KR20040093253A - 16x16 intra luma prediction mode determining method and apparatus - Google Patents

Abstract

PURPOSE: A method and an apparatus for determining a 16x16 intra luminance prediction mode are provided to apply only certain modes without applying 4 modes, perform prediction, and determine an optimal 16x16 intra luminance prediction mode. CONSTITUTION: A prediction mode information analyzing unit(1310) receives mode information of sub-blocks of a block to be predicted, analyzes received mode information, and outputs a prediction mode of the block to be predicted. A 16x16 block prediction performing unit(1320) applies the prediction mode received from the prediction mode information analyzing unit, and performs prediction with respect to the block to be predicted. A residue calculating unit(1330) calculates residue with respect to a prediction sample outputted from the 16x16 block prediction performing unit. A residue comparing unit(1340) compares residues calculated in the residue calculating unit, selects a mode with the minimum residue, and outputs the selected mode.

Description

16 × 16 intra luma prediction mode determining method and apparatus

The present invention relates to encoding / decoding of video data, and more particularly, to a method and apparatus for determining a prediction mode used for video encoding / decoding.

With the advent of digital TV and DVD-Video, broadcast and home entertainment services have changed significantly. Various application services provided by such broadcasting services and home entertainment services are made possible by standardization of video compression technology. Among the standards for video compression technology, MPEG4-visual enabled Internet-based video application services, and ITU-T's H.263 standard is widely used in videoconferencing systems.

MPEG4-visual and H.263 are standards based on video compression. Moving Picture Experts Group (MPEG) and Video Coding Experts Group (VCEG) involved in these standards are higher-quality and lower-bitrate than MPEG4 and H.263. The next step was to establish a standard for the next level of technology that would allow more compression while supporting streaming video.

After the completion of the H.263 standard, ITU-T's Video Coding Experts Group (VCEG) worked on the H.263 version 2 standard, which added additional functionality to H.263, and a new standard for low-bit rate video communications. Started research on. The new standard for low bit rate video communication is named ITU-T H.26L. In addition, ISO's Video Experts Group (MPEG) recognizes the importance of this new standard and has formed a joint video team (JVT) with ITU-T's Video Coding Experts Group (VCEG) to build the H.26L model. The study began to set international standards. The standard thus defined is ISO MPEG4 Part 10, that is, ITU-T H.264. The official name of this standard is MPEG-4 Advanced Video Coding (AVC), but is better known as H.264.

1 is a block diagram of an H.264 encoder.

The H.264 encoder largely includes a prediction block 110, a transform and quantization block 120, and an entropy coding block 130.

The prediction unit 110 performs inter prediction and intra prediction. Interprediction refers to performing block prediction of a current picture by using a reference picture stored in a buffer by performing decoding and deblocking filtering. That is, prediction is performed by using information between pictures. To this end, a motion estimation block 111 and a motion compensation block 112 are provided. Intra prediction refers to performing prediction using pixel data of a block adjacent to a block to be predicted in a picture that has already been decoded.

The transform and quantization unit 120 transforms, predicts, and compresses the prediction samples obtained by performing the prediction in the prediction unit 110. The entropy coding unit 130 performs encoding on the quantized video data in a predetermined manner to form an H.264 bitstream.

2 is a block diagram of an H.264 decoder.

The H.264 decoder receives the bitstream encoded by the H.264 encoder, performs entropy decoding, performs inverse quantization and inverse transformation, and performs decoding using reference picture information that has been subjected to motion compensation or intra prediction.

FIG. 3 is a diagram illustrating a luma block P to be predicted and an adjacent block to be used for prediction.

If a block or macroblock is encoded in intra mode, block P 310 to be predicted is predicted using the decoded blocks A to L adjacent to the block. The prediction is performed not only on the luma block but also on the chromaticity (Cb, Cr) blocks. In the present invention, prediction of the luminance block will be described. The luminance prediction block P 310 is a 16x16 block composed of several 4x4 blocks. Here, a to p are 4x4 blocks to be predicted, and A, B, C, D and I, J, K, and L are blocks used for prediction.

There are two types of intra prediction, 4x4 prediction and 16x16 prediction, depending on the size of the block to be predicted. According to the prediction direction, there are nine modes for 4x4 prediction and four modes for 16x16 prediction. The 4x4 prediction mode is divided according to the prediction direction when a prediction sample is obtained using pixel values of blocks A, B, C, D and I, J, K, and L adjacent to the 4x4 block to be predicted.

4 is a diagram illustrating the types of 4x4 intra luma prediction modes.

Referring to FIG. 4, it can be seen that the 4x4 intra luminance prediction mode includes a vertical mode, a horizontal mode, a DC mode, a diagonal_down_left mode, a diagonal_down_right mode, a vertical_right mode, a horizontal_down mode, a vertical_up mode, and a horizontal_up mode. The direction in which the prediction is performed in each prediction mode will be described with reference to FIG. 5. The prediction of the block in each mode is described with reference to FIGS. 6A to 6I.

FIG. 5 is a diagram illustrating nine prediction directions in 4x4 intra luminance prediction of H.264.

Referring to FIG. 5, it can be seen that prediction of a block is performed in a vertical direction, a horizontal direction, or a diagonal direction to correspond to each mode name.

6A through 6I are diagrams for describing performing prediction according to a 4x4 intra luminance prediction mode.

For example, in mode 0 (vertical mode), the 4x4 blocks a, e, i, m are predicted using the pixel values of the A block, and b, f, j, n are predicted using the pixel values of the B block. , c, g, k, o are predicted using the pixel values of the C block, and d, h, l, p are predicted using the pixel values of the D block. Prediction methods in other modes are as detailed in the H.264 standard.

7 is a diagram illustrating the types of 16x16 intra luma prediction modes.

Referring to FIG. 7, it can be seen that there are four types of 16x16 intra luminance prediction modes: vertical mode, horizontal mode, DC mode, and plane mode. The direction in which prediction is performed in each prediction mode is described with reference to FIGS. 8A to 8D.

8A to 8D are diagrams for describing performing prediction according to a 16x16 intra luminance prediction mode.

Each block in the 16x16 luma macroblock is predicted using the pixel value of block H and the pixel value of block V. In other words, in mode 0 (vertical mode) each block in the 16x16 luma macroblock is predicted using the pixel value of block H, and in mode 1 (horizontal mode) each block in the 16x16 luma macroblock is the pixel value of block V. Predicted by using each block in a 16x16 luma macroblock in mode 2 (DC mode) is estimated by averaging the pixel values in block H and pixel values in block V, and in each of the 16x16 luma macroblocks in mode 3 (plane mode) The block is predicted using both the pixel value of block H and the pixel value of block V. Detailed prediction methods are described in detail in the MPEG-4 AVC Standard.

In H.264 encoding, prediction is performed by selecting a best mode among 16x16 intra luminance prediction modes. The compression efficiency of H.264 varies according to which mode is selected to perform the 16x16 block luminance prediction.To select the best mode, the prediction of the block is performed by applying all modes and a predetermined cost function Calculate the cost using, and select the mode with the lowest cost. Therefore, since all four modes must be applied to the block to be predicted to perform the prediction and the cost is calculated, the encoder becomes very complicated.

The technical problem to be achieved by the present invention is to determine the 16x16 intra luminance prediction mode, without applying all four modes of 16x16 intra luminance prediction to perform the prediction by applying only a few modes of the prediction 16x16 The present invention provides a method and apparatus for determining a 16x16 intra luminance prediction mode capable of determining an intra luminance prediction mode.

1 is a block diagram of an H.264 encoder.

2 is a block diagram of an H.264 decoder.

3 is a diagram illustrating a luminance block P to be predicted and an adjacent block used for prediction.

4 is a diagram illustrating the types of 4x4 intra luma prediction modes.

FIG. 5 is a diagram illustrating nine prediction directions in 4x4 intra luminance prediction of H.264.

6A through 6I are diagrams for describing performing prediction according to a 4x4 intra luminance prediction mode.

7 is a diagram illustrating the types of 16x16 intra luma prediction modes.

8A to 8I are diagrams for describing performing prediction according to a 16x16 intra luminance prediction mode.

9 is a diagram illustrating dividing a 4x4 intra luminance prediction mode into three groups.

10 is a flowchart of a method for determining a 16x16 intra luminance prediction mode of the present invention.

11 is a view for explaining another 16x16 intra luminance prediction mode determination method of the present invention.

12 is a view for explaining another 16x16 intra luminance prediction mode determination method of the present invention.

13 is a block diagram of the prediction mode determination apparatus of the present invention.

According to an aspect of the present invention, there is provided a method of determining a prediction mode, comprising: (a) receiving and analyzing prediction mode information of sub-blocks of a block to be predicted; And (b) selecting a prediction mode of the block to be predicted according to the analysis result.

In order to achieve the above object, the method of determining a prediction mode according to the present invention includes: (a) a prediction mode of a 4x4 luma block located in a first group and a first group to which a prediction mode of a 4x4 luma block located in a first column of a 16x16 luma block belongs; Counting the number of prediction modes of the third group to which the second group belongs and the prediction mode of other 4x4 luma blocks; And (b) selecting and outputting a prediction mode corresponding to the group having the largest number of prediction modes.

In order to achieve the above object, the prediction mode determination method according to the present invention includes (a) a higher macroblock sample and a left macroblock sample used for 16x16 intra luminance prediction, and an upper block sample and a left block used for 4x4 intra luminance prediction. Comparing pixel values of the sample in a row unit or a column unit; And (b) selecting and outputting a 16x16 prediction mode according to whether the comparison result is the same or within a predetermined difference.

According to an aspect of the present invention, there is provided a prediction mode determining apparatus comprising: a prediction mode information analysis unit configured to receive mode information of a subblock of a block to be predicted and to output the prediction mode of the block to be predicted; A block prediction performing unit for performing prediction on the block to be predicted by applying the prediction mode received from the subblock prediction mode information analyzing unit; A prediction error calculator configured to calculate a prediction error of the prediction sample output from the block prediction execution unit; And a prediction error comparison unit comparing the prediction error calculated by the prediction error calculator to select and output a mode having the smallest prediction error.

In order to achieve the above object, the present invention provides a computer-readable recording medium recording a program for executing the method on a computer.

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

9 is a diagram illustrating dividing a 4x4 intra luminance prediction mode into three groups.

The 4x4 intra luminance prediction mode is divided into three groups with similar prediction directions. Considering the similarity between the modes having similar prediction directions, the nine modes in 4x4 intra luminance prediction are divided into three groups and the result is used for 16x16 intra luminance prediction.

In other words, mode 7 and mode 5, including mode 0 (vertical mode) and adjacent directions, are grouped into one group 1 901, and mode 8 and mode 6, including mode 1 (horizontal mode) and adjacent directions. And group 2 into the other group 2 (902), and group 3 (mode 3 and mode 4) including the mode 2 (DC mode) and the remaining diagonal modes into one group 3 (903). In addition, it is determined whether the luminance prediction mode of the 4x4 block belonging to which group occurs the most. According to the result, the 16x16 intra luminance prediction mode is determined using various methods described below. Hereinafter, various methods of determining a 16x16 intra luminance prediction mode will be described in detail with reference to FIGS. 10 to 12.

10 is a flowchart of a method for determining a 16x16 intra luminance prediction mode of the present invention.

First, the number of 4x4 intra luminance prediction modes belonging to each group described with reference to FIG. 9 is counted (S1010). Then, one group or two groups are selected using coefficient values of each group and a predetermined reference value A. In other words, it is determined whether a difference between the coefficient value of one group and the coefficient values of other groups is greater than or equal to a predetermined reference value (S1020). The prediction mode is selected and output (S1030). Otherwise, if the difference between the coefficient values of each group does not exceed a predetermined reference value, all 16x16 intra luminance prediction modes corresponding to the group having similar coefficient values are selected and output (S1040). The predetermined reference value may be set differently according to the prediction performance and the complexity of the prediction method. If the predetermined reference value A is set large, the prediction performance is improved, but the prediction method is complicated. In one embodiment, when the predetermined reference value A is set to 0, the 16x16 intra luminance prediction mode can be selected more simply.

In step S1030 and step S1040, if the group 1 is selected, 16x16 intra luminance mode 0 is selected and outputted; if the group 2 is selected, 16x16 intra luminance prediction mode 1 is selected and outputted; if the group 3 is selected, 16x16 intra luminance prediction mode 2 and 16x16 intra luminance prediction mode 3 are selected and output. Therefore, the optimal mode can be determined even if the prediction is performed by applying the maximum of two modes without performing the prediction by applying all the 16x16 intra luminance prediction modes.

11 is a view for explaining another 16x16 intra luminance prediction mode determination method of the present invention.

It is checked to which group the intra luminance prediction mode of each 4x4 luma block belongs. That is, if the prediction modes of the 4x4 luma blocks located in the first column 1110 belong to group 1, the prediction modes of the 4x4 luma blocks located in the first row 1120 do not all belong to group 2, and others (1130). 16x16 intra luminance prediction mode 0 (vertical mode) is selected when the prediction mode of the 4x4 luma block located in the group 1 is greater than or equal to a predetermined number.

If the prediction modes of the 4x4 luma blocks located in the first column 1110 do not all belong to Group 1, the prediction modes of the 4x4 luma blocks located in the first row 1120 all belong to Group 2, and the 4x4 located in the other 1130. If the prediction mode of the luminance block belongs to Group 2 or more by a predetermined number of times, 16x16 intra luminance mode 1 (horizontal mode) is selected.

If the DC value of the 4x4 luminance block of the predetermined number or more located in the first column 1110 and the first row 1120 is the same as the DC value of the 16x16 macroblock, the 16x16 intra luminance mode 2 (DC mode) is determined. Choose.

In the other cases, 16x16 intra luminance mode 3 (plane mode) is selected.

In the above example, the predetermined number of settings is not specified. Increasing the value improves the prediction performance, but makes the prediction device very complicated. In consideration of the prediction performance and the complexity of the prediction apparatus, the complexity of the prediction apparatus can be greatly reduced by setting all blocks to be compared.

12 is a view for explaining another 16x16 intra luminance prediction mode determination method of the present invention.

16 4x4 blocks in the macroblock to be predicted are assigned numbers in raster scan order from 1 to 16. The samples used for 16 × 16 intra luminance prediction are referred to as upper macroblock sample 1210 and left macroblock sample 1220. The samples used for 4x4 intra luminance prediction are referred to as an upper block sample 1230 and a left block sample 1240. In FIG. 12, an example of an upper block sample and a left block sample is shown, and there are upper block samples and left block samples corresponding to 4 × 4 blocks.

If the 4x4 intra luminance prediction modes of the 4x4 blocks 1250 with the numbers {1, 2, 3, 4, 5, 6, 7, 9, 10, 13} all belong to group 1, 16x16 intra luminance prediction mode 0 And mode 3 are selected. The selection of mode 0 and mode 3 is as follows. Compares the upper block samples located in the same column with each upper macro block sample 1210, and if the pixel values are all the same or a difference less than or equal to the predetermined setting is less than the predetermined number of 4x4 block settings, 16x16. Prediction is performed by applying intra luminance prediction mode 0. Otherwise, when a difference occurs in which the pixel value of the 4x4 block exceeding the predetermined 4x4 block setting number exceeds the predetermined setting value, prediction is performed by applying the 16x16 intra luminance prediction mode 3. In the above-described example, a method of comparing the upper macroblock sample and the upper block samples may be performed by comparing each pixel value or by averaging all pixel values and then comparing the average value.

If the 4x4 intra luminance prediction modes of the 4x4 blocks 1250 with the numbers {1, 2, 3, 4, 5, 6, 7, 9, 10, 13} all belong to group 2, 16x16 intra luminance prediction mode 1 And mode 3 are selected. At this time, the left block samples located in the same row are compared with each left macro block sample, and when the pixel values are all the same or a difference less than or equal to the predetermined setting is less than the predetermined number of 4x4 block settings, 16x16 intra luminance is used. Prediction is performed by applying prediction mode 1. Otherwise, when a difference occurs in which the pixel value of the 4x4 block exceeding the predetermined 4x4 block setting number exceeds the predetermined setting value, prediction is performed by applying the 16x16 intra luminance prediction mode 3. In the above-described example, a method of comparing the left macroblock sample and the left block samples may be performed by comparing each pixel value or by averaging all pixel values and then comparing the average value.

In the above example, the predetermined number of 4x4 block settings and the predetermined setting value may be differently set according to the prediction performance and the complexity of the prediction method. If the predetermined number of settings and the predetermined value are set large, the prediction performance is improved, but the prediction method is complicated.

The DC value of one 16x16 macroblock is compared with the DC values of 16 4x4 blocks. If the DC value of 16 4x4 blocks is less than the predetermined value from the DC value of the 16x16 macroblock, it is 16x16 intra. Prediction is performed by applying luminance prediction mode 2 (DC mode). The predetermined set value is set in consideration of the performance of the prediction apparatus and the complexity of the prediction method. When comparing DC values, a method of comparing a difference of average or a sum of difference may be used.

The 16x16 intra luminance prediction modes determined according to the prediction modes of each 4x4 block are applied to perform prediction on the 16x16 macroblock, and the prediction cost is calculated. The prediction cost is a residual value obtained by subtracting the prediction value from the original pixel value of the sample to be predicted, that is, a value calculated by SSD, SAD or SATD based on a prediction error. These cost values are compared with each other to determine the 16x16 intra luminance prediction mode having the optimal cost value, and the cost value at that time is compared with the cost value when the prediction is performed by applying the 4x4 intra luminance prediction mode that has already been calculated.

The calculation of the cost can be performed by several methods. Cost functions used include sum of absolute value (SAD), sum of absolute transformed difference (SATD), sum of squared difference (SSD), mean of absolute difference (MAD), LaGrange function, and the like. SAD is a value obtained by taking the absolute value of each 4x4 block prediction error value and adding the values together. The SATD is obtained by adding an absolute value of coefficients generated by applying a Hadamard transform to a prediction error value of each 4x4 block. The SSD is a value obtained by multiplying a prediction error value of each 4x4 block prediction sample, and the MAD is a value obtained by taking an absolute value of the prediction error value of each 4x4 block prediction sample and obtaining an average. The Lagrange function is a new cost function created by including the length information of a bitstream in the cost function.

Since the compression efficiency is better when the 16x16 intra luminance prediction mode is used than when the 4x4 intra luminance prediction mode is used, a weighting factor α is used when comparing the costs. α> 1 and can be adjusted. When comparing using weights, if the cost value obtained by applying the 4x4 intra luminance prediction mode to each 4x4 block is multiplied by the weight, the value of 16x16 is greater than or equal to the cost value using the 16x16 intra luminance prediction mode. The prediction is performed by applying the intra luminance prediction mode.

If the value obtained by multiplying the cost value when the 4x4 intra luminance prediction mode is applied to each 4x4 block is smaller than the cost value when the 16x16 intra luminance prediction mode is applied, the 4x4 intra luminance prediction mode is applied. Make predictions.

Since the Lagrange function is a new cost function created by including the bitstream length information in the conventional cost function, there is no need to use weights separately.

13 is a block diagram of the prediction mode determination apparatus of the present invention.

The prediction mode determining apparatus includes a 4x4 block prediction mode information analyzing unit 1310, a 16x16 block prediction performing unit 1320, a prediction error calculating unit 1330, and a prediction error comparing unit 1340.

The 4x4 block prediction mode information analyzer 1310 receives mode information of 16 4x4 blocks. And 4x4 prediction mode information is analyzed. That is, each 4x4 luminance prediction mode is classified by group and the number of modes belonging to each group is counted. Then, one or two pieces of 16x16 intra luminance prediction mode information are selected and output by the method described with reference to FIGS. 10 to 12.

The 16x16 block prediction performing unit 1320 performs the prediction by applying the 16x16 intra luminance prediction mode received from the 4x4 block prediction mode information analyzing unit 1310.

The prediction error calculator 1330 calculates and outputs a cost for the prediction sample output from the 16x16 block prediction execution unit 1320. The prediction error is a residue value obtained by subtracting the predicted pixel value from the pixel value of the predicted sample.

The prediction error comparing unit 1340 compares the cost value calculated by the prediction error calculating unit 1330 and selects and outputs a mode having the minimum value. Finally, the cost value obtained by applying the 4x4 intra luminance prediction mode is compared with the cost value obtained by applying the 16x16 intra luminance prediction mode by using the weight to select and output a mode having a lower cost value.

The encoder obtains a prediction error sample by applying the selected prediction mode through the above-described process, and performs transformation and quantization on the prediction error sample. Therefore, the decoder also performs the intra prediction by applying the selected prediction mode.

The invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like, and may also be implemented in the form of a carrier wave (for example, transmission over the Internet). Include. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

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. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. 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.

As described above, when the prediction mode determination method and apparatus of the present invention are used, in selecting the optimal mode in the 16x16 intra prediction mode, the optimum mode can be selected without applying all four modes. There is an effect that can prevent the deterioration.

In other words, by selecting the 16x16 intra luminance prediction mode using the 4x4 intra luminance prediction mode information calculated previously, the optimal 16x16 intra luminance prediction mode can be selected by comparing the prediction errors by applying only one or two prediction modes. It has an effect.

Claims (17)

(a) receiving and analyzing prediction mode information of sub-blocks of a block to be predicted; And (b) selecting a prediction mode of the block to be predicted according to the analysis result. The method of claim 1, wherein step (a) (a1) receiving prediction mode information of sub-blocks of a block to be predicted and grouping them into a predetermined group; (a2) counting the number of prediction modes of the subblocks belonging to each group; And and (a3) selecting and outputting one group or two groups using coefficient values of each group and a predetermined reference value. The method of claim 2, The block to be predicted is a 16x16 luminance block, and the subblock is a 4x4 luminance block. The method of claim 2, wherein step (a1) According to the direction in which prediction is performed, the vertical prediction mode and its adjacent modes are grouped into a first group, the horizontal prediction mode and its adjacent modes are grouped into a second group, and the remaining prediction modes are grouped into one third group. Prediction mode determination method characterized in that the grouping. The method of claim 2, wherein step (a3) If it is determined whether the difference between coefficient values of one group and coefficient values of other groups is greater than or equal to a predetermined reference value, and if the threshold value is greater than or equal to the reference value, a group having a coefficient value greater than or equal to the reference value is selected and outputted, If the difference does not exceed a predetermined reference value, all the groups having a similar coefficient value is selected and outputted. The method of claim 5, The predetermined reference value is arbitrarily set according to the prediction performance and the complexity of the prediction method. The method of claim 2, wherein step (b) The prediction mode determination method, characterized in that for selecting and outputting the prediction mode corresponding to the group output in the step (a3). The method of claim 7, wherein step (b) In step (a3), if Group 1 to which the vertical prediction mode belongs is selected, 16x16 intra luminance mode 0 is selected and output. If Group 2 to which the horizontal prediction mode belongs is selected, 16x16 intra luminance prediction mode 1 is selected. And 16x16 intra luminance prediction mode 2 and 16x16 intra luminance prediction mode 3 are selected and output when the group 3 to which the other prediction mode belongs is selected. The method of claim 1, wherein step (a) The first group belongs to the prediction mode of the 4x4 luma block located in the first column of each 16x16 luma block, the second group belongs to the prediction mode of the 4x4 luma block located in the first row, and the other includes the prediction mode of the 4x4 luma block. The method of claim 1, wherein one or two groups are selected and output according to the number of prediction modes of the third group. The method of claim 1, wherein step (a) Pixel values of the upper macroblock sample and the left macroblock sample used for the 16x16 intra luminance prediction, and the upper block sample and the left block sample used for the 4x4 intra luminance prediction are compared in a row unit or a column unit to determine whether they are equal or predetermined. A prediction mode determination method comprising selecting and outputting a prediction mode for a block to be predicted according to whether the difference is within a difference. (a) The first group to which the prediction mode of the 4x4 luma block located in the first column of the 16x16 luma block belongs, the second group to which the prediction mode of the 4x4 luma block is located in the first row, and other prediction modes of the 4x4 luma block Counting the number of prediction modes of the third group to which the device belongs; And (b) selecting and outputting a prediction mode corresponding to the group having the largest number of prediction modes. (a) comparing pixel values of upper macroblock samples and left macroblock samples used for 16x16 intra luminance prediction, and upper block samples and left block samples used for 4x4 intra luminance prediction in a row unit or a column unit; And and (b) selecting and outputting a 16x16 prediction mode according to whether the comparison result is the same or within a predetermined difference. The method of claim 12, And comparing the upper and left macroblock samples with the upper and left block samples for each pixel value or averaging all pixel values and then comparing the average values. A prediction mode information analyzer configured to receive mode information of a subblock of a block to be predicted and to output a prediction mode of the block to be predicted; A block prediction performing unit for performing prediction on the block to be predicted by applying the prediction mode received from the subblock prediction mode information analyzing unit; A prediction error calculator configured to calculate a prediction error of the prediction sample output from the block prediction execution unit; And And a prediction error comparison unit for comparing the prediction errors calculated by the prediction error calculation unit to select and output a mode having a minimum prediction error. The method of claim 14, And a block to be predicted is a 16x16 luma block and a subblock is a 4x4 luma block. 15. The method of claim 14, wherein the prediction mode information analysis unit Classify each 4x4 luma prediction mode by group, count the number of modes belonging to each group, and 16x16 intra corresponding to one or two groups whose coefficient value is largest or greater than a predetermined difference greater than the coefficient value of another group Prediction mode determination device characterized in that for selecting and outputting the luminance prediction mode information. (a) receiving and analyzing prediction mode information of sub-blocks of a block to be predicted; And and (b) selecting a prediction mode of the block to be predicted according to the analysis result. A computer-readable recording medium having recorded thereon a program for executing the method of determining a prediction mode in a computer.