KR101099257B1 - Method for fast determining encoding mode of macro block in H.264/AVC - Google Patents

Abstract

The present invention relates to a fast method for determining the encoding mode of a macroblock in H.264 / AVC to determine the encoding mode of the macroblock at high speed. The rate-distortion cost of the skip mode is smaller than the rate-distortion cost of the inter 16 × 16 mode. In this case, the early skip mode is determined, and if the rate-distortion cost of the skip mode is greater than the rate-distortion cost of the inter 16 × 16 mode, it is determined whether the rate-distortion cost of the inter 16 × 16 mode is the minimum value. If the rate-distortion cost of the inter 16 × 16 mode is the minimum value, the determination of all inter modes is omitted. If the rate-distortion cost of the inter 16 × 16 mode is not the minimum value, the inter 16 × 8 or inter 8 The optimum mode of the sub macroblock in the 8x8 mode is set depending on whether or not the x16 values are the minimum values, and there are two submacroblocks 8x8 in which the optimum mode is set, and the submacroblock 4 in which the optimum mode is set. Of × 4 two In the case of the negative, when 8 × 8 is 2 and 4 × 4 is 2, an intra 4 × 4 mode decision algorithm is performed using a 4 × 4 square mask and interblock correlation, and the rate-distortion cost is calculated. After determining the optimal mode, if 8x8 is not 2 or 4x4 is not 2, depending on whether 2 or more submacroblocks 8x8 or 2 or more submacroblocks 4x4 are used. The process of determining the prediction mode is omitted and the optimal mode is determined by calculating the rate-distortion cost.

Description

Method for fast determining encoding mode of macro block in H.264 / AVC}

The present invention relates to a fast method for determining an encoding mode of a macroblock in H.264 / AVC, which determines an encoding mode of a macroblock at high speed in H.264 / AVC. More specifically, the present invention relates to a method of determining a fast encoding mode of a macro block in H.264 / AVC, which can reduce the complexity of an encoder and a decoder by determining a fast encoding mode of a macro block in an inter mode or an intra mode.

H.264 / AVC is a digital video codec standard with very high data compression rates. The Video Coding Experts Group (VCEG) and the International Standardization Organization (ISO / IEC) of the International Telecommunications Union-Telecommunication Standardization Sector (ITU-T). The International Electro-technical Commission's Moving Picture Experts Group (MPEG) jointly formed the Joint Video Team (JVT) to standardize.

The H.264 / AVC video coding standard adds new coding tools that were not included in the previous video coding standards (MPEG2, MPEG4 Part 2, H.263). The performance is improved by about 40 to 50% compared to MPEG2 and about 30 to 40% compared to MPEG4.

In particular, the recently proposed Rate Distortion Optimization (RDO) greatly improves the compression efficiency of H.264 / AVC.

On the other hand, in H.264 / AVC, only the improvement of the compression efficiency is considered as a top priority without considering the complexity, and the computational amount of the encoder and the decoder is extremely increased to have a very high complexity.

Among the new coding schemes, the use of multiple reference frames, motion estimation in units of 1/4 pixels, and iterative rate-distortion cost (RDO_cost) function calculations for multiple prediction modes of variable block size are the complexity of the encoder and decoder. It is a major factor that increases.

In the H.264 / AVC encoder, the rate-distortion cost must be calculated for various combinations of inter and intra modes in order to select an optimal inter / intra prediction mode.

To calculate the amount of bits used in the rate-distortion cost function, DCT / Hadamard transform and quantization / dequantization, inverse DCT / Hadamard transform, and entropy encoding for each mode are performed, which greatly increases the complexity of the encoder. .

Accordingly, in order to speed up the determination of the intra mode, a method of reducing the number of candidate modes requiring the rate-distortion calculation using a Sobel mask and an edge histogram has recently been proposed.

In the above method, the intra mode decision process is omitted by using a specific threshold value, thereby speeding up the decision of the intra mode or by speeding up the decision of the reference frame by using the characteristics of the multiple reference frames, and using the selective intra mode decision method, Is reducing.

However, there is a disadvantage in that the calculation amount is not sufficiently reduced while maintaining the performance of encoding when using the specific threshold.

Therefore, an object of the present invention is to provide a fast method for determining the encoding mode of a macro block in H.264 / AVC, which can reduce the complexity of the encoder and the decoder by efficiently determining the inter mode and the intra mode of the macro blocks.

Problems to be solved by the present invention are not limited to the above-mentioned technical problems, and other technical problems not mentioned above may be clearly understood by those skilled in the art to which the present invention pertains. will be.

In the H.264 / AVC of the present invention, the method of fast determining the encoding mode of a macroblock includes determining an early skip mode when the rate-distortion cost of the skip mode is less than the rate-distortion cost of the inter 16 × 16 mode. When the rate-distortion cost of the skip mode is greater than the rate-distortion cost of the inter 16 × 16 mode, it is determined whether the rate-distortion cost of the inter 16 × 16 mode is the minimum value, Omitting the determination of all inter modes when the rate-distortion cost is the minimum value, and inter 16 × 8 or inter 8 × 16 minimum when the rate-distortion cost of the inter 16 × 16 mode is not the minimum value. Setting the optimum mode of the sub macroblock in the 8x8 mode according to whether or not the value is a value; and there are two submacroblocks 8x8 in which the optimum mode is set, and submacroblock 4 in which the optimum mode is set. Whether or not × 4 is two And 2 sub macroblocks 8 × 8 in which the optimum mode is set and 2 sub macroblocks 4 × 4 in which the optimum mode is set, and the correlation between the 4 × 4 square mask and the block is determined. Performing an intra 4 × 4 mode decision algorithm, calculating a rate-distortion cost, and then determining an optimal mode; and as a result of the determination, there are not two or optimal sub macroblocks 8 × 8 that set the optimal mode. When there are not two sub macroblocks 4 × 4 in which the mode is set, the process of determining the prediction mode is omitted according to whether two or more sub macroblocks 8 × 8 or two or more sub macroblocks 4 × 4 are used. Calculating a rate-distortion cost to determine an optimal mode.

The determining of the early skip mode may include calculating the rate-distortion cost of the skip mode and the rate-distortion cost of the inter 16 × 16 mode, and the calculated rate-distortion cost and the inter 16 × 16 mode of the skip mode. Comparing the rate-distortion cost of the step and determining the early skip mode when the rate-distortion cost of the skip mode is less than the rate-distortion cost of the inter 16 × 16 mode. It is done.

The setting of the optimum mode of the sub macroblock in the 8x8 mode includes determining whether the rate-distortion cost of the inter 16x16 mode is a minimum value, and the rate-distortion of the inter 16x16 mode. Omit all inter-mode decision processes when the cost is the minimum value, and whether the rate-distortion cost for the inter 16 × 8 mode is the minimum value when the rate-distortion cost of the inter 16 × 16 mode is not the minimum value. Or determining whether the rate-distortion cost for the inter 8 × 16 mode is a minimum value, and if the rate-distortion cost for the inter 16 × 8 mode is a minimum value, inter 16 × 16 mode, inter 8 × Setting the optimum mode of the sub macroblock in the 8x8 mode by omitting the mode decision process for the 16 mode and the sub macroblock 4x8 mode, and the rate-distortion cost for the inter 8x16 mode is the minimum value. In case of inter 16 × 16 mode, inter 16 × 8 Setting the optimal mode of the sub macroblock in the 8x8 mode by omitting the mode decision process for the mode and the submacroblock 8x4 mode, the rate-distortion cost for the inter16x8 mode, and the inter 8 And setting the optimum mode of the sub macroblock in the 8x8 mode when the rate-distortion cost for the x16 mode is not the minimum value.

The method of determining a prediction mode according to whether there are two or more sub macroblocks 8 × 8 or two or more sub macroblocks 4 × 4 is omitted, and determining an optimal mode by calculating a rate-distortion cost. When the number of sub macroblocks 8 × 8 is two or more, skipping an intra 4 × 4 prediction mode determination step and calculating a rate-distortion cost to determine an optimal mode; and two sub macroblocks 4 × 4 In this case, the process of determining an intra 16 × 16 prediction mode is omitted, calculating a rate-distortion cost to determine an optimal mode, and the sub macroblock 8 × 8 is not two or more, and the sub macroblock 4 And determining the optimal mode by calculating the rate-distortion cost when x4 is not two or more.

In the H.264 / AVC of the present invention, the encoding mode fast decision method of the macroblock has an effect that the generated bit rate is greatly reduced, as a result, the overall encoding efficiency is improved and the encoding time is greatly reduced.

Hereinafter, the present invention will be described in more detail with reference to the accompanying drawings, which do not limit the present invention, and like reference numerals designate like elements in some drawings.
FIG. 1A illustrates an intermode for visual prediction in H.264 / AVC; FIG.
1B is a diagram illustrating an intra mode for prediction in a spatial domain in H.264 / AVC;
FIG. 2A is a diagram for describing each prediction mode of the intra 16 × 16 mode of FIG. 1B in more detail. FIG.
2B is a diagram for explaining a prediction mode of an intra 4x4 mode in more detail;
3 is a diagram illustrating an example of a similarity between blocks and a prediction direction according to pixel information of each block in an intra 4 × 4 mode;
4 and 5 are diagrams for explaining a method of reducing a calculation process for determining an optimal mode in an intra 4x4 mode according to the present invention;
6 shows an optimal mode distribution of a submacroblock, and
7A and 7B are signal flow diagrams illustrating an operation of a preferred embodiment of a fast mode for encoding an encoding mode of a macroblock according to the present invention.

The following detailed description is only illustrative, and merely illustrates embodiments of the present invention. In addition, the principles and concepts of the present invention are provided for the purpose of explanation and most useful.

Accordingly, various forms that can be implemented by those of ordinary skill in the art, as well as not intended to provide a detailed structure beyond the basic understanding of the present invention through the drawings.

FIG. 1A illustrates an intermode for visual prediction in H.264 / AVC, and FIG. 1B illustrates an intramode for prediction in spatial domain in H.264 / AVC.

The H.264 / AVC international video standard uses a motion prediction and mode determination method. The motion prediction and mode determination method employs a motion compensation technique using a variable block, and the size of the macroblock is from 16 × 16 to 4 × 4. After dividing, rate-distortion optimization technique is used, and inter mode and intra mode are set separately for video processing.

Referring to FIG. 1A, five inter modes are supported according to a unit size of a macroblock as a mode for visual prediction. That is, there are a skip mode, an inter 16x16 mode, an inter 8x16 mode, an inter 16x8 mode, and an inter 8x8 mode.

8 × 8 macroblocks included in the inter 8 × 8 mode may be divided into sub-macroblocks having sizes of 8 × 8, 8 × 4, 4 × 8, and 4 × 4.

Referring to FIG. 1B, three intra modes are supported according to a unit size of a macroblock as a mode for prediction in a spatial domain. That is, there are intra 16x16 mode (luminance), intra 8x8 mode (color), and intra 4x4 mode (luminance). Since the intra 8x8 mode is provided only in the FRExt profile, the present invention will be described with the exception of the intra 8x8 mode.

Intra 16x16 mode supports four prediction modes and directions, and intra 4x4 mode supports eight directional modes and a DC mode, which is an averaged prediction mode without directionality. Intra mode predicts the current macroblock using the boundary pixels of the previously encoded valid neighboring macroblocks (pixels on the left, top, top right or top left).

FIG. 2A is a diagram for describing each prediction mode in detail in the intra 16 × 16 mode of FIG. 1B, and FIG. 2B is a diagram for describing the prediction mode in the intra 4 × 4 mode in more detail.

Referring to FIGS. 2A and 2B, pixels of a block to be currently encoded are predicted by using previously encoded pixels of the upper, left, upper right, or upper left, and the intra 16 × 16 mode predicts using the upper pixels. Mode 0 (vertical prediction), Prediction mode 1 (horizontal prediction) using the pixels on the left side, Prediction mode 2 (average value prediction), using the left and top pixels Prediction mode 3 (average prediction) has a prediction, and the intra 4x4 mode has a total of nine prediction modes from prediction mode 0 to prediction mode 8.

H.264 / AVC calculates the rate-distortion cost of each of the various modes as described above, and determines the optimal mode to encode the mode having the minimum value.In the case of intra 4 × 4 mode, according to a total of nine prediction directions It has a relatively high complexity compared to intra 16 × 16 because it is necessary to determine the optimal mode and perform the prediction coding on the granular block.

For example, as illustrated in FIG. 3, pixel information of a current block (a block to be currently encoded) and a neighboring block are similar in a region where a change of a video signal is small in a general image such as a surrounding background. In addition, if the optimal prediction direction of the current block is the same as the prediction direction of the neighboring block in the intra 4 × 4 mode, the pixel information of the current block and the neighboring block is also likely to be similar.

Therefore, when the similarity between blocks is high, the current information is determined using the neighboring mode information that has already been determined, rather than calculating the rate-distortion cost for each of the nine candidate modes to determine the optimal mode of the current block of the intra 4 × 4 mode. Determining the optimal mode of the block reduces the complexity.

That is, in the present invention, in order to reduce complexity by encoding a macroblock, the rate-distortion cost for each of the nine candidate modes is not calculated in the intra 4 × 4 mode, but the mode of the predetermined neighboring block is calculated. The optimal mode is determined by using the information. First, the similarity between blocks according to the difference between the pixel distribution of the current block and the pixel distribution of the neighboring block is determined to reduce the number of candidate modes.

In more detail through the equation, the pixel distribution diagram Cdis in the current block, the pixel distribution difference Lcor between the current block and the left block, and the pixel distribution difference Ucor between the current block and the upper block are as follows. It is obtained through Equation 1.

Here, Cavr is an average value of pixels in the current block, and each of CPx, LPx, and UPx is a pixel value corresponding to the x index of the current, left and upper blocks.

The inter-block difference Ldif according to the pixel value distribution between the current block and the left block and the pixel distribution difference Udif between the current block and the upper block are obtained through Equation 2 below.

Here, Cdis is a pixel distribution in the current block, Lcor is a pixel distribution difference between the current block and the left block, and Ucor is a pixel distribution difference between the current block and the upper block.

When the calculated Ldif and Udif are compared with a preset threshold T, and the result of the comparison Ldif and Udif is less than the threshold T, the pixel distribution of the current block and the neighboring block have a high correlation, The best mode is predicted from the mode of the neighboring block already determined (encoded).

Accordingly, the candidate mode to be the optimal mode of the current block in the intra 4x4 mode is compressed into the L mode (LMODE) and the U mode (UMODE) as shown in FIG. In addition, when the L mode and the U mode are the same, there is one selectable candidate mode. In other words, by determining the similarity of the pixel distribution between blocks to determine a small number of candidate modes from the prediction mode of the intra 4x4 mode.

On the other hand, if the calculated Ldif or Udif is greater than the threshold value T, the directionality and correlation between the pixel values in the block is high, and as shown in FIG. 5, the intra 4 × is achieved using the square 4 × 4 mask of (a). Four directions of eight prediction directions (excluding the average value prediction mode) of four modes are obtained using Equation 3 below.

Here, a to p are pixel values at corresponding positions of the square 4x4 mask.

As a result, the candidate mode finally determined among the prediction modes of the intra 4x4 mode is one candidate mode in which the operation value of Equation 3 is minimum or is determined as a maximum of four candidate modes.

Accordingly, the present invention can reduce the calculation process by determining a few candidate modes in the intra 4x4 mode as described above and calculating the rate-distortion cost only for each of the determined candidate modes.

7A and 7B are signal flow diagrams illustrating an operation of a preferred embodiment of a fast mode for encoding an encoding mode of a macroblock according to the present invention. First, referring to FIG. 7A, first, a rate-distortion cost (RDCost_Skip) for a skip mode is calculated (S702), and a rate-distortion cost (RDCost_inter 16 × 16) of an inter 16x16 mode is calculated (S704). ), The rate-distortion cost (RDCost_Skip) for the calculated skip mode and the rate-distortion cost (RDCost_Inter16 × 16) of the inter 16x16 mode are compared (S706).

As a result of the comparison, when the rate-distortion cost (RDCost_Skip) for the skip mode is smaller than the rate-distortion cost (RDCost_Inter16 × 16) of the inter 16x16 mode, the determination of all modes is skipped, and the optimal mode is the skip mode. (S708), the optimum mode is determined as the skip mode (S710).

As a result of the comparison of step S706, the rate-distortion for inter 16 × 8 when the rate-distortion cost RDCost_Skip for the skip mode exceeds the rate-distortion cost RDCost_Inter16 × 16 of the inter 16 × 16 mode. The cost (RDCost_Inter16x8) and the rate-distortion cost (RDCost_Inter8x16) for the inter 8x16 mode are calculated (S712).

Then, it is determined whether the rate-distortion cost RDCost_Inter16 × 16 of the inter 16x16 mode is the minimum value MIN_RDCost (S714), and the rate-distortion cost RDCost_Inter16 × 16 of the inter 16x16 mode is the minimum value MIN_RDCost. In one case, the determination process of all the following inter modes is omitted (S716).

When the rate-distortion cost RDCost_Inter16 × 16 of the inter 16 × 16 mode is not the minimum value MIN_RDCost as a result of the determination in step S714, the rate for the calculated inter 16 × 8 as shown in FIG. 7B. -Whether the distortion cost RDCost_Inter16x8 and the rate-distortion cost RDCost_Inter8x16 for the inter 8x16 mode is the minimum value MIN_RDCost is determined (S718, S720).

As a result of the determination in step S718, when the rate-distortion cost RDCost_Inter16 × 8 for the inter 16 × 8 is the minimum value MIN_RDCost, the inter 16 × 16 mode, the inter 8 × 16 mode, and the sub macroblock 4 × The mode decision process for the eight modes is omitted (S722), and the optimum mode of the sub macroblock in the 8x8 mode is set (S726).

If the rate-distortion cost RDCost_Inter8 × 16 for the inter 8 × 16 mode is the minimum value MIN_RDCost as a result of the determination in step S720, the inter 16 × 16 mode, the inter 16 × 8 mode, and the sub macroblock 8 The mode decision process for the x4 mode is omitted (S724), and the optimum mode of the sub macroblock in the 8x8 mode is set (S726).

In addition, as a result of the determination of the steps S718 and S720, the rate-distortion cost RDCost_Inter16 × 8 for inter 16 × 8 and the rate-distortion cost RDCost_Inter8 × 16 for inter 8 × 16 mode are both minimum values MIN_RDCost. If not, the optimum mode of the sub macroblock in the 8x8 mode is set (S726).

Among the sub macroblocks, it is determined whether there are two sub macroblocks 8x8 set to an optimal mode and two submacroblocks 4x4 (S728).

As a result of the determination, when there are not two sub macroblocks 8 × 8 or two sub macroblocks 4 × 4, it is determined whether two or more sub macroblocks 8 × 8 set to an optimal mode are two or more (S730). ).

If there are two or more submacroblocks 8x8 set to the optimal mode in step S730, an intra 4x4 prediction mode determination process is omitted (S732). When the number of the sub macroblocks 8 × 8 set to the optimal mode is not more than two in step S730, it is determined whether the number of the submacroblocks 4 × 4 set to the optimal mode is two or more (S734), When there are two or more submacroblocks 4x4 set to the optimal mode, the process of determining an intra 16x16 prediction mode is omitted (S736).

In addition, when there are two sub macroblocks 8 × 8 set to the optimal mode in step S728 and two sub macroblocks 4 × 4 set to the optimal mode, the correlation coefficient between the 4 × 4 square mask and the block is used. An intra 4 × 4 mode determination algorithm is performed (S738).

The rate-distortion cost is calculated (S740), and an optimal mode is determined (S742).

The present invention has been described in detail with reference to exemplary embodiments, but those skilled in the art to which the present invention pertains can make various modifications without departing from the scope of the present invention. I will understand.

Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined by the claims below and equivalents thereof.

Claims (4)

Determining an early skip mode if the rate-distortion cost of the skip mode is less than the rate-distortion cost of the inter 16 × 16 mode;
When the rate-distortion cost of the skip mode is greater than the rate-distortion cost of the inter 16 × 16 mode, it is determined whether the rate-distortion cost of the inter 16 × 16 mode is the minimum value, and the inter 16 × 16 mode Omitting the determination of all inter modes when the rate-distortion cost of the is the minimum value;
Setting an optimal mode of a sub macroblock in an 8 × 8 mode according to whether inter 16 × 8 or inter 8 × 16 are minimum values when the rate-distortion cost of the inter 16 × 16 mode is not the minimum value ;
Determining whether there are two sub macroblocks 8x8 in which the optimum mode is set and two submacroblocks 4x4 in which the optimum mode is set;
As a result of the determination, when there are two sub macroblocks 8 × 8 in which the optimal mode is set, and two sub macroblocks 4 × 4 in which the optimum mode is set, intra 4 is determined using a 4 × 4 square mask and inter-block correlation. Performing an x4 mode decision algorithm, calculating a rate-distortion cost and determining an optimal mode; And
As a result of the determination, when there are not two sub macroblocks 8 × 8 in which the optimum mode is set or two sub macroblocks 4 × 4 in which the optimum mode is set, two or more sub macroblocks 8 × 8 or sub macros are set. Omitting the prediction mode determination process according to whether two or more blocks 4 × 4 are present, and calculating an optimal mode by calculating a rate-distortion cost; and including the encoding mode of the macroblock in the H.264 / AVC. Fast decision making method.
The method of claim 1, wherein the determining of the early skip mode comprises:
Calculating the rate-distortion cost of the skip mode and the rate-distortion cost of the inter 16 × 16 mode, respectively;
Comparing the calculated rate-distortion cost of the skip mode with the rate-distortion cost of the inter 16 × 16 mode; And
Determining an early skip mode when the rate-distortion cost of the skip mode is less than the rate-distortion cost of the inter 16 × 16 mode as a result of the comparison. Fast encoding method of the encoding mode.
2. The method of claim 1, wherein setting an optimal mode of the sub macroblock in the 8x8 mode;
Determining whether the rate-distortion cost of the inter 16 by 16 mode is a minimum value;
Omitting all inter-mode determination when the rate-distortion cost of the inter 16 by 16 mode is a minimum value;
If the rate-distortion cost of the inter 16 × 16 mode is not the minimum value, the rate-distortion cost of the inter 16 × 8 mode is the minimum value, or the rate-distortion cost of the inter 16 × 16 mode is the minimum value. Determining whether or not;
When the rate-distortion cost for the inter 16 × 8 mode is the minimum value, the mode decision process for the inter 16 × 16 mode, the inter 8 × 16 mode, and the sub macroblock 4 × 8 mode is omitted, and the sub within 8 × 8 mode is omitted. Setting an optimal mode of a macroblock;
When the rate-distortion cost for the inter 8 × 16 mode is a minimum value, the mode decision process for the inter 16 × 16 mode, the inter 16 × 8 mode, and the sub macroblock 8 × 4 mode is omitted, and the sub within 8 × 8 mode is omitted. Setting an optimal mode of a macroblock; And
Setting an optimal mode of a sub macroblock in an 8x8 mode when both the rate-distortion cost for the inter 16x8 mode and the rate-distortion cost for the inter 8x16 mode are not minimum values; Fast mode for encoding a macroblock in H.264 / AVC.
The method of claim 1, wherein a prediction mode determination process is omitted according to whether two or more sub macroblocks 8x8 or two or more submacroblocks 4x4 are used, and a rate-distortion cost is calculated to optimize Determining the mode;
Skipping an intra 4 × 4 prediction mode determination process when the sub macroblocks 8 × 8 are two or more, and determining an optimal mode by calculating a rate-distortion cost;
Skipping an intra 16 × 16 prediction mode determination process when the sub macroblocks 4 × 4 are two or more, and determining an optimal mode by calculating a rate-distortion cost; And
Determining an optimal mode by calculating a rate-distortion cost when the sub macroblocks 8 × 8 are not two or more and the sub macroblocks 4 × 4 are not two or more. Fast mode for encoding a macroblock in H.264 / AVC.

Images