KR101668851B1 - Method and apparatus of adaptive coding mode ordering for early mode decision of hevc - Google Patents

Abstract

The present invention relates to an image encoding and decoding technique, and a method of determining an encoding mode according to an embodiment of the present invention includes: predicting a rate- Calculating a predicted value of the amount; And a predicted value of the rate-distortion cost is compared with a predetermined optimum rate-distortion cost value, and the calculated bit-amount predictive value is compared with a preset optimum bit-amount to perform a coding operation on the encoding mode Determining whether to omit it; Determining a coding mode for the current coding block using the determined minimum rate-distortion cost value; And updating the order of the encoding mode after performing the encoding. According to an embodiment of the present invention, an adaptive coding mode ordering method and apparatus for determining an HEVC fast coding mode can be provided.

Description

Field of the Invention < RTI ID = 0.0 > [0001] < / RTI > The present invention relates to an adaptive coding mode ordering method and apparatus for determining an HEVC fast encoding mode,

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to image encoding and decoding techniques, and more particularly, to an adaptive encoding mode ordering method and apparatus for determining a high-efficiency video coding (HEVC) fast encoding mode.

Recently, as broadcasting system supporting high definition (HD) resolution has expanded not only in Korea but also in the world, many users are getting used to high definition and high definition video. In addition, many organizations are spurring the development of next generation video equipment. In addition, as interest in ultra high definition (UHD) with resolution of four times or more of HDTV is increased along with high definition television (HDTV), compression technology for higher resolution and high image quality is required .

For image compression, pixel information of the current picture can be encoded by prediction. For example, an inter prediction technique for predicting pixel values included in a current picture from temporally preceding and / or following pictures, an inter prediction technique for predicting pixel values included in a current picture using pixel information in the current picture, ) Prediction technique can be applied.

In addition, an entropy encoding technique for assigning a short code to a symbol having a high appearance frequency and allocating a long code to a symbol having a low appearance frequency can be applied to reduce coding efficiency and the amount of information to be transmitted.

In the ITU-T / ISO / IEC Joint Collaborative Team on Video Coding (JCT-VC), High Efficiency Video Coding (HEVC) having a coding efficiency twice as high as that of the existing AVC / (FDIS) was published in January 2013, and the standardization work has been completed. Therefore, operators who want to use high-definition video service are increasingly demanding to replace the existing AVC / H.264 codec with HEVC. However, in HEVC technology, many techniques have been added to achieve high coding performance, which has greatly increased the complexity of the encoder. Therefore, in order to utilize HEVC technology in the market, it is necessary to reduce the complexity of the encoder and improve the coding speed accordingly.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and it is an object of the present invention to provide an adaptive coding mode ordering method and apparatus for determining a High Efficiency Video Coding (HEVC) fast coding mode.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are not restrictive of the invention, unless further departing from the spirit and scope of the invention as defined by the appended claims. It will be possible.

According to another aspect of the present invention, there is provided a method for determining a coding mode, the method comprising: calculating a predicted value of a rate-distortion cost and a predicted bit value for coding in a current coding mode through an initialization step; And a predicted value of the rate-distortion cost is compared with a predetermined optimum rate-distortion cost value, and the calculated bit-amount predictive value is compared with a preset optimum bit-amount to perform a coding operation on the encoding mode Determining whether to omit it; Determining a coding mode for the current coding block using the determined minimum rate-distortion cost value; And updating the order of the encoding mode after performing the encoding.

According to an embodiment of the present invention, an adaptive coding mode ordering method and apparatus for determining a High Efficiency Video Coding (HEVC) fast coding mode can be provided.

In addition, it can have excellent performance both in encoding speed and efficiency.

The effects obtained by the present invention are not limited to the above-mentioned effects, and other effects not mentioned can be clearly understood by those skilled in the art from the following description will be.

1 is an example of a block diagram of a coding apparatus according to an embodiment of the present invention.
2 is a block diagram of a decoding apparatus according to an embodiment of the present invention.
3 is a diagram illustrating an example of a CU divided into a quadtree structure.
4 is a diagram showing an example of a PU division mode applied to a CU.
5 is a diagram showing an example of a flowchart for determining a coding mode of a CU.
6 is a diagram illustrating an example of a quadtree structure having a minimum rate-distortion cost in the CTU.
FIG. 7 is a block diagram illustrating a random access configuration of an encoder according to an exemplary embodiment of the present invention. Referring to FIG.
FIG. 8 is a flowchart illustrating a method for determining a CU coding mode in the FMD step according to an exemplary embodiment of the present invention. Referring to FIG.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

In describing the embodiments, descriptions of techniques which are well known in the art to which the embodiments of the present invention belong, and which are not directly related to the embodiments of the present specification are not described. This is for the sake of clarity of the gist of the embodiment of the present invention without omitting the unnecessary explanation.

When an element is referred to herein as being "connected" or "connected" to another element, it may mean directly connected or connected to the other element, Element may be present. In addition, the content of "including" a specific configuration in this specification does not exclude a configuration other than the configuration, and means that additional configurations can be included in the scope of the present invention or the scope of the present invention.

Also, the terms first, second, etc. may be used to describe various configurations, but the configurations are not limited by the term. The terms are used for the purpose of distinguishing one configuration from another. For example, without departing from the scope of the present invention, the first configuration may be referred to as the second configuration, and similarly, the second configuration may be named as the first configuration.

In addition, the components shown in the embodiments of the present invention are shown independently to represent different characteristic functions, and do not mean that each component is composed of separate hardware or one software constituent unit. That is, each constituent unit is included in each constituent unit for convenience of explanation, and at least two constituent units of each constituent unit may form one constituent unit or one constituent unit may be divided into a plurality of constituent units to perform a function. The integrated embodiments and the separate embodiments of each component are also included in the scope of the present invention unless they depart from the essence of the present invention.

In addition, some of the components are not essential components to perform essential functions in the present invention, but may be optional components only to improve performance. The present invention can be implemented only with components essential for realizing the essence of the present invention, except for the components used for the performance improvement, and can be implemented by only including the essential components except the optional components used for performance improvement Are also included in the scope of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In the following description of the embodiments of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present disclosure rather unclear. DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. The following terms are defined in consideration of the functions of the present invention, and these may be changed according to the intention of the user, the operator, or the like. Therefore, the definition should be based on the contents throughout this specification.

1 is an example of a block diagram of a coding apparatus according to an embodiment of the present invention.

1, an encoding apparatus 100 according to an exemplary embodiment of the present invention includes a motion prediction unit 111, a motion compensation unit 112, an intra prediction unit 120, a switch 115, a subtractor 125, An inverse quantization unit 160, a transform unit 170, an adder 175, a filter unit 180, and a reference image buffer 190, And the like.

At this time, the encoding apparatus 100 may perform encoding in an intra mode or an inter mode with respect to an input image and output a bit stream. Intra prediction is intra prediction, and inter prediction is inter prediction. In the intra mode, the switch 115 is switched to the intra mode, and in the inter mode, the switch 115 can be switched to the inter mode. The image encoding apparatus 100 may generate a prediction block for an input block of an input image, and then may code a residual between the input block and the prediction block. At this time, the input image may mean an original picture.

In the intra mode, the intra prediction unit 120 may generate a prediction block by performing spatial prediction using the pixel value of the already coded block around the current block.

In the case of the intermodule, the motion predicting unit 111 may find a motion vector by searching an area of the reference image stored in the reference image buffer 190, which is best matched with the input block, in the motion prediction process. The motion compensation unit 112 may generate a prediction block by performing motion compensation using a motion vector. Here, the motion vector is a two-dimensional vector used for inter prediction, and can represent an offset between a current block and a block in the reference image.

In FIG. 1, the motion prediction unit 111, the motion compensation unit 112, and the intra prediction unit 120 are shown as separate components, but the present invention is not limited thereto. For example, the motion prediction unit 111 and the motion compensation unit 112 may constitute one inter prediction unit, and the motion prediction unit 111, the motion compensation unit 112, and the intra prediction unit 120 may generate one prediction And the like.

The subtractor 125 may generate a residual block by a difference between the input block and the generated prediction block. The transforming unit 130 may perform a transform on the residual block to output a transform coefficient. The quantization unit 140 may quantize the input transform coefficient according to the quantization parameter to output a quantized coefficient.

The entropy encoding unit 150 may output a bit stream by performing entropy encoding based on the values calculated by the quantization unit 140 or the encoding parameter values calculated in the encoding process.

When entropy encoding is applied, a small number of bits are allocated to a symbol having a high probability of occurrence, and a large number of bits are allocated to a symbol having a low probability of occurrence, thereby expressing symbols, The size of the column can be reduced. Therefore, the compression performance of the image encoding can be enhanced through the entropy encoding. The entropy encoding unit 150 may use an encoding method such as Exponential-Golomb Code, Context-Adaptive Variable Length Coding (CAVLC), and Context-Adaptive Binary Arithmetic Coding (CABAC) for entropy encoding.

Since the image encoding apparatus according to the embodiment of FIG. 1 performs inter prediction encoding, that is, inter-view prediction encoding, the currently encoded image needs to be decoded and stored for use as a reference image. Accordingly, the quantized coefficients are inversely quantized in the inverse quantization unit 160 and inversely transformed in the inverse transformation unit 170. The inverse quantized and inverse transformed coefficients become reconstructed residual blocks and added to the prediction block through an adder 175 to generate reconstructed blocks.

The restoration block passes through the filter unit 180 and the filter unit 180 applies at least one of a deblocking filter, a sample adaptive offset (SAO), and an adaptive loop filter (ALF) can do. The filter unit 180 may be referred to as an in-loop filter. The deblocking filter may remove block distortion and / or blocking artifacts that have occurred at the boundary between the blocks. The SAO may add a proper offset value to the pixel value to compensate for coding errors. ALF can perform filtering based on the comparison between the reconstructed image and the original image. The restoration block having passed through the filter unit 180 may be stored in the reference image buffer 190.

2 is a block diagram of a decoding apparatus according to an embodiment of the present invention.

2, the image decoding apparatus 200 includes an entropy decoding unit 210, an inverse quantization unit 220, an inverse transform unit 230, an intra prediction unit 240, a motion compensation unit 250, an adder 255, a filter unit 260, a reference image buffer 270, and the like.

The video decoding apparatus 200 receives the bit stream output from the encoder and decodes the video stream into the intra mode or the inter mode, and outputs the reconstructed video, that is, the reconstructed video. In the intra mode, the switch is switched to the intra mode, and in the inter mode, the switch can be switched to the inter mode. The image decoding apparatus 200 may obtain a reconstructed residual block from the input bitstream, generate a prediction block, and add the reconstructed residual block and the prediction block to generate a reconstructed block, i.e., a reconstructed block .

The entropy decoding unit 210 may entropy-decode the input bitstream according to a probability distribution to generate symbols including a symbol of a quantized coefficient type. The entropy decoding method is similar to the entropy encoding method described above.

When the entropy decoding method is applied, a small number of bits are assigned to a symbol having a high probability of occurrence, and a large number of bits are assigned to a symbol having a low probability of occurrence, so that the size of a bit string for each symbol is Can be reduced.

The quantized coefficients are inversely quantized using the quantization parameters in the inverse quantization unit 220 and inversely transformed in the inverse transformation unit 230. The reconstructed residual blocks can be generated as a result of inverse quantization / inverse transformation of the quantized coefficients.

In the intra mode, the intraprediction unit 240 may generate a prediction block by performing spatial prediction using the pixel value of the already decoded block around the current block. In the inter mode, the motion compensation unit 250 can generate a prediction block by performing motion compensation using a motion vector and a reference image stored in the reference image buffer 270. [

In the embodiment of FIG. 2, the intraprediction unit 240 and the motion compensation unit 250 are shown as separate components, but the present invention is not limited thereto. For example, the intraprediction unit 240 and the motion compensation unit 250 may constitute one prediction unit.

The reconstructed residual block and the prediction block are added through the adder 255, and the added block can be passed through the filter unit 260. The filter unit 260 may apply at least one of a deblocking filter, SAO, and ALF to a restoration block or a restored picture. The filter unit 260 may output a reconstructed image, that is, a reconstructed image. The restored image is stored in the reference image buffer 270 and can be used for inter prediction.

Hereinafter, a block means a unit of encoding and decoding. In the encoding and / or decoding process, the image is divided into a predetermined size and then encoded and / or decoded. Therefore, the block may be referred to as a macro block (MB), a coding unit (CU), a prediction unit (PU), a transform unit (TU) It may be divided into sub-blocks of small size.

On the other hand, in a conventional video encoder, a high-efficiency video coding (HEVC) encoder performs coding based on a 16 × 16 (macroblock) pixel. However, in a high efficiency video coding (HEVC) . CTUs can have sizes of 64x64, 32x32, and 16x16, and generally use 64x64. Quadtree coding is performed based on the CTU. In the CTU, prediction and transcoding are performed based on a coding unit (CU) corresponding to each block divided into a quadtree type. Lt; / RTI > The CU can have values of 64x64, 32x32, 16x16, and 8x8, and has a quad tree structure in the CTU, and is configured to minimize the rate-distortion cost.

3 is a diagram illustrating an example of a CU divided into a quadtree structure.

Referring to FIG. 3, there is shown a CU that requires encoding in order to find a structure in which a 64x64 CTU has a quad tree structure and has a minimum rate-distortion cost. When encoding is performed based on CTU, the CTU can be divided based on a quadtree structure to find a coding structure having a minimum rate-distortion cost. The quadtree structure is a recursive tree structure, and can be divided into four child nodes with one CTU as a root node. Then, the divided child nodes can be divided into four child nodes for each parent node.

For example, in a 64x64 CTU with YUV 4: 2: 0 format, if the optimal structure for the Y component is sought, the 64x64 CTU is divided into a quad tree structure to form one 64x64 CU 310, four 32x32 CUs 320, 16 16x16 CUs 330, and 64 8x8 CUs 340. As shown in FIG. At this time, the coding scheme having a minimum rate-distortion cost can be found by performing coding in a coding mode that the CU can have for each size of CU.

On the other hand, the coding mode used in the HEVC technology includes an inter mode for performing inter picture prediction and an intra mode for performing intra picture prediction as described above.

4 is a diagram showing an example of a PU division mode applied to a CU.

As described in connection with FIG. 3, the CTU may be divided into a quad tree structure, for example, 64x64, 32x32, 16x16, and 8x8 size CUs. The CU of each size can be coded in a mode having a minimum rate-distortion cost among a skip mode, a merge mode, an inter mode, and an intra mode. At this time, the CU of each size can be encoded into a partitioned block shape according to the encoding mode.

At this time, as shown in FIG. 4, the CU of each size may have eight types of prediction unit (PU) partition mode. In this case, in the inter prediction mode, the CU is divided into a maximum of eight division modes (2N x 2N, 2N x N, N x 2N, N x N, 2N x nU, 2N x nD, nL x 2N, 420, 430, 440, 450, 460, 470, 480). In the intra prediction mode, the CU can be coded using up to two divided modes (2N x 2N, N x N) 410 and 440. Then, by checking the coding modes, an encoding mode having a minimum rate-distortion cost can be found.

At this time, the rate-distortion cost can be calculated according to the following equation (1).

D is the Sum of Square Error (SSE) value for the difference between the original image and the reconstructed image, λ is a Lagrangian constant, B is the bit generated during encoding, . At this time, the Lagrangian constant value can be determined by a variable such as a QPI, a slice type, and a GOP (group of pictures) structure.

FIG. 5 illustrates an example of a flowchart for determining a coding mode of a CU, and FIG. 6 illustrates an example of a quadtree structure having a minimum rate-distortion cost in a CTU.

Referring to FIG. 5, the encoder can find a quadtree structure having a minimum rate-distortion cost by calculating a bit amount B generated in coding according to a mode and a rate-distortion cost J corresponding thereto.

First, in step 510, the maximum value (J _max ) of the J value and the maximum value (B _max ) of the B value can be set to the optimum J value (J _best ) and the optimum B value (B _best ), respectively. In step 520, the J value for each mode M _i can be calculated.

Thereafter, in step 530, the J value for the i-th mode M _i calculated in step 520 may be compared with the currently set optimum J value J _best . And the comparison i-th mode (M _i) J is a currently established optimal value J (J _best) is less than the current calculated in 540 steps i-th mode, the J value optimum J values for (M _i) of the (J _best ). And the current i-th mode ( _Mi ) can be set to the optimum mode.

On the other hand, the result of comparison, if it is less than the i th mode (M _i) optimal value J (J _best) J values are currently set for the, whether Check the J value for all candidate modes in step 550. In the step 530 Can be determined. If the J value of all the candidate modes is checked as a result of the determination, the procedure is terminated. Otherwise, the J value of the next mode may be checked in step 560. That is, if the J value for the i-th mode (M _i ) is not smaller than the currently set optimum J value (J _best ), the currently set optimum J value (J _best ) Value and compares it with the set optimal J value (J _best ).

On the other hand, according to the embodiment shown the best of the J value by calculating back the J value only if it is not be considered to determine whether the J values are taken into account for the next mode at step 570 is set this current, as (J _best ). ≪ / RTI >

Referring to FIG. 6, an example of a quadtree structure having the minimum rate-distortion cost thus determined in the CTU is shown. For example, when a 64x64 CTU carries out a quad-tree-based encoding process as described with reference to FIG. 5, a CTU having a size of 64x64 as shown in FIG. 6 has a quad- Can be derived in a form that minimizes the distortion cost.

In such a case, it is necessary to find a coding mode having a minimum rate-distortion cost by checking a maximum of 10 coding modes in each CU. Therefore, the process of finding a quadtree structure having a minimum rate- It has a lot of complexity.

In order to reduce the complexity, HM (HEVC Test Model) which is a reference SW of the HEVC, a method for quickly determining a skip and a merge in a skip / merge 2N x 2N coding mode (JCTVC (JCTVC-E316), a method of omitting intra prediction when the inter prediction residual signal is small (JCTVC-N1002v1), and the like can be applied . However, these methods have difficulty in reflecting the characteristics of the moving picture and the characteristics of the coding structure that change with time because the order of the mode in which the rate-distortion cost is calculated is fixed.

Accordingly, an adaptive encoding mode ordering method and apparatus for determining an HEVC fast encoding mode according to an embodiment of the present invention includes an initialization step and a fast mode decision (FMD) decision step. And thus the order of the encoding mode for calculating the minimum rate-distortion cost according to the distribution of the best mode may be updated. Meanwhile, the terms 'initialization step' and 'fast mode determination step' are for convenience of description and are not necessarily interpreted to be limited to these terms.

FIG. 7 is a block diagram illustrating a random access configuration of an encoder according to an exemplary embodiment of the present invention. Referring to FIG.

에서 s는 부호화 단계를 나타내는데, I는 초기화(initialization) 단계를 나타내고, F는 고속 모드 결정(FMD) 단계를 나타낸다. l과 k는 템포럴 레이어(temporal layer) 및 해당 단계의 템포럴 레이어(temporal layer)에서 부호화되는 순서를 각각 나타낸다. 이때, 초기화 단계에서는 FMD 단계에서 사용될 초기 파라미터와 부호화 모드 순서가 도출될 수 있다. According to an embodiment of the present invention, a coding structure that supports random access such as a broadcasting service can be applied. Referring to FIG. 7, there is shown a random access configuration of a configuration of an encoder, for example, an HM encoder. In Figure 7, the numbers to the left of the rectangle are the coding order of the inter-coded frames.

Where s denotes an encoding step, I denotes an initialization step, and F denotes a fast mode decision (FMD) step. l and k denote the order in which they are coded in the temporal layer and the temporal layer of the corresponding stage, respectively. At this time, the initialization parameters to be used in the FMD step and the coding mode order can be derived in the initialization step.

FIG. 8 is a flowchart illustrating a method for determining a CU coding mode in the FMD step according to an exemplary embodiment of the present invention. Referring to FIG.

Referring to FIG. 8, a prediction value (J _pred value) of the rate-distortion cost and a bit amount prediction value (B _pred value) generated in encoding can be calculated using the initial parameters derived in the initialization step. If the J _pred value is greater than the J _best value and the B _pred value is greater than the B _best value, the process of calculating the rate-distortion cost for the corresponding parenting mode may be omitted.

In this case, the initial encoding mode order of the FMD step may be determined by analyzing the encoding result of the initialization step so that the distribution of the best mode M ₁ to M ₁₁ is high.

The distribution of the best mode is calculated for each CU size of the temporal layer. In the FMD step, the coding mode order can be updated every time the frame is coded.

At this time, J _pred can be determined by the following formula (2) and (3) as a predicted value of the rate-distortion cost (or estimated (prediction) rate-distortion cost).

Here, l denotes a temporal layer, and k denotes a coding order. H denotes the height of the CU, and i denotes the coding mode index of the CU.

At this time, the encoding mode M _i may be as follows according to the value of i.

When i is 1, the encoding mode may be a 2N x 2N block-shaped skip mode.

When i is 2, the encoding mode may be a merge mode of 2N x 2N block shape.

When i is 3, the encoding mode may be a 2N x 2N block-shaped inter mode.

When i is 4, the encoding mode may be an N x 2N block-shaped inter mode.

When i is 5, the encoding mode may be an inter mode of a 2N x N block shape.

When i is 6, the encoding mode may be an inter mode of a 2N x nD block shape.

When i is 7, the encoding mode may be an inter mode of a 2N x nU block shape.

When i is 8, the encoding mode may be nL x 2N block-shaped inter mode.

When i is 9, the encoding mode may be nR x 2N block-shaped inter mode.

When i is 10, the encoding mode may be a 2N x 2N block-shaped intra mode.

When i is 11, the encoding mode may be an intra mode of N x N block shape.

는 템포럴 레이어(temporal layer)(l), 부호화 순서(k)에서 얻어진 샘플 회귀선(sample regression line)의 기울기로, 다음 [수학식 4]에 의해 계산될 수 있다. On the other hand, min {} is an operator that outputs the minimum value.

Is a slope of a sample regression line obtained in the temporal layer 1 and the coding order k, and can be calculated by the following equation (4).

의 템포럴 네이버(temporal neighbor)에 위치하는 CTU의 율-왜곡 비용을 나타내는 확률 변수이다. Where J and J- ₁ are

Is a probability variable that represents the rate-distortion cost of a CTU located in the temporal neighbor of.

는 템포럴 레이어(temporal layer)(l), 부호화 순서(k)에서 얻어진 상관 계수로, 다음 [수학식 5]에 의해 계산될 수 있다.

Is a correlation coefficient obtained in the temporal layer 1 and the coding order k, and can be calculated by the following equation (5).

Where X and X _-1 are probability variables for rate-distortion cost of the CTU at the same location in the current CTU and temporal neighbor, respectively. And, std (X) represents the standard deviation of X, and cov (X, X _-1 ) represents the covariance of X and X _-1 .

On the other hand, d _-1 is the distortion cost of the CTU at the same position of the temporal neighbor.

와

는 샘플 회귀선(sample regression line)의 기울기와 상관 계수에 대한 웨이팅 계수(weighting factor)이다. m₁은 비트 비용에 대한 최대 값을 제한하기 위한 변수이다. 그리고

은 확률 변수 R에 대한 누적 분포(cumulative distribution) 함수로, R은 h, l, M_i에 해당하는 CU의 비트 비용을 나타낸다. 또한, 상기 [수학식 3]에 나타난 것과 같이 k 값이 2 이상인 경우에는, k-1인 경우의 R 값을 이용하여 R을 계산할 수 있다.

은 비트 비용에 대한 웨이팅 계수(weighting factor)이다.

는

의 비트 량으로,

는 초기화(initialization) 단계의 동일한 템포럴 레이어(l)에서 평균 비트량을 나타낸다.

와 는 각각 초기화(initialization) 단계의 동일한 템포럴 레이어(l)에서 얻어지는 기울기와 상관 계수 값들의 평균 값을 의미한다. Then, R (l, k, h, i) represents the bit cost.

Wow

Is the slope of the sample regression line and the weighting factor for the correlation coefficient. m ₁ is a variable for limiting the maximum value for the bit cost. And

Is a cumulative distribution function for the random variable R, where R represents the bit cost of the CU corresponding to h, l, M _i . When the k value is 2 or more as shown in Equation (3), R can be calculated using the R value in the case of k-1.

Is the weighting factor for the bit cost.

The

As the bit amount of "

Represents the average amount of bits in the same temporal layer (l) of the initialization step.

Wow Means an average value of the slope and correlation coefficient values obtained in the same temporal layer 1 of the initialization step, respectively.

On the other hand, the B _pred value is a predicted bit amount (or estimated (predicted) bit amount) generated in encoding, and can be determined by the following equation (6).

는 상기 [수학식 3]의 m₁,

과 동일한 역할을 하는 값이다. 또한, 상기 [수학식 6]에 나타난 것과 같이 k 값이 2 인 경우에는, k가 1인 경우의 B_pred 값을 이용하여 B_pred 값을 계산하고, k 값이 2보다 큰 경우에는 k가 1인 경우의 B_pred 값과 k-1인 경우의 B_pred 값을 이용하여 B_pred 값을 계산할 수 있다.At this time, m ₂ ,

Is expressed by m ₁ ,

Is a value that plays the same role as. Further, if the k value is 2, as shown in [Equation 6] it is, for calculating a B _pred values using the B _pred values when k is 1, and a k value greater than 2, k is 1, _pred the B value can be calculated by using a B _pred values when the B value _pred k-1 and in the case of.

,

,

, m₁, p₁}은 {0.5, 0.5, 0.8, 3, 0.01}을 사용하고, {m₂,

, p₂}는 {2, 0.5, 0.003}을 사용할 수 있다. The parameter values of the equations (2), (3) and (6) can be determined through experiments. For example, in one embodiment of the present invention,

,

,

, M _1, p _1}, and is used to {0.5, 0.5, 0.8, 3 , 0.01}, {m 2,

, p ₂ } can be {2, 0.5, 0.003}.

Referring back to FIG. 8, in step 810, the maximum value J _max (J _max ) of the J value is set as the optimal J value (J _best ) and the optimum B value (B _best ) for the first coding mode ) And the maximum value (B _max ) of the B value can be set. And you can set the bSkipFlag value to false.

In step 815, it can be determined whether the value of l is greater than one. That is, it can determine whether the current temporal layer 1 is the first one.

If it is determined in step 815 that the value of l is not greater than 1, it may be determined in step 825 whether i is greater than 3.

If it is determined in step 825 that the i value is not greater than 3, it may be determined in step 830 whether i is greater than 2 and bSkipFalg is set to false. If it is determined in step 830 that the i value is not greater than 2 (i is 1 or bSkipFlag is true), the process proceeds to step 855 where the current mode M _i The J value and the B value can be calculated.

On the other hand, if it is determined in step 830 that the i-value is greater than 2 and bSkipFalg is false, it is determined whether the bit-amount predicted value (B _pred ) is less than the optimal B-value (B _best ) (B _pred (l, k, h, i) < B _best ). If the B _pred value is smaller than the B _best value, the J value and the B value for the current mode M _i can be calculated in step 855. However, if the B _pred value is not smaller than the B _best value, the calculation of the rate-distortion cost for the current encoding mode may be omitted and the process may proceed to step 870.

Further, even if the result value of i is determined in the step 825 is greater than 3, it can be determined whether the bit rate prediction value from the 835 step (B _pred value) is smaller than the optimal value of B (B _best) (B _pred (l, k, h, i) < B _best . At this time, if the B _pred value is smaller than the B _best value, the J value and the B value for the current mode M _i can be calculated in step 855. However, if the B _pred value is not smaller than the B _best value, the calculation of the rate-distortion cost for the current encoding mode may be omitted and the process may proceed to step 870.

On the other hand, if it is determined in step 815 that the value of 1 is greater than 1, it may be determined in step 820 whether i is greater than 3. If it is determined in step 820 that the i value is not greater than 3, it may be determined in step 840 whether i is greater than 2 and bSkipFalg is set to false. If it is determined that the i value is not greater than 2 (i is 1) or bSkipFlag is true, the predicted value (J _pred value) of the rate-distortion cost is calculated in step 845 it can be determined whether or not smaller than the optimal value J _{(J best) (J pred (} l, k, h, i) <J best). If the J _pred value is smaller than the J _best value, the J value and the B value for the current mode M _i can be calculated in step 855. However, if the J _pred value is not smaller than the J _best value, the calculation of the rate-distortion cost for the current encoding mode may be omitted and the process may proceed to step 870.

On the other hand, if it is determined in step 840 that the i-value is greater than 2 and the bSkipFalg is set to false, the prediction value ( _Jred value) of the rate-distortion cost is smaller than the optimal J-value (J _best ) , bit amount predicted value (B _pred value) can determine whether or not smaller than the optimal value of B (B _best) (J _pred (l, k, h, i) <J _best and B _pred (l, k , h, i) < B _best . Even if the result i determined in step 820 is greater than 3, even if the predicted value (J _pred ) of the rate-distortion cost is smaller than the optimal J value ( _Jbest ) and the bit amount predicted value (B _pred ) may be the determination of optimum whether less than the B value _{(B best) (J pred (} l, k, h, i) <J best and B pred (l, k, h, i) <B best).

If it is determined in step 850 that the J _pred value is smaller than the J _best value and the B _pred value is smaller than the B _best value, the J value and the B value for the current mode M _i can be calculated in step 855. However, if the J _pred value is not smaller than the J _best value and the B _pred value is not smaller than the B _best value, the calculation of the rate-distortion cost for the current encoding mode is omitted, and in step 870 You can proceed.

On the other hand, after calculating the J value and the B value for the current mode M _i in step 855, it can be determined whether the J value calculated in step 860 is smaller than the currently set optimum J value J _best . If the J value calculated determination result is less than J _best value, the i th mode of the current calculated in step 865 the best of the J value of the J value for the (M _i) (J _best) to set, and the current calculated i-th mode The B value for the input image M _i can be set to the optimal B value B _best . And the current i-th mode ( _Mi ) can be set to the optimum mode. In step 870, it is determined whether the J value of all candidate modes is checked.

On the other hand, without changing the result of the determination, if the calculated value J is not less than J _best value J _best value is set, B _best value and the optimum mode in the 860 step, the process proceeds to 870 steps for all candidate modes J value is checked.

As a result of the determination in step 870, if the J value is checked for all candidate modes, the procedure can be terminated. Otherwise, the J value for the next encoding mode can be checked.

In this case, it may be determined whether the encoding mode M _i currently selected in step 875 is a 2N x 2N block-shaped skip mode (skip 2N x 2N). That is, whether the current i value is 1 or not. If it is determined that the current M _i is a 2N x 2N block-shaped skip mode, bSkipFlag may be set to true in step 880. In step 885, the i value may be incremented by one. However, if it is determined in step 875 that the current M _i is not the 2N x 2N block-shaped skip mode, the process may proceed directly to step 885 to increment i by one.

After incrementing the value of i by one in step 885, the process returns to step 815 and the steps 815 to 885 described above may be repeated.

The results according to an embodiment of the present invention may be as shown in Table 1 below.

Sequence
(resolution) Number of frames
(frame rate) QPI △ ETR BD-rate _YUV HM12.1 Proposed HM12.1 Proposed Traffic
(1260x1600) 150
(30 fps) 22 -44.79 -47.64 0.55 0.24 27 -50.68 -57.22 32 -53.51 -64.93 37 -55.09 -71.36 PeopleOnStreet
(2160x1600) 150
(30 fps) 22 -30.38 -23.09 1.25 0.46 27 -36.18 -31.23 32 -40.90 -37.49 37 -44.65 -42.68 Kimono
(1920x1080) 240
(24 fps) 22 -39.94 -40.92 0.64 0.28 27 -45.59 -48.90 32 -49.59 -55.86 37 -52.71 -63.21 Cactus
(1920x1080) 500
(50 fps) 22 -39.99 -43.36 0.69 0.40 27 -47.09 -51.79 32 -50.49 -58.45 37 -52.31 -65.83 BQTerrace
(1920x1080) 600
(60 fps) 22 -34.75 -27.32 0.68 0.35 27 -49.84 -53.51 32 -53.89 -69.36 37 -55.52 -79.20 BQMall
(832x480) 600
(60 fps) 22 -40.95 -41.56 0.89 0.30 27 -45.76 -48.94 32 -49.14 -53.69 37 -51.40 -62.28 PartyScene
(832x480) 500
(50 fps) 22 -35.46 -37.79 0.45 0.34 27 -42.36 -46.99 32 -47.26 -53.37 37 -50.45 -58.40 BasketballDrill
(832x480) 500
(50 fps) 22 -36.24 -41.30 0.80 0.33 27 -41.47 -46.50 32 -45.95 -51.57 37 -49.42 -59.21 BlowingBubbles
(416x240) 500
(50 fps) 22 -26.83 -29.63 0.41 0.31 27 -43.59 -49.16 32 -47.98 -54.52 37 -51.22 -62.55 BasketballPass
(416x240) 500
(50 fps) 22 -32.49 -33.69 0.86 0.74 27 -36.90 -38.86 32 -41.50 -46.27 37 -45.83 -53.99 ChinaSpeed
(1024x768) 500
(30 fps) 22 -33.92 -27.74 0.75 0.54 27 -39.54 -38.66 32 -45.42 -48.26 37 -49.87 -58.38 SlideEditing
(1280x720) 300
(30 fps) 22 -61.18 -80.31 0.15 0.11 27 -60.82 -81.58 32 -60.35 -82.54 37 -59.82 -84.14 SlideShow
(1280x720) 500
(20 fps) 22 -53.25 -75.13 0.17 -0.06 27 -53.83 -77.50 32 -54.92 -79.48 37 -55.47 -81.29 Avg. -46.51 -54.21 0.64 0.33

At this time,? ETR (Encoding Time Reduction) indicates a decrease in the time required for coding. The BD-rate (Bjontegaard Delta-rate) is a value indicating the degree of loss of the coding efficiency. At this time,? ETR can be obtained by the following equation (7).

Here, T denotes a time required for encoding using the fast encoding mode or HM12.1 according to an embodiment of the present invention. T2 is a method for quickly determining Skip and Merge in HM12.1 (JCTVC-H178), a method for reducing the amount of computation of AMP (JCTVC-E316), a method of omitting Intra prediction when the residual signal is small JCTVC-N1002v1) is excluded, it means time required for coding.

Referring to Table 1, it can be seen that the coding efficiency according to an embodiment of the present invention improves the coding rate of about 54.21% and has a coding efficiency loss of 0.33%.

Meanwhile, although not shown, a rate control apparatus for a video parallel coder according to an embodiment of the present invention may include a control unit.

The control unit controls the apparatus to perform the operation of any one of the above-described embodiments. For example, the control unit may allocate bits to frames to be processed in parallel, determine encoding parameters, perform encoding in parallel with the determined encoding parameters, and update the actually generated bit amount.

The embodiments disclosed in the present specification and drawings are merely illustrative of specific examples for the purpose of easy explanation and understanding, and are not intended to limit the scope of the present invention. It is to be understood by those skilled in the art that other modifications based on the technical idea of the present invention are possible in addition to the embodiments disclosed herein.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, And is not intended to limit the scope of the invention. It is to be understood by those skilled in the art that other modifications based on the technical idea of the present invention are possible in addition to the embodiments disclosed herein.

100: encoding device 200: decryption device
111: motion prediction unit 112: motion compensation unit
120: intra prediction unit 115: switch
125: subtracter 130:
140: quantization unit 150: entropy coding unit
160: Inverse quantization unit 170: Inverse transform unit
175: adder 180: filter section

Claims (9)

Determining a coding mode order for at least one coding mode using parameters obtained in the coding of a previous frame for the same temporal layer in coding for any frame in the fast mode determination step;
Calculating a rate-distortion cost for each of the at least one encoding mode of the current encoded block, based on the encoding mode order;
And performing encoding of the current encoded block using an encoding mode in which the calculated rate-distortion cost is minimized,
Wherein the step of determining the encoding mode order comprises:
And determining the encoding mode order using the parameter obtained in the encoding process in the initialization step if the arbitrary frame is the first frame of the fast mode determination step,
Wherein the step of calculating the rate-distortion cost comprises:
Calculating a rate-distortion cost prediction value and a bit cost prediction value of the current coded block for an arbitrary order coding mode according to the coding mode order;
Calculating the rate-distortion cost for the random-order encoding mode if the rate-distortion cost estimate and the bit-cost estimate are less than a minimum rate-distortion cost and a minimum bit cost computed up to the previous-order encoding mode; And
If the rate-distortion cost prediction value and the bit cost prediction value are equal to or greater than the minimum rate-distortion cost and the minimum bit cost calculated up to the previous order coding mode, the calculation of the rate-distortion cost for the arbitrary order coding mode is skipped , &Lt; / RTI &gt;
The initialization process includes:
Wherein the coding process for the first n frames is a coding process for the first n frames.
Here, n is a natural number. 2. The method of claim 1, wherein determining the encoding mode order comprises:
And determining the coding mode order in a descending order of distributions of optimal modes applied in the initialization process. delete 2. The method of claim 1, wherein the rate-
Is calculated by the following equation.

Here, J _pred is the rate-distortion cost prediction value, l is the temporal layer of the coded block (temporal layer), k is the coding order of the frame located in the same temporal layer, h is the height of the encoding block, the encoding mode index i,

Is a weight for the slope of the sample regression line,

Is a weight for the correlation coefficient of the sample regression line,

Is the average of the slope of the sample regression line obtained from the temporal layer l of the coding block and the coding order k,

Is the average of the correlation coefficients obtained in the temporal layer l and coding order k of the coding block, d _-1 is the distortion cost of the coding block in the same position as the coding block in the temporal neighbor, min {} is the minimum value output operator , m ₁ is the variable which controls the maximum value for bit cost, R (l, k, h, i) is a temporal hierarchy l, the encoding procedure of the encoding block k, the encoding block height h, coding bit cost of the block index i , and lambda is a Lagrangian constant. 5. The method of claim 4,

Quot;
Is calculated by the following equation.

J and J _-1 are probability parameters for the rate-distortion cost of the coding block located in the temporal layer 1, the coding order k, and the frame located in the temporal neighbors of the coding block in the initialization process, cov J, J _-1 ) is the covariance of J and J _-1 , and var (J) is the variance of j. 5. The method of claim 4,

Quot;
Is calculated by the following equation.

Wherein, X and X _-1 is a rate of the encoded block in the same position as each coding block and the coding block in temporal Naver-probability for distortion cost variable, cov (X, X _-1) is of J and J _-1 The covariance, std (X), is the standard deviation of X. 5. The method of claim 4, wherein R (l, k, h, i)
Is calculated by the following equation.

here,

Is a cumulative distribution function for a random variable R representing the bit cost of the encoding block, P ₁ is an arbitrary real number value between 0 and 1, Z is an integer set,

Is the weight for the bit cost,

The temporal layer l of the coded block, the bit cost for the frame of the coding order k,

Is the average bit cost at the same temporal layer l of the initialization process. 2. The method of claim 1,
Is calculated by the following equation.

Where B _pred is the predicted bit cost, l is the temporal layer of the encoded block, k is the encoding order of the frame located in the same temporal layer, h is the height of the encoded block, i is the encoding mode index,

Is a cumulative distribution function for a random variable R representing the bit cost of the encoding block, P ₂ is any real value between 0 and 1, Z is an integer set, m ₂ is the rate-distortion cost estimate J _pred A variable that limits the maximum value for the bit cost,

Is a weight for bit cost,

The temporal layer l of the coded block, the bit cost for the frame of the coding order k,

Is the average bit cost at the same temporal layer l of the initialization process. The method according to claim 1,
And updating the coding mode order using parameters obtained in the coding of the arbitrary frame when the coding of the arbitrary frame is completed.

Images