록에 대해 엔트로피 인코딩을 수행하여, 비트스트림을 생성할수 있다. 도 15는 본 발명이 적용되는 실시예로서 ¾ DST4(Discrete Sine Transform-4) 및 DCT4(Discrete Cosine Transfomi-4)를 순방향 DCT2 (forward DCT2) 또는 역방향 DCT2 (inverse DCT2)로 수행하는 디코딩 흐름도를 나타낸다.
디코더는 비트스트림으로부터 변환 인덱스를 획득할 수 있다 (S1510). 여기서, 상기 변환 인덱스는 본 명세서에서 설명한 다른 실시예들이 적용될 수 있다. 예를 들어, 상기 도 6의 실시예들, 이후 설명하게 될 도 26 및 도 27의 실시예들 중 적어도 하나를 포함할 수 있다. 2019/190284 1»(:1^1{2019/003743
39 상기 디코더는 상기 변환 인덱스에 대응되는 수평 변환 및 수직 변환을 유도할 수 있다 1520). 이때, 상기 수평 변환 및/또는 상기 수직 변환의 후보는 상기 도 6의 실시예들, 이후 설명하게 될 도 26 및/또는 도 27의 실시예들 중 적어도 하나를 포함할 수 있다.
다만, 상기 81510 및 81520 단계는 일실시예이며, 본 발명은 이에 한정되지 않는다. 예를 들어, 상기 디코더는 현재 블록의 예측 모드, 블록 모양 및/또는 블록 크기 중 적어도 하나에 기초하여 수평 변환 및 수직 변환을 유도할 수 있다. 다른 예로, 상기 변환 인덱스는 수평 변환에 대응되는 수평 변환 인덱스 및 수직 변환에 대응되는 수직 변환 인덱스를 포함할 수 있다.
한편, 상기 디코더는 상기 비트스트림을 엔트로피 디코딩하여 변환 계수 블록을 획득하고, 상기 변환 계수 블록에 대해 역양자화를 수행할 수 있다 (81530). 상기 디코더는 상기 역양자화된 변환 계수 블록에 대해 상기 수직 변환을 이용하여 수직 방향으로 역방향 변환을 수행할 수 있다 1540).
그리고, 상기 디코더는 상기 수평 변환을 이용하여 수평 방향으로 역방향 변환을 수행할 수 있다 (81550).
본 실시예에서는, 수직 변환을 적용한 후 수평 변환을 적용하지만, 본 발명은 이에 한정되지 않는다. 즉, 수평 변환을 먼저 적용한 후 수직 변환을 적용할 수 있다.
일실시예로, 상기 81540 단계의 수직 방향 역방향 변환에서 역방향 14가 적용되고, 상기 81440 단계의 수평 방향 역방향 변환에서 역방향 å)(그4가 적용될 수 있다. 또는 반대의 경우도 가능하다.
일실시예로, 상기 수평 변환과 상기 수직 변환의 조합는 상기 도 6의 2019/190284 1»(:1^1{2019/003743
40 실시예들, 이후 설명하게 될 도 26 및 도 27의 실시예들 중 적어도 하나를 포함할수 있다.
상기 디코더는 S1550 단계를 통해 레지듀얼 블록을 생성하고, 상기 레지듀얼 블록과 예측블록이 더해져서 복원 블록이 생성된다. 도 16은 본 발명이 적용되는 실시예로서, DST4 및 DCT4를 순방향 DCT2 로 수행시 변환 블록 사이즈 (N) 및 우측 쉬프트 양 (S0의 쌍에 대한 대각 요소들을 나타낸다.
본 발명은 비디오 압축을위한 변환 커널 중 DST4(Discrete Sine Transform- 4) 및 DCT4(Discrete Cosine Transform-4)에 대한 메모리 사용 및 연산 복잡도를 감소시키는 방법을 제안하고자 한다.
일실시예로, DST4(Discrete Sine Transform-4) 및 DCT4(Discrete Cosine Transform-4)를 순방향 DCT2 (forward DCT2)로수행하는 방법을 제공한다.
일실시예로, DST4 및 DCT4를 역방향 DCT2 (inverse DCT2)로 수행하는 방법을 제공한다.
일실시예로, DST4 및 DCT4를 MTS (Multiple Transform Selection)가 적용되는 변환설정 그룹 (transform configuration group)에 적용하는 방법을 제공한다. 실시예 1 : Design of DST4 and DCT4 with DCT2
DST4 및 DCT4의 행렬을유도하기 위한수학식은 다음과 같다.
【수학식 4】 2019/190284 1»(:1/10公019/003743
41
Figure imgf000043_0001
여기서, n (0, ... N-1) 은 행 인덱스 (row index) 를 나타내고, k (0, ... N-1) 는 열 인덱스 (column index) 를,나타낸다. 이 경우, 상기 수학식 4 및 5는 각각 DST4 및 DCT4 의 역방향 변환 행렬을 생성한다. 그리고, 이들의 트랜스포즈 (transpose)는 순방향 변환 행렬을 나타낸다.
DST4 (DCT4) 역방향 변환 행렬을 (成1 ( (C^ ) ) 로 나타내면, 다음 수학식들 6 및 7의 관계를 확인할 수 있다.
【수학식 6]
Figure imgf000043_0002
상기 수학식 6 및 7에 따라, 본 발명은, 선처리 단계 (pre-processing stage) 또는 후처리 단계 (post-processing stage)를 통해 입력 순서 또는 줄력 순서 (input or output order)를 바꾸고 부호 (sign)를 변경함으로써, DCT4 (DST4) 역방향 변환 행렬 \ 02019/190284 1»(:1/10公019/003743
42
(S^) (((才)) 로부터, 각각 (0014) 역방향 변환 행렬 (<) ((( )) 을 유도할 수 있다.
따라서, 본 발명을 통해 0 4 또는 00141- 수행할 경우 부가적인 계산 없이 하나로부터 다른 하나를 쉽게 유도할 수 있다.
본 발명의 일실시예에서, £>(刀4는 £)(刀2를 이용하여 다음과 같이 표현될 수 있다.
【수학식 8】
(Cn !vJ = (C;I )=AN(C! !:JM
Figure imgf000044_0001
여기서, 는 후처리 행렬을 나타내고, 山은 선처리 행렬을 나타낸다.
예로서,
Figure imgf000044_0002
본 발명은, 상기 수학식 8로부터 1X^4가 후처리 행렬 71八, 선처리 행렬 및 1X72 에 기초하여 디자인될 수 있다는 것을 확인할 수 있다. 여기서, 2019/190284 1»(:1/10公019/003743
43 후처리 행렬 및 선처리 행렬 의 경우 단지 소량의 곱셈만이 추가된다. 그리고, DCT2 는, 저장되어야 할 계수들의 수를 줄일 수 있고, DCT2 행렬 내 계수들간의 대청성 (symmetry)에 기반한 고속 실행 (fast implementation)을 위한 변환으로 잘 알려져 있다.
따라서, 약간의 곱셉 팩터를 추가함으로써, 저복잡도로 £)(刀4 의 고속 실행을 실현할 수 있다. 이는 0요14의 경우에도 마찬가지이다.
후처리 행렬 ᆻ 및 선처리 행렬 잤의 역행렬은 다음 수학식 9와 같이 표현될 수 있다.
【수학식 9】
Figure imgf000045_0001
상기 수학식 9의 A -', 를 이용하여, 본 발명은 1) 4 및 0012 2019/190284 1»(:1/10公019/003743
44 사이의 다음 수학식 10과 같은 또다른 관계식을 유도할 수 있다.
【수학식 10]
[cfh [C = M- cfjAN-'
여기서, 상기 A-] , M~] 은 ( ) 보다 더 심플한 곱셉들로 구성되어 있어서, 저복잡도로 DCT4의 고속 실행(fast implementation)이 가능하다. 그리고, A-' 은 과 v 보다 더 적은 수의 덧셈 또는 뺄셈을 초래하지만, MJ- 내 계수들은 MN 보다 더 넓은 레인지를 갖는다. 따라서, 본 발명은, 복잡도와 퍼포먼스 간의 트레이드 오프를 고려하여, 상기 수학식 9 및 10 에 기초하여 변환 타입을 디자인할 수 있다.
상기 수학식 7, 8, 10으로부터, 본 발명은 DCT2의 고속 실행(fast implementation)을 재사용함으로써 저복잡도의 DST4를 실행할 수 있다. 이는 다음 수학식 11 및 12에서 보여진다.
【수학식 1 1】
(에) =(섟,)=(幻세 } · )
Figure imgf000046_0001
2019/190284 1»(:1/10公019/003743
45
Figure imgf000047_0005
실시예 2 : 순방향 DCT2에 의한 DST4 및 DCT4의 실행 (Implementation of DST4 and DCT4 with forward DCT2)
DST4 의 실행을 위해 상기 수학식 11 이 이용되는 경우, 먼저 길이 N의 입력 벡터는 、MNJn) 만큼 스케일되어야 한다. 마찬가지로, DCT4의 실행을 위해 상기 수학식 8이 이용되는 경우, 먼저 길이 N의 입력 벡터는 ( Mn ) 만큼 스케일되어야 한다.
M N 내 대각 요소들 (diagonal elements)은 부동 소수점 수들 (floating point numbers)이고, 이는 고정점 (fixed-point) 또는 정수 곱셈 (integer multiplications)에서 이용되기 위해 적절히 스케일되어야 한다. 정수화된 (integerized) {MNJn) 및
Figure imgf000047_0001
(MNjJ 및
Figure imgf000047_0003
로 표시한다면
Figure imgf000047_0002
및 ik/;는 각각 다음 수학식 13에 따라 계산될 수 있다.
【수학식 13】
Figure imgf000047_0004
2019/190284 1»(:1/10公019/003743
46
Figure imgf000048_0001
상기 도 16은 # 및 此 에 기초한 MN' 예들을 나타낸다. 여기서, diag(·)는 argument matrix 를 argument matrix 내 대각 요소들을 구성하는 연관 벡터로 변환하는 것을 의미한다. 동일한 (ᄊ 幻)의 산 ((% ·과) 는 각 벡터의 요소 순서를 변경함으로써 상기 도 16으로부터 쉽게 유도될 수 있다. 예를 들어, [251 ,213,142,50] 는 [50,142,213,251] 로 변경할 수 있다.
본 발명은 각 에 대해 5/을 다르게 설정할 수도 있다. 예를 들어, 4x4 변환에 대해 을 7로 설정하고, 8x8 변환에 대해 5;을 8로 설정할 수 있다. 상기 수학식 13의
Figure imgf000048_0002
은 2"1 만큼 스케일링하기 위한 좌측 쉬프트 양을 나타내고,‘ 이피선” 연산자는 적절한 1>01111(111 을 수행한다. 八 및 ( ) 는 대각 행렬들이고, 입력 벡터 X 의 1* 요소(< 11( (3 公 ) 는 [好 ,, 만큼 곱해진다. 입력 벡터 X 의 곱셈 결과 및 대각
Figure imgf000048_0003
행렬들은 다음 수학식 14와 같이 나타낼 수 있다.
【수학식 14
Figure imgf000048_0004
상기 수학식 14의 도는 곱셈 결과를 나타낸다. 다만, X 는 이후에 스케일 다운되어야 한다. ^ 의 다운 스케일링은 1X72를 적용하기 전에 수행되거나, 2019/190284 1»(:1/10公019/003743
47
0(그2를 적용한 후에 수행되거나, 山 ( (公; 0 ) 에 1X74 肝4) 를 곱한 후에 수행될 수 있다. 만약, 도의 다운 스케일링은 DCT2를 적용하기 전에 수행되는 경우, 다운 스케일된 것 (the down-scaled one), x 는 다음 수학식 15에 기초하여 결정될 수 있다.
【수학식 15】
Figure imgf000049_0001
상기 수학식 15에서, ¾ 는 상기 과 같은 값일 수 있다. 다만, 본 발명은 이에 한정되지 않으며, 상기 ¾ 는 상기 과 다른 값을 가질 수도 있다.
상기 수학식 15에서, 스케일링 및 라운딩의 어떠한 타입들도 이용가능하며, 일실시예로 상기 수학식 15의 (1), (2)가 이용될 수 있다. 즉, 상기 수학식 15에 보여진 것처럼, (1), (2) 또는 다른 함수들이 국 를 찾기 위해 적용될 수 있다. 도 17 및 도 18은 본 발명이 적용되는 실시예들로서, 도 17은 DST4 또는 DCT4에 적용될 수 있는 DCT2 커널 계수들의 세트들을 나타내고, 도 18은 DCT2 커널 계수의 세트로부터 생성된 순방향 DCT2 행렬을 나타낸다.
본 발명의 일실시예는 HEVC와 동일한 DCT2 커널 계수 (kernel coefficient)를 이용할 수 있다. 32x32 까지의 모든 사이즈의 모든 DCT2 커널 계수들 간의 대칭성에 의해 이용되는 (which is facilitated by symmetries among all DCT2 kernel coefficients of all sizes up to 32x32), DCT2의 31개의 다른 계수들을 유지할 필요가 있다.
만약 기존 DCT2 실행 (implementation)을 재사용한다면, DST4 또는 DCT4 2019/190284 1»(:1^1{2019/003743
48 에서 이용되는 DCT2의 부가적인 계수들을 저장할 필요가 없다.
만약 기존 DCT2(existing DCT2)가 아닌 특정 DCT2 커널 (specific DCT2 kernel)을 이용한다면, 본 발명은 동종의 대칭성을 이용하는 31개 계수들인 DCT2 커널 계수들의 단 하나의 세트를 추가할 수 있다. 즉, 2n x 2n DCT2 까지 지원된다면,본 발명은 (2n - l)개의 다른 계수들만이 필요하다.
이러한부가적인 세트는 기존보다 더 높거나낮은 정확도를 가질 수 있다. 만약 x 의 동적 범위 (dynamic range)가 기존 DCT2 디자인에 의해 지원되는 범위를 초과하지 않는다면, 본 발명은 내부 변수들 (internal variables)의 비트 길이를 확장하지 않고 DCT2와 동일한루틴 (routine)을 재사용할 수 있고, DCT2의 레거시 디자인 (legacy design)을 재사용할수 있다.
DCT2보다 DST4/DCT4 의 더 산술적인 정확도를 필요로 한다고 할지라도, 더 높은 정확도를 축적할 수 있는 업데이트된 루틴 (routine) 또한 기존 DCT2를 수행하기에 충분하다. 예를 들어, DCT2 계수들의 더 정확한 세트들이 스케일링 팩터들에 따라상기 도 17에 리스트되어 있다.
상기 도 17의 각 계수는 기저 벡터들 (basis vectors) 간의 직교성을 향상시키기 위해 더 조정될 수 있고, 각 기저 벡터의 norm은 1에 가깝게 하고, floating-point accurate DCT2 kernel로부터 Frobenius norm error를 줄일 수 있다. 만약, 계수 세트가 (a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t,u八 w,x,y,z,A,B,C,D,E)로 주어지면, 상기 계수 세트로부터 생성된 순방향 DCT2는 도 18과 같이 설정될 수 있다.
상기 도 18에서, 각 DCT2 계수 세트 (상기 도 18의 각 행)는 (a,b?c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t,u,v,w,x,y,z,A,B,C,D,E) 형태로 기술된다. 이는, 31개의 2019/190284 1»(:1^1{2019/003743
49 다른 계수들만 (only 31 possibly different coefficients)이 32x32 보다 크지 않은 사이즈의 모든 DCT2 변환들을 위해 필요하다는 것을 반영한다.
DCT2 변환의 출력은 DCT4 (또는 DST4) 의 행렬 AN (또는 DNAN)을 통해 후처리되어야 한다. 입력 벡터를 상기 DCT4 (또는 DST4) 의 행렬 AN (또는 끄 비에 제공하기 전에, 상기 입력 벡터로서 DCT2 출력 벡터는 한정된 비트 길이의 변수들로 저장하기 위해 정확도 조정을 위한 값으로 라운드될 수 있다. 스케일링 및 라운딩 전의 상기 DCT2 출력 벡터를 y 라 하면, 라운드된 값 (rounded one) y는 다음 수학식 16으로부터 결정될 수 있다. 상기 수학식
15처럼,스케일링 및 라운딩의 다른 형태들도상기 수학식 16에 적용가능하다.
【수학식 16】
Figure imgf000051_0001
상기 수학식 16에서, S3 가 0 인 경우, 어떠한스케일링 및 라운딩도 少,에 적용되지 않는다. 즉 X = yf 이다.
AN 또는 (£)^ 를 에 곱한 후의 최종 출력 벡터를 X라 하자. 곱셈의 대부분은 처음 1/V2 곱셈을 제외하고는, 간단한 덧셈 또는 텔셈에 의해 대체될 수 있다. 여기서, 1/V2 팩터는상수 (constant number)이므로 다음수학식 17에 의해 보여지는 것처럼 우즉 쉬프트에 의한 하드웨어 곱셈 (hardwired multiplication)만큼 근사화될 수 있다. 상기 수학식 15처럼, 스케일링 및 라운딩의 다른 형태들도 상기 수학식 17에 적용가능하다.
【수학식 17】
x0-(y0-F+ 0« (¾ - 1))) » ¾ 2019/190284 1»(:1/10公019/003743
50 상기 수학식 17에서, 및 公 는 F»S4 가 1/V2 에 매우 근접하다는 조건을 만족시켜야 한다. (F, S4) 쌍을 획득하는 방법 중 하나는 F = roundel V2 ) « SA } 를 이용하는 것이다.
본 발명은, 1/V2 에 더 정확한 근사화를 위해, 公를 증가시칼 수 있지만, S4 의 증가는 더 긴 길이의 중간 변수들 (intermediate variables of longer length)을 필요로 하며, 이는 실행 복잡도를 증가시킬 수 있다. 아래 표 1는 1/V2 에 근사화하는 (F, S4) 의 가능한 쌍들을 나타낸다.
【표 1】
Figure imgf000052_0002
상기 수학식 17에서, 본 발명은 전체적인 스케일링을 변화하지 않기 위해, 의 좌측 쉬프트와 동일한 양의 우측 쉬프트 (公) 를 적용한다는 것을 가정하지만, 반드시 그럴 필요는 없다. 만약, S4 대신에 ¾ (< S4) 만큼 우측 쉬프트를 적용할 경우, 본 발명은
Figure imgf000052_0001
모든 j)를 스케일 업해야 한다. DCT4 (또는 DST4) 계산 (5V· , 여기서 positive value는 right shift를 의미함) 후의 예즉된 결과 스케일링 (expected resultant scaling)과 이전 수학식들의 모든 쉬프트들을 고려할 때, 본 발명은 모든 스케일링 비트 쉬프트 값들 (all the scaling bit shift values)을 가진 다음 수학식 18을 설정할 수 있다. 2019/190284 1»(:1^1{2019/003743
51
【수학식 18】
Sr = (S^ - S2) + Sc - S3 + (S4 - S5) - S0
상기 수학식 18에서, Sc는 DCT2 정수 곱셈으로 인한 좌측 쉬프트 양 (left shift amount)을 나타내고, 이는 상기 도 17에서처럼 비정수 값 (non-integer value)일 수 있다. So 는 DCT4 (또는 DST4) 의 최종 출력 ( )을 산출하는 우측 쉬프트 양 (right shift amount)을 나타낸다.상기 수학식 18에서 몇몇 부분들은 0일 수 있다. 예을들어, (5/ - &), ¾또는 (Ss - S4) 는 0일 수 있다. 도 19 및 도 20은 본 발명이 적용되는 실시예들로서, 도 19는 DST4 에 대한 출력 단계의 코드 실행을 나타내고, 도 20은 DCT4 에 대한 출력 단계의 코드실행을 나타낸다.
최종 출력 벡터의 i번째 요소를 라 하면, 본 발명의 일실시예는 상기 도 19에서 같이 (公 의 곱셈에 대응되는 DST4에 대한 최종 단계의 코드 실행 예를 제공할수 있다.
또한, 본 발명의 다른 일실시예는 상기 도 20에서와 같이 의 곱셈에 대응되는 DCT4에 대한최종단계의 코드실행 예를 제공할수 있다.
상기 도 19에서의 cutoff 는 벡터 X 에서 계수들의 유효 개수를 나타낸다. 예를 들어,상기 cutoff 는 N 일 수 있다.
상기 도 19에서, S 1910 단계 및 S1920 단계는 다음 수학식 19와 같이 하나의 계산과정으로 머지될 수 있다.
【수학식 19】
X。 =〔7江» 3(£; 次«/»«", £ /½3)«>«·, {% F + (1 « (>55 + 5 - 1)) » (5 + ¾)) 2019/190284 1»(:1^1{2019/003743
52 상기 수학식 15처럼, 스케일링 및 라운딩의 다른 형태들도 상기 도 19 및 상기 수학식 19에 적용가능하다.
상기 도 20에서, S2010 단계 및 S2020 단계는 다음 수학식 20과 같이 하나의 계산 과정으로 머지될 수 있다.
【수학식 20]
X0 -- Clip3(、clipMinimum, clipMaximwn, (y0 - F + (1 « (S5 + S0 - 1)) » (S5 + SQ)) 상기 수학식 15처럼, 스케일링 및 라운딩의 다른 형태들도 상기 도 20 및 상기 수학식 20에 적용가능하다.
상기 도 19 및 도 20에서, Clip3 은 argument value 를 양끝 (clipMinimum, clipMaximum)에 클립하는 연산을 나타낸다.
AN (또는 {DNAn) )^\ 각 행 (each row)은 이전 행의 통상적인 패턴 (common patern with its previous row)을 가지며, 본 발명은 적절한 부호 전환 (proper sign reversal)에 따라 이전 행의 결과 (a result of the previous row)를 재이용할 수 있다. 이러한 패턴은 상기 도 19 및 도 20에서 변수 z, prev 를 통해 활용될 수 있다. 여기서, 변수 z, prev 는 AN (또는 (公 ᆻ))의 곱셈 계산을 감소시킨다.
상기 변수 z, prev 에 의해, 본 발명은 각 출력마다 한번의 곱셈 또는 한번의 덧셈/뺄셈만을 필요로 한다. 예를 들어, 곱셈은 최초 출력 요소에서만 필요할 수 있다. 도 21은 본 발명이 적용되는 실시예로서, DST4 및 DCT4를 순방향 DCT2 로 수행시 DST4 및 DCT4 에 대한 파라미터 세트 및 곱셈 계수들의 구성을 나타낸다. 2019/190284 1»(:1^1{2019/003743
53 상기 도 21은 DST 4 및 DCT4의 파라미터 세트 및 곱셈 계수들 (parameter set and multiplication coefficients for DST4 and DCT4)의 구성 (configuration)을 나타낸다. 다른 사이즈의 각 변환은 개별적으로 구성될 수 있다. 즉, 다른 사이즈의 각 변환은 각자의 파라미터 세트 및 곱셈 계수들을 가질 수 있다.
예를 들어, DST4의 파라미터 세트의 구성이 (Si, S2, Ss, S4, S5, So) 일 때, 모든블록크기에 대해 곱셈 계수값은 (8, 8, 0, 8, 8, identical to HEVC) 일 수 있다. 그리고, DCT4의 파라미터 세트의 구성이 (Si, S2, S3, S4, Ss, So) 일 때, 모든 블록 크기에 대해 곱셈 계수값은 (8, 8, 0, 8, 8, identical to HEVC) 일 수 있다.
또한, 파라미터 세트의 구성이 MN, 일 때 각 블록 사이즈마다 상기 도 21에 기재된 각각의 곱셈 계수값을 가질 수 있다.
본 발명에 따르면,상기 수학식 18에 의해, 역방향 DST4 [DCT4]의 실행은 순방향 DST4 [DCT4] 와동일하다. 도 22 및 도 23은 본 발명이 적용되는 실시예들로서, 도 22는 DCT4 에 대한 전처리 단계의 코드 실행을 나타내고, 도 23은 DST4 에 대한 전처리 단계의 코드 실행을 나타낸다.
실시예 3: Alternative implementation of DST4 and DCT4 with inverse DCT2 본 발명은, 상기 수학식 10 및 12를 통해 각각 DCT4 및 DST4 를 실행하는 방법을 제공한다.
섟, (AN-'Jn) , M-\ 및 (DNMJ- ) 는 An , {DNAn) , Mn , 및 {MNJn、 대신 이용될 수 있고, 이들 각각은 DCT2와 비교해서 더 작은 계산량을 필요로 한다. 역방향 DCT2는상기 수학식 10 및 12에서 순방향 DCT2 대신에 적용된다. 2019/190284 1»(:1/10公019/003743
54 상기 수학식 8 및 11과 대조적으로, A-' 또는 (A~'Jn) 는 입력 벡터 x 에서 적용되고, M-' 또는 (DNMJ- ) 는 DCT2의 출력 벡터에서 적용된다.
상기 수학식 9 및 12처럼, 하나의 요소만이 AJ- 및 (^1^) 에서 V2 만큼 곱해진다. 여기서, A ' 및 (A~]Jn) 는 우측 쉬프트만큼 정수 곱셈으로 근사화될 수 있다.
상기 수학식 10 에서 DCT4 의 전처리 단계의 코드 실행의 예는 도 22와 같으며, 이는 A~' 의 곱셈에 대응된다. 또한, 상기 수학식 12에서 DST4의 전처리 단계의 코드 실행의 예는 도 23과 같으며, 이는 (세시의 곱셈에 대응된다.
상기 수학식 15처럼, 스케일링 및 라운딩의 다른 형태들도 다음 표 8, 9에 적용가능하다.
상기 도 22, 23 에서, N 은 입력 벡터 x의 길이 뿐 아니라 변환 기저 벡터 (transform basis vector)의 길이를 나타낸다. F 및 51/ 은, 관계식
Figure imgf000056_0001
의 V 를 근사화하기 위한 곱셈 팩터 (multiplication factor) 및 우즉 쉬프트 양 (right shift amount)을 나타낸다.
상기 도 22, 23 에서, 2S1_S2만큼 입력 벡터를 스케일 업해야 하므로, S2 는
Si 대신 라운딩을 위해 이용된다. Si 이 Si 과 같으면, 입력 벡터에 스케일링을 할 필요가 없다. 다음 표 2는 、反 곱셈을 근사화하기 위한 (F, S0 쌍의 예를 나타낸다.
【표 2】
Figure imgf000056_0002
2019/190284 1»(:1/10公019/003743
55
Figure imgf000057_0002
상기 수학식 16에서와 같이, 본 발명은 더 짧은 비트 길이의 변수를 이용하기 위해 역방향 1X^2 출력을 스케일 다운할 수 있다. 역방향 1X^2 출력 벡터를 ^ 라 하고 ᅵ번째 요소를 ᄊ라 하면, 다음 수학식 21 에 따라 스케일된 출력 벡터 를 획득할 수 있다. 상기 수학식 15처럼, 스케일링 및 라운딩의 다른 형태들도 상기 수학식 21에 적용가능하다.
【수학식 21 ]
女 = (刀 + (1 « (¾ - 1))) ñ ñ公3I = 0,1, ..., ^ - 1
상기 수학식 10 및 12에서, 후처리 단계들은 각각 및 (、å> ) 에 대응된다. 여기서, 연관된 대각 계수들은 고정점 또는 정수 곱셈을 위해 스케일 업될 수 있다. 이러한 스케일 업은 다음 수학식 22와 같이 적절한 좌측 쉬프트들로 수행될 수 있다.
【수학식 22】
Figure imgf000057_0001
도 24는 본 발명이 적용되는 실시예로서 , 0814 및 1X74를 역방향 0012 2019/190284 1»(:1^1{2019/003743
56 로 수행시 변환 블록 사이즈어) 및 우측 쉬프트 양 4)의 쌍에 대한 대각 요소들을 나타낸다. 의 대각 요소들의 예들은 상기 도 24의 및 幻의 다양한 조합으로 보여질 수 있다.
상기 실시예 2에서처럼, S4 는 각 변환 사이즈에 대해 다르게 설정될 수 있다. 상기 도 24에서, (N, S4) 이 (32, 9) 일 경우, ‘10431, 과 같은 큰 수들은, 더 짧은 비트 길이의 연산자부의 곱셈에 적합한, 아래 수학식 23과 같이 분해될 수 있다. 이는,큰 수의 곱셈이 나타나는 경우에 적용될 수 있다.
【수학식 23]
10431 - x = (8096 + 2048 + 287) · x = (JC « 13) + (x « 1 1) + (287 · x)
Figure imgf000058_0001
대응되는 예들은 상기 도 24로부터 유도될 수 있다. 예를 들어, (N, S4) 이 (4, 9) 인 경우, 벡터는 [261, -308, 461, -1312] 이다.
0이 아닌 요소들 (non-zero elements)은 M 및 {DNM^ ) 에서 대각 라인들상에서만 이용가능하고, 연관된 행렬 곱셈은 다음 수학식 24와 같이, 간단한단위별 곱셈 (element-wise multiplication) 에 의해 수행될 수 있다.
【수학식 24】
Figure imgf000058_0002
최종 출력 벡터를 X 라 하면, 상기 수학식 24로부터 산출된 1는 미리 2019/190284 1»(:1^1{2019/003743
57 주어진 예즉된 스케일링 (expected scaling)을 만족시키기 위해 적절히 스케일되어야 한다. 예를 들어, 상기 최종 출력 벡터 X를 획득하기 위한 좌측 쉬프트 양 (left shift amount)이 So 이고 상기 예측된 스케일링이 ST이면,切 및 切와 함께 쉬프트 길이들 간의 전체적인 관계는 다음 수학식 25와 같이 설정될 수 있다.
【수학식 25]
X, = (i-, + (1 « (¾ - 1))) » ¾, z = 0,1,...,iV - 1
Figure imgf000059_0001
여기서, ST 는 음수 (negative value) 뿐만 아니라 비음수값 (non-negative value)을 가질 수 있다. Sc 는 상기 수학식 18에서와 같은 값을 가질 수 있다. 상기 수학식 15처럼, 스케일링 및 라운딩의 다른 형태들도 상기 수학식 25에 적용가능하다. 도 25는 본 발명이 적용되는 실시예로서, DST4 및 DCT4를 역방향 DCT2 로 수행시 DST4 및 DCT4 에 대한 파라미터 세트 및 곱셈 계수들의 구성을 나타낸다.
상기 도 25는 DST4 및 DCT4 에 대한 다른 실행 (alternative implementation)에서의 파라미터 세트 및 곱셈 계수들 (parameter set and multiplication coefficients)의 구성 (configuration)을 나타낸다. 다른 사이즈의 각 변환은 개별적으로 구성될 수 있다. 즉, 다른 사이즈의 각 변환은 각자의 파라미터 세트 및 곱셈 계수들을 가질 수 있다.
예를 들어, DST4의 파라미터 세트의 구성아 (Si, S2, S3, S4, Ss, So) 일 때, 모든 블록 크기에 대해 곱셈 계수값은 (8, 8, 0, 8, 8, identical to HEVC) 일 수 있다. 2019/190284 1»(:1^1{2019/003743
58 그리고, DCT4의 파라미터 세트의 구성이 (Si, S2, S3, S4, S5, So) 일 때, 모든 블록 크기에 대해 곱셈 계수값은 (8, 8, 0, 8, 8, identical to HEVC) 일 수 있다. 또한, 파라미터 세트의 구성이 MJ~ 일 때 각 블록 사이즈마다 상기 도 25에 기재된 각각의 곱셈 계수값을 가질 수 있다.
본 발명에 따르면,상기 수학식 18에 의해, 역방향 DST4 [DCT4]의 실행은 순방향 DST4 [DCT4] 와동일하다. 도 26 및 도 27은 본 발명이 적용되는실시예들로서,도 26은 인트라 예측 레지듀얼에 대한 MTS 매핑을 나타내고, 도 27은 인터 예측 레지듀얼에 대한
MTS 매핑을 나타낸다. 실시예 4: Possible Multiple Transform Selection (MTS) mapping with DST4 and DCT4 본 발명의 일실시예로, DCT4 및 DST4 는 MTS 매핑을 생성하기 위해 이용될 수 있다. 예를 들어, DST7 및 DCT8 이 DCT4 및 DST4 에 의해 대체될 수 있다.
다른 실시예로, DCT4 및 DST4 만이 MTS 를 생성하기 위해 이용될 수 있다. 예를 들어, 다음 표 13 및 14는 각각 인트라 예측된 레지듀얼 및 인터 예측된 레지듀얼에 대한 MTS 예를 나타낸다.
본 발명의 다른 실시예로, DST4, DCT4, DCT2 등의 다른 조합들에 의해 매핑 또한가능하다.
다른 일실시예로, DCT4를 DCT2로 대체하는 MTS구성이 가능하다. 2019/190284 1»(:1^1{2019/003743
59 다른 일실시예로, 1X78 / 0817 으로 구성된 인터 예측된 레지듀얼에 대한 매핑은 그대로 유지하고, 인트라 예측된 레지듀얼에 대해서만 대체할 수도 있다. 다른 일실시예로, 상기 실시예들의 조합도 가능하다. 도 28은 본 발명이 적용되는 실시예로서, 컨텐츠 스트리밍 시스템 구조도를 나타낸다.
상기 도 28을 살펴보면, 본 발명이 적용되는 컨텐츠 스트리밍 시스템은 크게 인코딩 서버, 스트리밍 서버, 웹 서버, 미디어 저장소, 사용자 장치 및 멀티미디어 입력 장치를 포함할 수 있다.
상기 인코딩 서버는 스마트폰, 카메라, 캠코더 등과 같은 멀티미디어 입력 장치들로부터 입력된 컨텐츠를 디지털 데이터로 압축하여 비트스트림을 생성하고 이를 상기 스트리밍 서버로 전송하는 역할을 한다. 다른 예로, 스마트폰, 카메라, 캠코더 등과 같은 멀티미디어 입력 장치들이 비트스트림을 직접 생성하는 경우, 상기 인코딩 서버는 생략될 수 있다.
상기 비트스트림은 본 발명이 적용되는 인코딩 방법 또는 비트스트림 생성 방법에 의해 생성될 수 있고, 상기 스트리밍 서버는 상기 비트스트림을 전송 또는 수신하는 과정에서 일시적으로 상기 비트스트림을 저장할 수 있다. 상기 스트리밍 서버는 웹 서버를 통한 사용자 요청에 기초하여 멀티미디어 데이터를 사용자 장치에 전송하고, 상기 웹 서버는 사용자에게 어떠한 서비스가 있는지를 알려주는 매개체 역할을 한다. 사용자가 상기 웹 서버에 원하는 서비스를 요청하면, 상기 웹 서버는 이를 스트리밍 서버에 전달하고, 상기 스트리밍 서버는 사용자에게 멀티미디어 데이터를 전송한다. 이때, 2019/190284 1»(:1^1{2019/003743
60 상기 컨텐츠 스트리밍 시스템은 별도의 제어 서버를 포함할 수 있고, 이 경우 상기 제어 서버는 상기 컨텐츠 스트리밍 시스템 내 각 장치 간 명령/응답을 제어하는 역할을 한다.
상기 스트리밍 서버는 미디어 저장소 및/또는 인코딩 서버로부터 컨텐츠를 수신할 수 있다. 예를 들어, 상기 인코딩 서버로부터 컨텐츠를 수신하게 되는 경우, 상기 컨텐츠를 실시간으로 수신할 수 있다. 이 경우, 원활한 스트리밍 서비스를 제공하기 위하여 상기 스트리밍 서버는 상기 비트스트림을 일정 시간동안 저장할 수 있다.
상기 사용자 장치의 예로는, 휴대폰, 스마트 폰 (smart phone), 노트북 컴퓨터 (laptop computer), 디지털방송용 단말기, PDA(personal digital assistants), PMP(portable multimedia player), 네비게이션, 슬레이트 PC(slate PC), 태블릿 PC(tablet PC), 울트라북 (ultrabook), 웨어러블 디바이스 (wearable device, 예를 들어, 워치형 단말기 (smartwatch), 글래스형 단말기 (smart glass), HMD(head mounted display)), 디지털 TV, 데스크탑 컴퓨터, 디지털 사이니지 등이 있을 수 있다.
상기 컨텐츠 스트리밍 시스템 내 각 서버들은 분산 서버로 운영될 수 있으며, 이 경우 각 서버에서 수신하는 데이터는 분산 처리될 수 있다. 상기 기술된 것과 같이, 본 발명에서 설명한 실시예들은 프로세서, 마이크로 프로세서, 컨트롤러 또는 칩 상에서 구현되어 수행될 수 있다. 예를 들어, 각 도면에서 도시한 기능 유닛들은 컴퓨터, 프로세서, 마이크로 프로세서, 컨트롤러 또는 칩 상에서 구현되어 수행될 수 있다.
또한, 본 발명이 적용되는 디코더 및 인코더는 멀티미디어 방송 송수신 2019/190284 1»(:1^1{2019/003743
61 장치, 모바일 통신 단말, 홈 시네마 비디오 장치, 디지털 시네마 비디오 장치, 감시용 카메라, 비디오 대화 장치, 비디오 통신과 같은 실시간 통신 장치, 모바일 스트리밍 장치, 저장 매체, 캠코더, 주문형 비디오 (VoD) 서비스 제공 장치, OTT 비디오 (Over the top video) 장치, 인터넷 스트리밍 서비스 제공 장치, 3차원 (3D) 비디오 장치, 화상 전화 비디오 장치, 및 의료용 비디오 장치 등에 포함될 수 있으며, 비디오 신호 또는 데이터 신호를 처리하기 위해 사용될 수 있다. 예를 들어, OTT 비디오 (Over the top video) 장치로는 게임 콘솔, 블루레이 플레이어, 인터넷 접속 TV, 홈시어터 시스템, 스마트폰, 태블릿 PC, DVR(Digital Video Recoder) 등을 포함할 수 있다.
또한, 본 발명이 적용되는 처리 방법은 컴퓨터로 실행되는 프로그램의 형태로 생산될 수 있으며, 컴퓨터가 판독할 수 있는 기록 매체에 저장될 수 있다. 본 발명에 따른 데이터 구조를 가지는 멀티미디어 데이터도 또한 컴퓨터가 판독할 수 있는 기록 매체에 저장될 수 있다. 상기 컴퓨터가 판독할 수 있는 기록 매체는 컴퓨터로 읽을 수 있는 데이터가 저장되는 모든 종류의 저장 장치 및 분산 저장 장치를 포함한다. 상기 컴퓨터가 판독할 수 있는 기록 매체는, 예를 들어, 블루레이 디스크 (BD), 범용 직렬 버스 (USB), ROM, PROM, EPROM, EEPROM, RAM, CD-ROM, 자기 테이프, 플로피 디스크 및 광학적 데이터 저장 장치를 포함할 수 있다. 또한, 상기 컴퓨터가 판독할 수 있는 기록 매체는 반송파 (예를 들어, 인터넷을 통한 전송)의 형태로 구현된 미디어를 포함한다. 또한, 인코딩 방법으로 생성된 비트스트림이 컴퓨터가 판독할 수 있는 기록 매체에 저장되거나 유무선 통신 네트워크를 통해 전송될 수 있다.
또한, 본 발명의 실시예는 프로그램 코드에 의한 컴퓨터 프로그램 2019/190284 1»(:1^1{2019/003743
62 제품으로 구현될 수 있고, 상기 프로그램 코드는 본 발명의 실시예에 의해 컴퓨터에서 수행될 수 있다. 상기 프로그램 코드는 컴퓨터에 의해 판독가능한 캐리어 상에 저장될 수 있다.
【산업상 이용가능성】
이상, 전술한 본 발명의 바람직한 실시예는, 예시의 목적을 위해'개시된 것으로, 당업자라면 이하 첨부된 특허청구범위에 개시된 본 발명의 기술작 사상과 그 기술적 범위 내에서, 다양한 다른 실시예들을 개량, 변경, 대체 또는 부가등이 가능할 것이다.

Claims

2019/190284 1»(:1^1{2019/003743 63 【청구의 범위】
【청구항 1】
저복잡도 변환 실행에 기초하여 비디오 신호를 복원하는 방법에 있어서, 상기 비디오 신호로부터 현재 블록의 변환 인덱스를 획득하는 단계, 여기서 상기 변환 인덱스는 0814 및/또는 0014 의 조합으로 구성된 복수개의 변환 조합들 중 어느 하나에 대응됨;
상기 변환 인덱스에 대응되는 변환 조합을 유도하는 단계, 여기서 상기 변환 조합은 수평 변환 및 수직 변환으로 구성되고, 상기 수평 변환 및 상기 수직 변환은 상기 0814 또는 상기 0014 중 어느 하나에 대응됨;
상기 14를 이용하여, 상기 현재 블록에 대해 수직 방향으로 역방향 변환을 수행하는 단계;
상기 1)(그14를 이용하여, 상기 현재 블록에 대해 수평 방향으로 역방향 변환을 수행하는 단계; 및
상기 역방향 변환된 현재 블록을 이용하여 상기 비디오 신호를 복원하는 단계
를 포함하는 것을 특징으로 하는 방법.
【청구항 2]
제 1항에 있어서,
상기 0814 및/또는 상기 DCT4³ 순방향 1X72 또는 역방향
Figure imgf000065_0001
이용하여 실행되는 것을 특징으로 하는 방법.
【청구항 3】
제 2항에 있어서, 2019/190284 1»(:1/10公019/003743
64 상기 14 및/또는 상기 1)(그4는 상기 순방향 1) 2 또는 상기 역방향 후처리 행렬 및 선처리 행렬 울 적용하는 것을 특징으로 하는
Figure imgf000066_0002
사이즈를 나타냄)
【청구항 4】
제 1항에 있어서,
상기 수직 변환이 상기 14일 때 상기 14의 역변환을 열마다 적용하고, 상기 수평 변환이 1X714일 때 상기 1X714의 역변환을 행마다 적용하는 것을 특징으로 하는 방법.
【청구항 5】
제 1항에 있어서,
상기 변환 조합 (수평 변환, 수직 변환)은, 4, 14), (1X74, 14),
Figure imgf000066_0001
포함하는 것을 특징으로 하는 방법.
【청구항 6】
제 5항에 있어서,
상기 현재 블록이 인트라 예측된 레지듀얼일 때, 상기 변환 조합은 상기 변환 인덱스 0,1,2, 3에 대응되는 것을 특징으로 하는 방법.
【청구항 7】 2019/190284 1»(:1^1{2019/003743
65 제 5항에 있어서,
상기 현재 블록이 인터 예측된 레지듀얼일 때, 상기 변환 조합은 상기 변환 인덱스 3, 2, 1,0에 대응되는 것을 특징으로 하는 방법.
【청구항 8】
저복잡도 변환 실행에 기초하여 비디오 신호를 복원하는 장치에 있어서, 상기 비디오 신호로부터 현재 블록의 변환 인덱스를 획득하는 파싱부, 여기서 상기 변환 인덱스는 0814 및/또는 Dm 의 조합으로 구성된 복수개의 변환 조합들 중 어느 하나에 대응됨;
상기 변환 인덱스에 대응되는 변환 조합을 유도하고, 0 14를 이용하여 상기 현재 블록에 대해 수직 방향으로 역방향 변환을 수행하고, 1X74를 이용하여 상기 현재 블록에 대해 수평 방향으로 역방향 변환을 수행하는 변환부, 여기서 상기 변환 조합은 수평 변환 및 수직 변환으로 구성되고, 상기 수평 변환 및 상기 수직 변환은 상기 0814 또는 상기 DCT4 중 어느 하나에 대응됨;
상기 역방향 변환된 현재 블록을 이용하여 상기 비디오 신호를 복원하는 복원부
를 포함하는 것을 특징으로 하는 장치.
【청구항 9]
제 8항에 있어서,
상기 0814 및/또는 상기 £)(그4는 순방향 1X72 또는 역방향
Figure imgf000067_0001
이용하여 실행되는 것을 특징으로 하는 장치.
【청구항 10】
제 9항에 있어서, 2019/190284 1»(:1/10公019/003743
66 상기 0814 및/또는 상기 :0(그4는 상기 순방향 £)아2 또는 상기 역방향
1X72에 후처리 행렬 및 선처리 행렬 을 적용하는 것을 특징으로 하는
Figure imgf000068_0001
사이즈를 나타냄)
【청구항 11】
제 8항에 있어서,
상기 수직 변환이 상기 14일 때 상기 14의 역변환을 열마다 적용하고, 상기 수평 변환이 0( 4일 때 상기 1X^4의 역변환을 행마다 적용하는 것을 특징으로 하는 장치.
【청구항 12】
제 8항에 있어서,
상기 변환 조합 (수평 변환, 수직 변환)은, 必肝4, 0814), (1) 4, 14), (0814,0 4) 및 (0 4,1)014)를 포함하는 것을 특징으로 하는 장치.
【청구항 13】
제 12항에 있어서,
상기 현재 블록이 인트라 예측된 레지듀얼일 때, 상기 변환 조합은 상기 변환 인덱스 0, 1,2,3에 대응되는 것을 특징으로 하는 장치.
【청구항 14】 2019/190284 1»(:1/10公019/003743
67 제 12항에 있어서,
상기 현재 블록이 인터 예측된 레지듀얼일 때, 상기 변환 조합은 상기 변환 인덱스 3, 2, 1, 0에 대응되는 것욜 특징으로 하는 장치.
PCT/KR2019/003743 2018-03-29 2019-03-29 비디오 압축을 위한 변환 커널의 저복잡도 연산을 수행하는 방법 및 장치 WO2019190284A1 (ko)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/042,722 US20210021871A1 (en) 2018-03-29 2019-05-01 Method and apparatus for performing low-complexity operation of transform kernel for video compression

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
KR20180036418 2018-03-29
KR10-2018-0036418 2018-03-29

Publications (1)

Publication Number Publication Date
WO2019190284A1 true WO2019190284A1 (ko) 2019-10-03

Family

ID=68060603

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/KR2019/003743 WO2019190284A1 (ko) 2018-03-29 2019-03-29 비디오 압축을 위한 변환 커널의 저복잡도 연산을 수행하는 방법 및 장치

Country Status (2)

Country Link
US (1) US20210021871A1 (ko)
WO (1) WO2019190284A1 (ko)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11665367B2 (en) * 2018-03-30 2023-05-30 Sony Corporation Image processing apparatus and method
CN113365055B (zh) * 2019-03-09 2023-03-14 杭州海康威视数字技术股份有限公司 进行编码和解码的方法、解码端、编码端和系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101091479B1 (ko) * 2007-06-15 2011-12-07 콸콤 인코포레이티드 인트라 예측 모드에 따른 레지듀얼 블록들의 적응형 변환
KR20130098360A (ko) * 2010-09-08 2013-09-04 삼성전자주식회사 내부 예측을 위한 적응적 이산 코사인 변환 및 이산 사인 변환을 이용한 낮은 복잡도의 변환 코딩
KR20170117112A (ko) * 2015-03-06 2017-10-20 한국과학기술원 저 복잡도 변환에 기반한 영상 부호화 및 복호화 방법 및 이를 이용하는 장치
US20180020218A1 (en) * 2016-07-15 2018-01-18 Qualcomm Incorporated Look-up table for enhanced multiple transform
KR20180028468A (ko) * 2015-08-06 2018-03-16 엘지전자 주식회사 싱글톤 계수 업데이트를 이용하여 변환을 수행하는 방법 및 장치

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101091479B1 (ko) * 2007-06-15 2011-12-07 콸콤 인코포레이티드 인트라 예측 모드에 따른 레지듀얼 블록들의 적응형 변환
KR20130098360A (ko) * 2010-09-08 2013-09-04 삼성전자주식회사 내부 예측을 위한 적응적 이산 코사인 변환 및 이산 사인 변환을 이용한 낮은 복잡도의 변환 코딩
KR20170117112A (ko) * 2015-03-06 2017-10-20 한국과학기술원 저 복잡도 변환에 기반한 영상 부호화 및 복호화 방법 및 이를 이용하는 장치
KR20180028468A (ko) * 2015-08-06 2018-03-16 엘지전자 주식회사 싱글톤 계수 업데이트를 이용하여 변환을 수행하는 방법 및 장치
US20180020218A1 (en) * 2016-07-15 2018-01-18 Qualcomm Incorporated Look-up table for enhanced multiple transform

Also Published As

Publication number Publication date
US20210021871A1 (en) 2021-01-21

Similar Documents

Publication Publication Date Title
CN112166613B (zh) 使用缩减二次变换来处理视频信号的方法和装置
RU2745021C1 (ru) Способ и устройства для конфигурирования преобразования для сжатия видео
TWI771679B (zh) 以區塊為基礎之預測技術
US11368691B2 (en) Method and device for designing low-complexity calculation DST7
US11350130B2 (en) Method and apparatus for processing video signal by using approximation transform on basis of preprocessing/postprocessing matrix
CN111615830A (zh) 基于选择性变换的图像编码方法和用于该方法的装置
US20240114159A1 (en) Image encoding/decoding method and device therefor
CN113170197A (zh) 基于二次变换的图像编码方法及其装置
CN113273214A (zh) 基于二次变换的图像编码方法及其装置
WO2019190284A1 (ko) 비디오 압축을 위한 변환 커널의 저복잡도 연산을 수행하는 방법 및 장치
WO2019194503A1 (ko) 분할된 블록에 2차 변환을 적용하여 비디오 신호를 처리하는 방법 및 장치
CN113940074A (zh) 基于广角帧内预测和变换的图像编码方法和装置
CN113228685A (zh) 用于基于二次变换编码图像的方法及其装置
WO2019194505A1 (ko) 축소된 회전기반 변환을 이용하여 비디오 신호를 처리하는 방법 및 장치
RU2795258C1 (ru) Способ и устройство для конфигурирования преобразования для сжатия видео
US11290748B2 (en) Method and device for designing low complexity DST7
CN117652143A (zh) 用于设计低频不可分离变换的方法和装置
WO2019190282A1 (ko) 곱셈 없는 회전 기반 변환에 기초하여 비디오 신호를 처리하는 방법 및 장치

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19777196

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19777196

Country of ref document: EP

Kind code of ref document: A1