KR20060021149A - Method for selecting macro block type - Google Patents

Abstract

The present invention relates to a method for determining a macroblock type for quickly determining a macroblock type when encoding a video and to encoding a video using the macroblock type. The present invention relates to a macroblock determination method for encoding a video. Calculating a variance of pixel values constituting the circuit; And determining an inter macroblock type according to the variance value of the macroblock.

According to the present invention, since the time required to determine the macroblock type and the amount of computation can be significantly reduced when encoding an image, the possibility of real-time implementation of an image encoder can be enhanced.

H.264, macroblocks, motion estimation

Description

Method for selecting macro block type}

1 is a diagram illustrating an example of macroblock partitioning for motion estimation and compensation in the conventional H.264 standard.

2 is a flowchart illustrating an example of a method for determining a macroblock type in a conventional H.264 encoder.

3 is a flowchart illustrating another example of a method for determining a macroblock type in a conventional H.264 encoder.

4 is a block diagram of an H.264 encoder in accordance with a preferred embodiment of the present invention.

5 is a block diagram schematically illustrating a configuration of a motion estimation unit of FIG. 4.

6 is a flowchart illustrating a method of determining a macroblock type of an H.264 encoder according to the present invention.

7 to 9 illustrate embodiments for explaining a method of selecting a candidate type macroblock in an H.264 encoder according to the present invention.

FIG. 10 is a diagram for describing a candidate type selection embodiment of a distribution calculator having a function of further selecting a candidate inter macroblock type; FIG.

<Description of Main Parts of Drawings>

100: motion estimation unit 101: P_16 x 16 type motion estimation unit

102: P_Skip type determination unit 103: variance calculation unit

104: P_16x16 type determination unit 105: candidate type selection unit

106: candidate type motion estimation unit 107: inter macroblock type determination unit

The present invention relates to a video compression system, and more particularly, to a method of determining a macroblock type of a video encoder for rapidly determining a macroblock type when encoding a video and encoding a video using the macroblock type.

H.264 is the next generation video compression standard jointly promoted by ITU-T (International Telecommunication Union), ISO (International Organization for Standardization) and IEC (International Electrotechnical Commission). MPEG-4 Part 10 or MPEG-4 AVC (Advanced) Video Codec). The H.264 provides a data compression capacity of three times or more than MPEG-2 and twice or more than MPEG-4.

In addition, in motion estimation and compensation process of existing MPEG-2, MPEG-4, and H.263, one reference frame is used for each of 16x16 or 8x8 macroblocks in forward and backward directions. In contrast, the H.264 encoder can use a plurality of reference frames for motion estimation and compensation, and supports motion estimation using seven motion estimation units and 1/4 pixel motion vector accuracy.

In the H.264 standard, one picture is divided into one slice of I, B, and P slices, and each slice is divided into a plurality of macroblocks. The macroblock is encoded in two ways, inter mode and intra mode.

First, an inter mode is a mode used in an interprediction encoding a difference between pixel values and motion vector information indicating positions of one or a plurality of blocks selected from a reference picture in order to encode a macroblock of a current picture. . Since the H.264 standard can have up to five reference pictures, the reference picture is searched in the frame memory in which the reference picture is stored to find a block to which the current macroblock refers. The reference picture stored in the frame memory may be a past picture or a future picture based on the current picture.

Next, the intra mode does not refer to a reference picture in order to encode a macroblock of a current picture, but uses a pixel value spatially adjacent to a macroblock to be encoded to obtain a prediction value for a macroblock to be encoded. After calculation, this mode is used in the intra prediction for encoding the difference between the predicted value and the pixel value.

1 is a diagram illustrating an example of macroblock partitioning for motion estimation and compensation in the conventional H.264 standard. One 16 × 16 macroblock can be divided into 16 × 16, 16 × 8, 8 × 16, or 8 × 8 blocks, with each 8 × 8 block being a smaller unit: 8 × 4, 4 × 8, It can be divided into 4 × 4 blocks. Each of the divided macroblocks is encoded in all encoding modes in inter-prediction and in all encoding modes in intra-prediction, and thereafter, the bit rate required for encoding the macroblock, and the original macroblock and the decoded macroblock One encoding mode is determined and encoded according to the degree of distortion.

2 is a flowchart illustrating an example of a method for determining a macroblock type in a conventional H.264 encoder according to the encoding method. Hereinafter, 'P_X × Y' refers to a macroblock type of size X × Y used in inter prediction, and 'I_X × Y' refers to a macroblock type of size X × Y used in intra prediction. .

First, it is checked whether the slice to which the macroblock to be encoded belongs is a P slice (S10). If the macroblock to be encoded is an I slice, the process proceeds directly to the intra encoding step (S13).

In the case of the P slice as a result of the determination in step S10, the motion estimating unit types 'P_16 × 16', 'P_16 × 8', 'P_8 × 16', 'P_8 × 8', and 'P_Skip' for each reference frame. Motion estimation is performed at step S11.

In operation S11, an optimal reference frame is determined for each type, and a motion vector and an optimal rate-distortion cost (RD cost) are calculated. The bit rate-distortion value of each type is compared based on the result of the calculation, and the type having the minimum bit rate-distortion value is determined as an optimal inter macroblock type, and the motion vector and the reference frame of the determined inter macroblock type have the final motion. It becomes a vector and a final reference frame (S12).

Next, the intra encoder encodes the macroblocks as 'I_4 × 4' and 'I_16 × 16', and then calculates bit rate-distortion values for each type (S13).

Based on the calculation result of step S13, the intra macroblock type determination unit compares the bit rate-distortion value of each intra macroblock type, and selects an intra macroblock type having the minimum bit rate-distortion value as an optimal intra macroblock type. Determined (S14).

The optimum macroblock type determination unit compares the optimal inter macroblock type determined in the above steps S12 and S14 with the optimal intra macroblock type to determine a type having a smaller bit rate-distortion value as the optimal macroblock type. In the case of an I slice, as described above, the motion estimation process S11 is omitted, and the optimal intra macroblock type is determined as the optimal macroblock type as it is.

However, the method of determining the macroblock type shown in FIG. 2 has an advantage of obtaining an optimal motion vector and an optimal macroblock type because motion estimation is performed for all estimation units. Real time coding is difficult because it requires a considerable amount of computation.

In order to overcome this disadvantage, another example of the macroblock type determination method in the H.264 encoder has been disclosed as shown in FIG. 3.

Referring to FIG. 3, another example of a method for determining a macroblock type in a conventional H.264 encoder divides motion estimation into a 'P_16 × 16' type and other types to perform motion estimation (S21 and S25). A step S22 of determining whether the P_Skip type is performed using the result of the motion estimation S21 for the x16 'type is included. This is based on the fact that many macroblocks to be coded are determined to be of the 'P_Skip' type, so as to filter out macroblocks to be previously determined to be of the 'P_Skip' type and to omit the motion estimation process from other types. That is, for a macroblock determined as a 'P_Skip' type, an encoding time is reduced by determining an optimal macroblock type as 'P_Skip' without a motion estimation in another type and a separate intra macroblock type determination process (S25 to S29). will be.

However, in the conventional macroblock determination method of FIG. 3, the process of determining the type of the macroblock that is not determined as the 'P_Skip' type still requires a large amount of computation and encoding time.

In addition, in a video to be encoded, there are more macroblocks determined by the 16 × 16 type than macroblocks determined by the other type. In particular, in the case of P slices, most of the macroblocks of the less complicated and uncomplicated parts are determined as 'P_Skip' or 'P_16 × 16' types. Although the above-mentioned 'P_Skip' type is previously screened out in FIG. 3 to avoid unnecessary motion estimation for other types, the macroblock to be determined as the 'P_16 × 16' type cannot be selected in advance.

The present invention has been made to solve the above problems, and provides a macroblock type determination method capable of quickly determining a macroblock type using variance of macroblocks without using all macroblock types for motion estimation. The purpose is to.

In addition, the present invention takes advantage of the fact that macroblocks that are temporally and spatially adjacent are likely to be encoded in the same type, and thus types of macroblocks to be encoded using information of macroblocks to be encoded and macroblock type information that are temporally and spatially adjacent to each other. It is an object of the present invention to provide a method for determining a macroblock type for estimating.

In order to solve the above problems, the present invention provides a method for determining a macroblock type, the method comprising: calculating a variance of pixel values constituting a macroblock; And determining an inter macroblock type according to the variance value of the macroblock.

In the determining step, when the variance value of the macroblock is less than or equal to a predetermined value, it is preferable to determine the inter macroblock type as a macroblock having a size of 16 × 16.

The determining may include determining an inter macroblock type by performing motion estimation on a predetermined candidate macroblock type when the variance value of the macroblock is greater than or equal to a predetermined value.

The predetermined candidate macroblock type is preferably selected using a 16 × 16 type motion vector.

The inter macroblock type is preferably determined to have a minimum bit rate distortion value by comparing the bit rate distortion value for the 16 × 16 macroblock type and the bit rate distortion value for the candidate macroblock type.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In particular, the preferred embodiment will be described based on the encoder according to the H.264 standard, but the idea of the present invention can be applied to other video encoders.

The H.264 encoder according to the present invention encodes video data. The video data is composed of a plurality of pictures arranged on the time axis, each picture is composed of a plurality of slices, and each slice is composed of a plurality of blocks. As shown in FIG. 1 described above, a block includes a macroblock, a subblock obtained by dividing or dividing a macroblock in a vertical or horizontal direction. When interprediction is performed using these various types of variable blocks, encoding is effectively performed according to characteristics and motions of an image.

4 is a block diagram of an H.264 encoder according to a preferred embodiment of the present invention. The H.264 encoder includes a motion estimator 100, a motion compensator 120, an intra prediction unit 122, a transform unit 124, a quantization unit 126, a reordering unit 128, and an entropy coding unit ( 130, an inverse quantizer 132, an inverse transformer 134, a filter 136, and a frame memory 138.

The motion estimation unit 100 performs a search for a prediction value of a macroblock to be currently encoded for interprediction in the reference picture. When the reference block is found in units of 1/2 or 1/4 pixels, the motion compensator 120 calculates these intermediate pixel values to determine the reference block data value. In this way, the interprediction is performed by the motion estimator 100 and the motion compensator 120.

In addition, the intra prediction unit 122 that finds the prediction value of the macroblock of the current picture within the current picture is performed by the intra prediction unit 122. Whether to perform inter-prediction or intra-prediction on the current macroblock is performed by calculating the bit rate-distortion value under all encoding modes, and determining the mode with the smallest value as the encoding mode of the block. Perform encoding on

As described above, if inter prediction or intra prediction is performed and prediction data to be referred to by the macroblock of the current frame is found, the transform unit 124 performs subtraction by subtracting the macroblock from the macroblock of the current picture and then performing quantization ( 126). The subtraction of the motion estimation reference block from the macroblock of the current frame is called residual. The residual value is encoded to reduce the amount of data during encoding. The quantized residual value passes through the reordering unit 128 for encoding in the entropy coding unit 130.

Meanwhile, in order to obtain a reference picture to be used for interprediction, the current picture is reconstructed through the inverse quantization unit 132 and the inverse transform unit 134. The reconstructed current picture is stored in frame memory and used to perform interprediction on the next picture. When the reconstructed picture passes the filter 136, it becomes a picture with some encoding error in the original picture.

FIG. 5 is a block diagram schematically illustrating a configuration of the motion estimation unit 100 of FIG. 4.

The motion estimator 100 includes a P_16 × 16 type motion estimator 100, a P_Skip type determiner 102, a variance calculator 103, a P_16 × 16 type determiner 104, and a candidate type selector 105. ), A candidate type motion estimation unit 106, and an inter macroblock type determination unit 107. As described above, in the following, 'P_X × Y' refers to a macroblock type of size X × Y used in inter prediction, and 'I_X × Y' refers to a macroblock of size X × Y used in intra prediction. It means type.

In the case of a macroblock belonging to a P slice, motion estimation is first performed by using a 16 × 16 type macroblock in the P_16 × 16 type motion estimation unit 101.

The P_Skip type determiner 102 determines whether the macroblock to be encoded is a P_Skip type by using the result of the motion estimation of the P_16 × 16 type motion estimator 101.

The skip mode is a mode for transmitting or storing only encoding mode information of a block to be encoded. In the background part of the image, a given block of the current picture is likely to have the same pixel value as the corresponding block of the reference picture, and in this case, the transmission or storage of separate coded data such as a residual signal or motion vector information This is because only encoding mode information needs to be transmitted or stored without requiring. In order to determine whether a macroblock to be encoded is a P_Skip type, the P_Skip type determiner 102 may determine whether the following four conditions are satisfied.

i) the size of the block used for interprediction is 16x16, ii) the reference picture used to encode the current block is the one closest in time to the current picture among the reference pictures stored in the frame memory, and iii) the motion vector is ( 0, 0) or the same as a predicted motion vector (PMV), and iV) a value after converting and quantizing a residual value that is a difference between pixel values of a current block and a block of a reference picture is 0.

Alternatively, all three cases are satisfied. i) the size of the block used for interprediction is 8x8, and ii) the reference picture and the motion vector used for encoding the current block are estimated from the reference picture and motion vector information of the spatial or temporal neighboring block of the current block. Iii) The value after converting and quantizing a residual value, which is a difference between pixel values of a block of a current block and a reference picture, is 0.

If all of the above skip conditions are satisfied, the P_Skip type determination unit 102 determines that the macroblock is a P_Skip type and ends without checking other macroblock types. However, the present invention is not limited to the skip condition, and a method of determining whether various P_Skip types are used may be used.

The variance calculator 103 calculates a variance for macroblocks not determined as P_Skip type by the P_Skip type determiner 102. That is, the variance of the macroblock is calculated by dividing the difference between the average of each pixel value constituting the macroblock to be encoded and the average pixel value constituting the macroblock by dividing by the number of pixels of the macroblock. Specific formulas will be described later.

The P_16x16 type determiner 104 is a P_16x16 type motion vector and reference frame information estimated by the P_16x16 type motion estimator 101, and a macroblock calculated by the variance calculator 103. It is determined whether the current macroblock is determined to be a P_16x16 type macroblock using the variance of.

When the P_16 × 16 type determination unit 104 determines the P_16 × 16 type, the optimal inter macroblock type is determined as 'P_16 × 16', and the optimal intra macro determined by the intra prediction unit 122 is determined. Compared to the block type, having a smaller bit rate-distortion value determines the optimal macroblock type.

The candidate type selector 105 may select 'P_16 × 16', 'P_16 × 8', 'P_8 × 16', for macroblocks not determined as P_16 × 16 type by the P_16 × 16 type determiner 104. An inter macroblock type having a high probability of being determined among macroblocks of the 'P_8 × 8' type is predicted and selected. At this time, the inter macroblock type is predicted and selected using the 'P_16 × 16' type motion vector obtained by the P_16 × 16 type motion estimation unit 101.

The candidate type motion estimation unit 106 performs motion estimation on the remaining types except for the 'P_16 × 16' type among the candidate types selected by the candidate type selection unit 105 and calculates a bit rate-distortion value. In this case, the reference frame determined by the P_16x16 type determination unit 104 is used only as the reference frame.

The inter macroblock type determiner 107 compares the candidate types with bit rate-distortion values in the P_16x16 type and determines a type having a minimum bit rate-distortion value as an optimal inter macroblock type.

The optimal inter macroblock type determined by the inter macroblock type determiner 107 is again compared with the optimal intra macroblock type determined by the intra prediction performer 122 to have a smaller bit rate-distortion value. Determined by macroblock type.

Hereinafter, a method of determining a macroblock type of an H.264 encoder according to a preferred embodiment of the present invention will be described in detail.

6 is a flowchart illustrating a method of determining a macroblock type of an H.264 encoder according to the present invention.

In the method of determining the macroblock type of the H.264 encoder according to the present invention, the H.264-based video encoding process is performed by pre-filtering not only the 'P_Skip' type but also the 'P_16 × 16' type inter macroblock. Is characterized by reducing the time spent in the complex motion estimation process, which accounts for most of the total processing time.

Specifically, first, it is determined whether the macroblock to be currently encoded belongs to the P slice (S100).

As a result of the determination of the step, in the case of the macroblock belonging to the P slice, the P_16x16 type motion estimator 100 estimates the motion by using the 16x16 type macroblock in all reference frames, and then uses the 16x16 type. A motion vector, an optimal bit rate-distortion value, and an optimal reference frame in the macroblock of are obtained (S102).

Using the result of the step S102, the P_Skip type determination unit 102 determines whether the skip condition as described above is satisfied, and determines whether the macroblock to be encoded is a P_Skip type (S104).

As a result of the determination in step S104, if it is determined that the macroblock to be encoded is a P_Skip type, the optimal macroblock type is determined as P_Skip, and the process ends without performing motion estimation in another macroblock type (S108).

If it is determined in step S104 that it is not the P_Skip type, the motion estimation process is performed using macroblock types of 'P_16 × 16', 'P_16 × 8', 'P_8 × 16', and 'P_8 × 8'. In the present invention, the macroblock to be encoded in the 'P_16 × 16' type may be determined in advance by calculating the variance of the macroblock to be encoded among the macroblock types.

Specifically, if it is determined that the step S104 is not of the P_Skip type, the variance calculator 103 calculates a variance of the macroblock to be encoded using Equation 1 below (S110). ).

In Equation 1, k denotes a number of macroblocks, and p (i, j) represents a pixel value spaced apart from the leftmost pixel of the macroblock by i in the horizontal direction and j in the vertical direction. In addition, MEAN (k) is expressed by the following equation 2 as an average of all pixel values constituting a macroblock to be encoded.

Using the variance calculated in step S110, the P_16 × 16 type determination unit 104 determines whether a macroblock to be encoded is determined to be an P_16 × 16 type inter macroblock (S112). Specifically, it is determined in advance whether it is an P_16x16 type inter macroblock using the following equation (3).

In Equation 3, k is a macroblock number, VAR (k) is variance, J _{P _ 16 × 16} (k) is a bit rate-distortion value at P_16 × 16, and mb_type _inter (k) is an finally determined inter macro. It means the block type. In addition, Th _VAR and Th _{RD cost} are user- _definable reference values, which can be flexibly changed according to the characteristics of the image. The greater the value of the Th _VAR and Th _{RD cost,} the higher the probability that the inter macroblock probability is determined to be 'P_16 × 16'.

As shown in Equations 1 to 3, in the macroblock type determination method of the H.264 encoder according to the present invention, the dispersion is large when the macroblock to be encoded is complex, and the dispersion is small when the relatively flat region is formed. In other words, the macroblock having a variance below a certain reference value takes advantage of the fact that it is likely to be of type P16 × 16.

At this time, in step S112, an optimal reference frame of the P_16x16 type obtained by the P_16x16 type motion estimation unit 101 is determined as the final reference frame, and the only reference in the motion estimation in the other macroblock type is determined. Used as a frame.

Next, it is determined whether the type of the macroblock to be encoded is P_16 × 16 type under the condition shown in Equation 2 (S114). In the case of the P_16 × 16 type, the optimal inter macroblock type is determined as P_16 × 16. The intra coding step S126 is reached.

If it is not determined that the P_16 × 16 type is determined, the candidate type selector 105 has a high probability of being determined among 'P_16 × 16', 'P_16 × 8', 'P_8 × 16', and 'P_8 × 8'. The candidate of the macroblock type is predictively selected (S116). At this time, the P_16x16 type motion vector obtained by the P_16x16 type motion estimation unit 101 is used.

7 to 9 are diagrams for describing an exemplary embodiment of selecting a candidate macroblock type.

 FIG. 7 illustrates a case in which a macroblock to be currently encoded is predicted from four macroblocks of a reference frame. As shown, when the current macroblock is predicted through four macroblocks of a reference frame indicated by the P_16 × 16 type motion vector, the number of pixels required to predict the current macroblock in each of the four macroblocks is calculated. . In the case of FIG. 7, the number of pixels used for prediction of the current macroblock is 25 for macroblock A, 55 for B, 55 for C, and 121 for D. Among macroblocks A, B, C, and D, the macroblock of the reference frame having the largest number of pixels used for prediction of the current macroblock among the remaining macroblocks is selected as a candidate type. If there are two or more macroblocks equal in the number of pixels, priority is given to the macroblock closest to the euclidian distance from the macroblock to be encoded. If both the number of pixels and the distance from the macroblock to be encoded are the same, priority is given arbitrarily.

In addition, when the type selected as the candidate type is only P_16x16 type, the inter macroblock type is determined as P_16x16 type without further motion estimation process. If macroblocks A, B, C, and D of the reference frame are all encoded in the intra mode or P_Skip type, P_16 × 8 and P_8 which are the remaining types except P_16 × 16 and P_Skip types are candidate types. Select both x16 and P_8x8.

FIG. 8 is a diagram for describing a candidate type selection method when a current macroblock is predicted through two macroblocks C and D of a reference frame as another embodiment of selecting a candidate type.

 As shown, a type of macroblock having a large number of pixels used for prediction of a macroblock to be currently encoded is selected as a candidate type from two macroblocks C and D of a reference frame. In the case of FIG. 8, unless the D macroblock is encoded in the intra mode or the P_Skip type, the macroblock type P_16 × 16 which is the type of the D macroblock is selected as the candidate type.

If the number of pixels used for prediction is the same, as in the case of FIG. 7, the type of the macroblock of the reference frame close to the current macroblock to be encoded is selected as the candidate type.

In the case where the selected type is P_16 × 16, the inter macroblock type is determined as P_16 × 16 without a separate motion estimation process, and both macroblocks C and D of the reference frame are encoded in the intra mode or the P_Skip mode. If so, macroblocks of type 'P_16 × 8', 'P_8 × 16', and 'P_8 × 8' are selected as candidate types. If only macroblock C of macroblocks C and D is encoded in the inter mode, the candidate type is determined as the type of macroblock C.

9 is a diagram illustrating a method of selecting a candidate type when a macroblock to be currently encoded is predicted from one specific macroblock in a reference frame.

In the case of FIG. 9, the macroblock type of the reference frame is selected as a candidate type as it is. However, when the selected candidate type is encoded in the intra mode or the P_Skip type, the candidate types are 'P_16 × 8', 'P_8 × 16', and 'P_8 × 8'.

If the candidate type is selected in step S116, it is determined whether there are macroblocks other than the P_16x16 type among the candidate types (S118), and if only P_16x16 type exists as the candidate type, the inter-macroblock type is selected. P_16 × 16 is determined, and the process of determining the optimal macroblock type is completed (S120).

As a result of the determination in step S118, when there are candidate types other than the P_16x16 type, the candidate type motion estimation unit 106 performs motion estimation on candidate types other than the P_16x16 type (S122). .

The reason why the motion estimation is not performed for the P_16x16 type is that the motion estimation for the P_16x16 type macroblock has already been performed in the P_16x16 type motion estimation unit 101. As described above, only the reference frame determined by the P_16x16 type determination unit 104 is used as the reference frame.

If the P_8x8 type is present as a candidate type in the motion estimation step S122, motion estimation is performed up to a sub-macroblock unit of 8x8, 8x4, 4x8, and 4x4 sizes. Through the motion estimation step S122, a motion vector and an optimal bit rate-distortion value are obtained for each candidate type.

Next, the inter macroblock type determination unit 107 determines the bit rate-distortion value for the P_16 × 16 type obtained by the P_16 × 16 type motion estimation unit 101 and the bit rate-distortion for each candidate type obtained in the step S122. By comparing the values, the macroblock type having the minimum bit rate-distortion value is determined as the optimal inter macroblock type (S128).

In the state where the optimal inter macroblock type is determined through the above process, the intra prediction processor 122 encodes the current macroblock into 'I_4 × 4' and 'I_16 × 16' types, and according to each type. Calculate the bit rate-distortion value. The intra macroblock type having the minimum bit rate-distortion value is determined as an optimal intra macroblock type by comparing bit rate-distortion values of respective intra macroblock types based on the calculation result.

Finally, by comparing the optimal inter macroblock type determined in the step S124 of determining the inter macroblock type with the optimal intra macroblock type determined in the step S128, a type having a smaller bit rate-distortion value is an optimal macroblock. Determine the type (S130). However, in the above process, the macroblock belongs to the P slice, and in the case of the I slice, the optimal intra macroblock type is determined as the optimal macroblock type.

Meanwhile, as another embodiment of the present invention, the variance calculator 103 of FIG. 5 may be one of 'P_16 × 8', 'P_8 × 16', and 'P_8 × 8' types separately from the candidate type selector 105. It may have a function of additionally selecting a candidate inter macroblock type. In this case, the candidate type motion estimation unit 106 performs motion estimation on the candidate type determined by the variance calculator 103 and the candidate type selection unit 105 having an extended function.

FIG. 10 is a diagram illustrating a candidate type selection embodiment of the variance calculator 103 having a function of further selecting a candidate inter macroblock type. In FIG. 10, A to H represent partitions of macroblocks, A and B are blocks of 16 × 8 size, C and D are blocks of 8 × 16 size, and E to H are blocks of 8 × 8 size. It is a block.

The variance VAR (k) expressed in Equation 3 is the sum of the squares of the difference between MEAN (k), which is the overall macroblock, and the partition pixel value for each macroblock partition of A to H, S _A , S _{B ,} S _{C, and} S _{D. ,} S _{E ,} S _{F ,} S _{G ,} S _H It can be expressed by using Equation 4 below.

In Equation 4, S _A , S _{B ,} S _C, S _D, S _{E ,} S _{F ,} S _{G ,} S _H Is used for candidate type determination, and the determination method is as follows.

이면, 후보 타입으로 'P_16×8'을 결정하고,i)

If it is, the candidate type 'P_16 × 8' is determined,

이고,

이면, 후보 타입으로 'P_8×16'을 결정하고,ii)

ego,

If it is, the candidate type 'P_8 × 16' is determined,

이고,

이고,iii)

ego,

ego,

이면, 후보 타입으로 'P_8×8'을 결정한다. 여기서, Th_{P _16 × 8}, Th_{P _8 × 16}, Th_{P _8 × 8}은 부호화하고자 하는 영상의 특성에 따라 임의로 설정할 수 있는 값이다.

In this case, 'P_8 × 8' is determined as the candidate type. Here, Th _{P _16 × 8, × 16 _8 P} Th, Th _{P _8 × 8} is a value that can be set arbitrarily according to the characteristics of the image to be coded.

In the motion estimation process of the macroblock to be encoded through the candidate type selection process as described above, the candidate macroblock type used in the motion estimation process can be quickly determined.

So far I looked at the center of the preferred embodiment for the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

According to the present invention as described above, it is possible to significantly reduce the amount of time and calculation amount for determining the macroblock type in the video encoder, it is possible to increase the possibility of real-time implementation of the video encoder.

Claims (5)

In the macroblock type determination method for encoding an image, Calculating a variance of pixel values constituting the macroblock; And And determining an inter macroblock type according to the variance value of the macroblock. The method of claim 1, wherein the determining step, And if the variance value of the macroblock is less than or equal to a predetermined value, determining an inter macroblock type as a macroblock having a size of 16 × 16. The method of claim 1, wherein the determining step, And determining the inter macroblock type by performing motion estimation on a predetermined candidate macroblock type when the variance value of the macroblock is greater than or equal to a predetermined value. The method of claim 3, wherein The predetermined candidate macroblock type is selected using a 16x16 type motion vector. The method of claim 3, wherein And wherein the inter macroblock type is determined as a type having a minimum bit rate distortion value by comparing the bit rate distortion value for the 16 × 16 macroblock type and the bit rate distortion value for the candidate macroblock type.

Images