KR20100020926A - Method for making thumb-nail image of image frame based on h.264 - Google Patents

Abstract

PURPOSE: A method for making thumbnail image of image frame based on h.264 is provided to quickly search specific image through thumbnail images in the multimedia terminal. CONSTITUTION: The prediction mode decision part(110) decides the intra prediction mode of the input N×N image block. The predicted block generating unit(120) creates the predicted block of the spatial domain about the N×N image block. The prediction reference pixel generating unit(180) creates the prediction reference pixel of the image block. The predicted block DC(Direct Current) calculation unit(190) calculates the DC coefficient of the predicted block about the unit image block. The thumbnail image forming unit(197) creates the thumbnail image about the video frame by the combination of the DC coefficient of all N×N image blocks.

Description

Method for making thumb-nail image of image frame based on H.264 standard

The present invention relates to a method for generating a thumbnail image for a video image. More specifically, the DC coefficient of a predictive block for a video block constituting the video image is directly calculated from a predictive reference pixel of an image block in a spatial domain and calculated. The present invention relates to a method of generating a thumbnail image of an image image by calculating a DC coefficient of an image block from a DC coefficient of a prediction block and an error block of the image block.

Video compression has become a key component of broadcast and entertainment media due to its widespread adoption in digital TV, Internet streaming video and DVD-video. The success of digital TV and DVD video is based on the MPEG-2 standard, 15 years after it was published. The utility of this technology has been proven, but it is now outdated. Now is the time to take advantage of advances in processing power to replace MPEG-2 with more effective and efficient technologies. The debate over which technology should replace MPEG-2 continues, but one of the foremost technologies considered in this debate is H.264 / AVC (Advanced VDeo cDing).

H.264 / AVC is a standard for the coded representation of visual information and was developed by the Video Coding Experts Group (VCEG), a research group of the International Telecommunication Union (ITU-T).

H.264 / AVC does not define a codec (enCDer / DECder) separately, but merely defines the syntax of the encoded video bitstream and how to decode such a bitstream. A coded picture consists of several macroblocks, each containing 16 × 16 luma samples and associated color difference samples. Macroblocks within each picture are arranged in a slice, where the macroblocks are present in sequential scan order. Macroblocks are included in I slices, P slices, and B slices, where I slices contain only intra macroblocks, P slices contain inter macroblocks and intra macroblocks, and B slices contain inter macroblocks. Block and intra macroblock. The double intra macroblock is predicted using intra prediction from the reference sample decoded and reconstructed within the current slice.

1 shows a functional block diagram for explaining intra prediction performed in a conventional H.264 / AVC encoder. The H.264 / AVC encoder includes two data flow paths, a "forward" path (left to right) and a "restore" path (right to left). A prediction block is generated from the reference block reconstructed in the "forward" path, and an error pixel block is generated by subtracting from the prediction block that has generated the video block to be currently encoded. Meanwhile, in the "restore" path, the generated error pixel block and the prediction block are added together to generate a reference pixel block of an image block to be encoded next.

The "forward" path and the "restore" path will be described in more detail with reference to FIG. 1. First, referring to the “forward” path, the prediction mode determiner 10 determines an intra prediction mode for an N × N image block of an input spatial region. The prediction block generator 20 generates a prediction block for the N × N image block of the input spatial region by using the stored reference block according to the determined intra prediction mode.

The error block generator 30 generates an N × N error block by subtracting the input N × N image block and the generated prediction block from each other. The transform unit 40 converts the N × N error block into a block-based transform method and converts the N × N error block into an N × N error coefficient block in the transform domain. Preferably, a discrete cosine transform (DCT) or an integer DCT scheme is used among the block-based transformation schemes.

The quantization unit 50 quantizes the N × N error coefficient blocks of the transform domain, and the encoding unit 60 quantizes the N × N error coefficients according to one of coding schemes such as prediction coding, variable length coding, and arithmetic coding. The block is encoded to generate a bitstream.

On the other hand, looking at the "restore" path, the quantized N × N error coefficient block output from the quantization unit 50 inverse quantized by the inverse quantizer 70, the inverse transform unit 80 is inverse quantized N × N The error coefficient block is inversely transformed to generate an N × N error block in a spatial domain. The reference block generator 90 reconstructs the reference block used to generate the prediction block by adding the N × N error block and the prediction block.

As described above, the H.264 / AVC encoder specifies to compress a video image using N × N error coefficient blocks. The prediction block used to generate the N × N error coefficient blocks is N × to be encoded. It is determined through the intra prediction mode of the N image block. In the H.264 / AVC encoder, nine intra prediction modes are defined for luminance 4 × 4 image blocks, four intra prediction modes are defined for luminance 16 × 16 image blocks, and 8 × 8 images for chrominance images. Four intra prediction modes are defined for the block.

Referring to FIG. 2, the nine intra prediction modes used in the 4 × 4 image block will be described in more detail as follows.

1) 0th prediction mode (vertical)

The vertical mode is a prediction mode using 4 pixels of the upper X image block of the block to be currently encoded.

A pixel is filled with 4 pixels of the first column part of the block, B pixel is filled with 4 pixels of the second column part of the block, and C and D pixels are also filled with 4 pixels corresponding to each block column.

2) first prediction mode (horizontal)

The horizontal mode is a mode for predicting using 4 pixels of the left Z image block of the block to be currently encoded.

The I pixel is filled with four pixels of the first row portion of the block, the J pixel is filled with four pixels of the second row portion of the block, and the K and L pixels are also filled with four pixels corresponding to each block row. .

3) second prediction mode (DC)

-DC mode predicts the average value of 4 pixels (I, J, K, L) of the left Z image block and 4 pixels of the upper X image block (A, B, C, D) of the block to be currently encoded. Mode.

4) third prediction mode (diagonal down-left)

-Diagonal down-left mode is a prediction mode using 4 pixels of the upper X image block and up / right Y image block of the block to be currently encoded.

It is filled with a 45 degree angle between the lower left and the upper right of the block to be encoded.

5) fourth prediction mode (diagonal down-right)

Diagonal down-right mode is a prediction mode using 4 pixels of the upper X image block, 1 pixel of the upper left S image block (Q) and 4 pixels of the left Z image block of the block to be encoded. .

-Filled in the 45 degree direction at the bottom right of the block to be encoded.

6) fifth prediction mode (vertical-right)

The vertical-right mode is a prediction mode using 4 pixels of the upper X image block, 1 pixel of the upper left S image block (Q) and 4 pixels of the left Z image block to be encoded.

-Filled in the direction of about 26.6 degrees to the right of the vertical. (Width / height = 1/2)

7) sixth prediction mode (horizontal-down)

-Horizontal-down mode is a prediction mode using 4 pixels of the upper X image block, 1 pixel (Q) of the upper left S image block, and 4 pixels of the left Z image block.

-Filled in the direction of about 26.6 degrees below the horizontal.

8) seventh prediction mode (vertical-left)

-The vertical-left mode is a prediction mode using 4 pixels of the upper X image block and 1 pixel E of the upper right Y image block of the block to be currently encoded.

-The left side of the vertical is filled in about 26.6 degrees.

9) eighth prediction mode (horizontal-up)

-Horizontal-up mode is a mode for predicting using 4 pixels of the left Z image block of the block to be currently encoded.

-Interpolate in the direction of about 26.6 degrees above the horizontal.

Meanwhile, a multimedia terminal such as a mobile phone or a digital TV searches for image data stored in the multimedia terminal using a small size image (hereinafter referred to as a thumbnail image) to preview the stored image image.

3 illustrates an example of a thumbnail image used in a mobile phone. Referring to the thumbnail image illustrated in FIG. 3, a plurality of video or video images stored in the mobile phone are previously displayed as thumbnail images on the display unit. When the user searches for an image or video to be played back through the displayed thumbnail image and selects a predetermined image from the searched video or video, the selected image is played back in the original size image.

Conventional methods for generating thumbnail images from the original image can be divided into two broad categories. The first method is to generate a thumbnail image of the video image by down sampling the video image of the spatial domain. The second method is to generate a thumbnail image of the original image by extracting only DC coefficients from each image coefficient block of the transform region constituting the image. The DC coefficient present at the top left of each image coefficient block in the transform region is an average value for each image pixel block in the spatial domain. The image image generated by extracting only the DC coefficient is downsampled the size of the original image image by 1 / N. Same as did.

In order to generate a thumbnail image of a video image of the H.264 standard using the first method described above, the inverse transform of the residual image block in the transform region to the spatial region is performed by adding the residual image block of the inverse transformed spatial region to the prediction block of the spatial region. After generating the reconstruction block of the region, the reconstruction block is down sampled again to generate a thumbnail image of the video image. Therefore, the first method has a problem in that the image image must be repeatedly transformed into the transform region and the spatial region in order to generate thumbnail images for the image image of the H.264 standard.

Meanwhile, in order to generate a thumbnail image of an H.264 video image by the second method described above, an error coefficient block of the transform domain is inversely transformed into a spatial domain, and an error pixel block of the inverse transformed spatial domain and a prediction block of the spatial domain are transformed. In addition, after restoring the reference pixel block of the spatial domain, the DC coefficient is extracted by converting the reference pixel block of the spatial domain back to the transform domain, which requires a long process time and complicated data processing.

In addition, only some of the reconstructed reference pixel blocks are used for the reference pixels used to generate the prediction blocks according to the intra prediction mode in the H.264 / AVC encoder. However, in the conventional H.264 / AVC encoder, the entire reference pixel block is reconstructed, and thus, many calculations are performed to reconstruct the entire reference pixel block. Furthermore, the conventional H.264 / AVC encoder requires a large storage space for storing reference pixels of an unused reference pixel block, and it takes a long time to generate a prediction block by increasing the number of times of accessing the storage space. Have

The object of the present invention is to solve the problems of the above-described conventional thumbnail image generation method, the object of the present invention is to calculate and calculate only the DC coefficient of the image block required for generating the thumbnail image with a small amount of calculation A method of generating thumbnail images from DC coefficients of an image block is provided.

Another object of the present invention is to generate a thumbnail image with a small process by generating only the reference pixels used to generate the prediction blocks in intra prediction, and to store only the reference pixels used to generate the prediction blocks. It is to provide a method of generating a thumbnail image that requires space.

In order to achieve the object of the present invention, in the method for generating a thumbnail image from the H.264-based N × N image blocks constituting the image frame according to the present invention, the prediction reference to the N × N image block Extracting a pixel, and using the predictive reference pixels of the extracted N × N image block based on the intra prediction mode of the N × N image block, DC coefficients of the prediction block for each unit image block constituting the N × N image block Calculating DC coefficients of the N × N image block by adding the DC coefficients of the prediction block for each unit image block and the DC coefficients of the error blocks for each unit image block, and calculating the DC coefficients of the image block, and N constituting the image frame. Extracting DC coefficients for the xN image blocks to generate thumbnail images of the image frames.

Here, the DC coefficient of the prediction block for the unit image block is calculated by the combination of prediction reference pixels for the N × N image block.

The predictive reference pixel may be a reference pixel around the N × N image block used for intra prediction of the N × N image block.

The predictive reference pixel may include calculating a prediction pixel from a prediction reference pixel of a previously stored unit image block according to an intra prediction mode of the unit image block, calculating an error pixel of the unit image block, and calculating the predicted pixel and the error pixel. Generating a reference reference pixel for the unit image block after the unit image block and storing the prediction reference pixel of the generated unit image block based on the unit image block constituting the N × N image block. It is characterized by.

Here, the error pixel of the unit image block is an error pixel of a pixel located at the boundary of the unit image block, and preferably, the error pixel of the unit image block is an error pixel of a pixel located at the right boundary and the bottom boundary of the unit image block. .

Preferably, when the size of the N × N image block is a 4 × 4 image block, the prediction reference pixels of the N × N image block are 13 reference pixels around the N × N image block used for intra prediction of the N × N image block. It is characterized by that.

According to the present invention, a method of generating a thumbnail image has various effects as compared with the conventional method of generating a thumbnail image.

First, the method of generating a thumbnail image according to the present invention calculates the DC coefficient of the image block directly by using the DC coefficient of the error coefficient block for the image block and the DC coefficient of the prediction block for the image block, thereby providing a space between the spatial domain and the transform domain. It is possible to generate thumbnail images for video images of the H.264 standard without converting the video blocks.

Second, in the method for generating a thumbnail image according to the present invention, a plurality of image images are stored by directly calculating a DC coefficient of an image block using a DC coefficient of an error coefficient block of the image block and a DC coefficient of a prediction block of the image block. In a multimedia terminal, a thumbnail image can be used to quickly search for a specific video image.

Third, the method of generating a thumbnail image according to the present invention calculates a prediction reference pixel by calculating only a reference pixel, that is, a prediction reference pixel used to calculate a DC coefficient of the prediction block, instead of calculating all the reference pixels of the reference pixel block. The amount of computation and memory access required is short, allowing quick thumbnail image creation.

Hereinafter, a method of generating a thumbnail image according to the present invention will be described in detail with reference to the accompanying drawings.

FIG. 4 is a functional block diagram of an apparatus for generating a thumbnail image from an image frame input or stored in an H.264 standard image encoding apparatus according to an embodiment of the present invention.

Referring to FIG. 4, when the “forward” path is first described, when the N × N image block of the spatial domain is input, the prediction mode determiner 110 determines an intra prediction mode of the input N × N image block. The prediction block generator 120 generates a prediction block of a spatial domain for the N × N image block input from the reference pixel previously reconstructed and stored based on the determined intra prediction mode of the N × N image block. The error pixel block generation unit 130 generates N × N error pixel blocks in a spatial domain by subtracting the input N × N image block and the generated prediction block. The transform unit 140 generates the N × N first error coefficient block of the transform domain by converting the generated N × N error pixel block of the spatial domain by a block-based transform scheme. Preferably, a discrete cosine transform (DCT) or an integer DCT scheme is used among the block-based transformation schemes.

 The quantization unit 150 quantizes the N × N first error coefficient block, and the encoding unit 160 quantizes the N × N first quantized according to one of encoding methods such as prediction encoding, variable length encoding, and arithmetic encoding. An error coefficient block is encoded to generate a bitstream.

Meanwhile, referring to the “restore” path, an N × N error coefficient block output from the quantization unit 150 to generate a prediction block for the N × N second image block input after the N × N first image block. Inverse quantization in inverse quantization unit 170. The N × N first error coefficient block output from the quantization unit 150 is inversely quantized by the inverse quantization unit 170.

The prediction reference pixel generator 180 generates a prediction reference pixel of the image block, wherein the prediction reference pixel means a reference pixel around the image block used to calculate the prediction pixel of the image block according to the intra prediction mode of the image block. do. For example, in FIG. 7, a prediction reference pixel for a 4x4 image block ("bar") is an intra of a 4x4 image block ("bar") among reference pixels adjacent to the 4x4 image block ("bar"). 13 reference pixels (A, B, C, D, E, F, G, H, I, J, K, L, M) that are directly used to calculate the prediction pixel according to the prediction mode.

The prediction reference pixel generator 180 does not generate all the reference pixels of the image block in order to generate the prediction reference pixels of the image block, but generates only the reference pixel used as the prediction reference pixel. That is, the prediction reference pixel generator 180 generates only reference pixels positioned at the boundary of the image block as prediction reference pixels, or only reference pixels positioned at the right boundary and the lower boundary of the image block as prediction reference pixels.

For example, the predictive reference pixel generator 180 reconstructs the entire 4x4 image block ("bar") from the 4x4 image block ("bar") so that all of the 4x4 image block ("bar") can be restored. Instead of generating a reference pixel, only seven reference pixels A ', B', C ', D', I ', J', and K 'used as predictive reference pixels are generated. The generated seven reference pixels A ', B', C ', D', I ', J', and K 'are used as some of the 13 predictive reference pixels of the next image block ("sa"). That is, among the seven reference pixels A ', B', C ', D', I ', J', and K ', the reference pixels I', J ', K', and D 'are prestored. Together with another reference pixel D, E, F, G, H, L, O, P, Q of the image block ("sa"), it is used as the predictive reference pixel of the image block ("sa").

The predictive reference pixel generator 180 may predict the predicted reference pixels A, B, C, D, E, of the 4x4 image block ("bar") according to the intra prediction mode of the 4x4 image block ("bar"). Calculation from prediction pixels of 4x4 image blocks ("bars") and error blocks of 4x4 image blocks ("bars") calculated using F, G, H, I, J, K, L, M The predicted reference pixels A ', B', C ', D', I ', J', and K 'for the next image block ("four") after the sum of the error pixels. Calculate).

The error pixel for each unit image block constituting the N × N image block used to generate predictive reference pixels of the N × N image block is calculated by inversely transforming an error coefficient block for each unit image block into a spatial domain. In the present invention, only an error pixel necessary for generating a predictive reference pixel of an N × N image block is calculated. Preferably, in order to reduce the amount of computation in the present invention, the prediction reference pixel of the N × N image block is transformed using a transform element value combined with the error coefficients of the error coefficient block for each unit image block constituting the N × N image block. Compute only the error pixels needed to generate.

In the same manner, the predictive reference pixels of all N × N image blocks constituting the image frame are sequentially calculated and stored in the predictive reference pixel generator 180. The predictive reference pixel generator 180 described above generates a predictive reference pixel in units of 4 × 4 image blocks, that is, in units of unit image blocks.

The prediction block DC calculator 190 configures a unit image constituting the N × N image block based on the intra prediction mode of the N × N image block determined by the prediction mode determiner 110 and the prediction reference pixels of the N × N image block. Compute the DC coefficients of the predictive block for the block. The DC coefficient of the prediction block is a DC coefficient of the prediction block for each unit image block generated by block-based transforming the prediction block for each unit image block constituting the N × N image block. Is calculated as the predictive reference pixel of the N × N image block. The image block DC calculator 195 calculates the DC coefficient of the N × N image block by adding the calculated DC coefficients of the prediction block for each unit image block and the DC coefficients of the error coefficient block for each unit image block. The thumbnail image generator 197 generates a thumbnail image of the image frame by combining DC coefficients of all N × N image blocks constituting the image frame. The unit image block constituting the N × N image block described above is a 4 × 4 image block as a unit size of an image block used to generate a predictive reference pixel of the N × N image block and calculate a DC coefficient. Therefore, in the case of an 8x8 image block, a prediction reference pixel is generated or a DC coefficient is calculated for each of the 4x4 unit image blocks constituting the 8x8 image block, and in the case of a 16x16 image block, 16 4x4 units A predictive reference pixel is generated or a DC coefficient is calculated for each image block.

FIG. 5 is a functional block diagram of an apparatus for generating a thumbnail image from an image frame received or stored in an H.264 standard image decoding apparatus according to an embodiment of the present invention.

Referring to FIG. 5, the prediction reference pixel generator 210 configures each N × N image block from the prediction reference pixels of the N × N image block based on the intra prediction mode of the N × N image block. The prediction pixel of the unit image block is calculated, and the error pixel of the error coefficient block for each unit image block is calculated. The predictive reference pixel generator 210 generates a predictive reference pixel in units of each image block by adding the prediction pixels and the error pixels of each unit image block constituting the calculated N × N image block.

The prediction block DC calculator 220 may determine a prediction block for each unit image block constituting the N × N image block based on the intra prediction mode of the N × N image block and the prediction reference pixels of the previously stored N × N image block. Calculate the DC coefficient. The DC coefficient of the prediction block is a DC coefficient of the prediction coefficient block generated by block-based transforming the prediction block for each unit image block constituting the N × N image block. In the present invention, the DC coefficient of the prediction block is N × N image. Computed with the predictive reference pixels of the block. The image block DC calculator 230 calculates the DC coefficient of each unit image block constituting the input N × N image block by adding the calculated DC coefficients of the prediction block and the DC coefficients of the error coefficient block for each unit image block. do. That is, the image block DC calculator 230 calculates the DC coefficient of each unit image block constituting the N × N image block in units of N × N image blocks. The thumbnail image generator 295 generates a thumbnail image of an image frame by combining DC coefficients of all image blocks.

6 is a functional block diagram for explaining in more detail the prediction reference pixel generator according to the present invention.

Referring to FIG. 6, the predictive reference pixel generator includes a transform element calculator 310, a predictive reference pixel calculator 320, and a predictive reference pixel storage 330. The predictive reference pixel generator generates a predictive reference pixel by adding the prediction pixels generated according to the intra prediction mode of each unit image block and the error pixels of each unit image block, and the predictive reference pixel generator generates a predictive reference pixel from the error coefficient block of each unit image block. The transform element value is used to reduce the amount of computation required to calculate the error pixel of each unit image block. The transform element calculator 310 converts the transform element values H ₀ , H ₁ , H ₂ , H ₃ , V ₀ , V ₁ used to calculate each error pixel of the unit image block from each error coefficient block of the unit image block. , V ₂ , V ₃ ) is calculated by the following equations (1) to (8).

[Equation 1]

[Equation 2]

[Equation 3]

[Equation 4]

[Equation 5]

[Equation 6]

[Equation 7]

[Equation 8]

Here, the error block W of the transform region for the unit image block is expressed by Equation 9 below.

[Equation 9]

The prediction reference pixel calculator 320 is a prediction pixel generated according to the intra prediction mode of the unit image block from the prediction reference pixels of the unit image block previously stored in the prediction reference pixel storage unit 330 and the error pixel of the unit image block. S is calculated to calculate a prediction reference pixel for the unit image block after the unit image block. That is, the prediction reference pixel calculator 320 calculates the prediction reference pixel by adding only the prediction pixel and the error pixel at the position used as the prediction reference pixel.

Referring to FIG. 7, the prediction reference pixel of the 4x4 image block calculated according to the intra prediction mode of the 4x4 image block will be described in more detail. Equation (10) below is an equation for generating a predictive reference pixel for the 4x4 image block according to the vertical mode which is the 0th mode in the 4x4 luminance mode.

[Equation 10]

A '= ((A) * a + V0 + V1 + V2 + 1/2 * V3 + a / 2) / a

B '= ((B) * a + V0 + 1/2 * V1-V2-V3 + a / 2) / a

C '= ((C) * a + V0-1/2 * V1-V2 + V3 + a / 2) / a

D '= ((D) * a + H0-H1 + H2-1/2 * H3 + a / 2) / a

I '= ((D) * a + H0 + H1 + H2 + 1/2 * H3 + a / 2) / a

J '= ((D) * a + H0 + 1/2 * H1-H2-H3 + a / 2) / a

K '= ((D) * a + H0-1/2 * H1 + H2 + H3 + a / 2) / a

L '= D'

Equation (11) is an equation for generating a predictive reference pixel for a 4x4 image block according to a horizontal mode which is the first mode in the 4x4 luminance mode.

[Equation 11]

A '= ((L) * a + V0 + V1 + V2 + 1/2 * V3 + a / 2) / a

B '= ((L) * a + V0 + 1/2 * V1-V2-V3 + a / 2) / a

C '= ((L) * a + V0-1/2 * V1-V2 + V3 + a / 2) / a

D '= ((L) * a + H0-H1 + H2-1/2 * H3 + a / 2) / a

I '= ((I) * a + H0 + H1 + H2 + 1/2 * H3 + a / 2) / a

J '= ((J) * a + H0 + 1/2 * H1-H2-H3 + a / 2) / a

K '= ((K) * a + H0-1/2 * H1 + H2 + H3 + a / 2) / a

L '= D'

Equation (12) is an equation for generating a predictive reference pixel for the 4x4 image block according to the DC mode which is the second mode in the 4x4 luminance mode.

[Equation 12]

Avg = (A + B + C + D + I + J + K + L) / 8

A '= ((Avg) * a + V0 + V1 + V2 + 1/2 * V3 + a / 2) / a

B '= ((Avg) * a + V0 + 1/2 * V1-V2-V3 + a / 2) / a

C '= ((Avg) * a + V0-1/2 * V1-V2 + V3 + a / 2) / a

D '= ((Avg) * a + H0-H1 + H2-1/2 * H3 + a / 2) / a

I '= ((Avg) * a + H0 + H1 + H2 + 1/2 * H3 + a / 2) / a

J '= ((Avg) * a + H0 + 1/2 * H1-H2-H3 + a / 2) / a

K '= ((Avg) * a + H0-1/2 * H1 + H2 + H3 + a / 2) / a

L '= D'

Equation (13) is an equation for generating a predictive reference pixel for a 4x4 image block according to the diagonal down-left mode, which is the third mode in the 4x4 luminance mode.

[Equation 13]

temp1 = (D + F + 2 * E + 2) / 4

A '= ((temp1) * a + V0 + V1 + V2 + 1/2 * V3 + a / 2) / a

temp2 = (E + G + 2 * F + 2) / 4

B '= ((temp3) * a + V0 + 1/2 * V1-V2-V3 + a / 2) / a

temp3 = (F + H + 2 * G + 2) / 4

C '= ((temp3) * a + V0-1/2 * V1-V2 + V3 + a / 2) / a

temp4 = (G + 3 * H + 2) / 4

D '= ((temp4) * a + H0-H1 + H2-1/2 * H3 + a / 2) / a

I '= A'

J '= B'

K '= C'

L '= D'

Equation (14) is an equation for generating a predictive reference pixel for a 4x4 image block in a 4 × 4 luminance mode according to a fourth diagonal mode down-right mode.

[Equation 14]

temp1 = (L + 2 * K + J + 2) / 4

A '= ((temp1) * a + V0 + V1 + V2 + 1/2 * V3 + a / 2) / a

temp2 = (K + 2 * J + O + 2) / 4

B '= ((temp2) * a + V0 + 1/2 * V1-V2-V3 + a / 2) / a

temp3 = (J + 2 * I + M + 2) / 4

C '= ((temp3) * a + V0-1/2 * V1-V2 + V3 + a / 2) / a

temp4 = (I + 2 * M + A + 2) / 4

D '= ((temp4) * a + H0-H1 + H2-1/2 * H3 + a / 2) / a

temp5 = (B + 2 * C + D + 2) / 4

I '= ((temp5) * a + H0 + H1 + H2 + 1/2 * H3 + a / 2) / a

temp6 = (A + 2 * B + C + 2) / 4

J '= ((temp6) * a + H0 + 1/2 * H1-H2-H3 + a / 2) / a

temp7 = (M + 2 * A + B + 2) / 4

K '= ((temp7) * a + H0-1/2 * H1 + H2 + H3 + a / 2) / a

L '= D'

Equation (15) is an equation for generating a predictive reference pixel for the 4x4 image block according to the vertical right mode which is the fifth mode in the 4x4 luminance mode.

[Equation 15]

temp1 = (I + 2 * J + K + 2) / 4

A '= ((temp1) * a + V0 + V1 + V2 + 1/2 * V3 + a / 2) / a

temp2 = (I + 2 * M + A + 2) / 4

B '= ((temp2) * a + V0 + 1/2 * V1-V2-V3 + a / 2) / a

temp3 = (M + 2 * A + B + 2) / 4

C '= ((temp3) * a + V0-1/2 * V1-V2 + V3 + a / 2) / a

temp4 = (A + 2 * B + C + 2) / 4

D '= ((temp4) * a + H0-H1 + H2-1/2 * H3 + a / 2) / a

temp5 = (C + D + 1) / 2

I '= ((temp5) * a + H0 + H1 + H2 + 1/2 * H3 + a / 2) / a

temp6 = (B + 2 * C + D + 2) / 4

J '= ((temp6) * a + H0 + 1/2 * H1-H2-H3 + a / 2) / a

temp7 = (B + C + 1) / 2

K '= ((temp7) * a + H0-1/2 * H1 + H2 + H3 + a / 2) / a

L '= D'

Equation (16) is an equation for generating a predictive reference pixel for the 4x4 image block according to the horizontal down mode which is the sixth mode in the 4x4 luminance mode.

[Equation 16]

temp1 = (K + L + 1) / 2

A '= ((temp1) * a + V0 + V1 + V2 + 1/2 * V3 + a / 2) / a

temp2 = (J + 2 * K + L + 2) / 4

B '= ((temp2) * a + V0 + 1/2 * V1-V2-V3 + a / 2) / a

temp3 = (J + K + 1) / 2

C '= ((temp3) * a + V0-1/2 * V1-V2 + V3 + a / 2) / a

temp4 = (I + 2 * J + K + 2) / 4

D '= ((temp4) * a + H0-H1 + H2-1/2 * H3 + a / 2) / a

temp5 = (A + 2 * B + C + 2) / 4

I '= ((temp5) * a + H0 + H1 + H2 + 1/2 * H3 + a / 2) / a

temp6 = (I + 2 * M + A + 2) / 4

J '= ((temp6) * a + H0 + 1/2 * H1-H2-H3 + a / 2) / a

temp7 = (M + 2 * I + J + 2) / 4

K '= ((temp7) * a + H0-1/2 * H1 + H2 + H3 + a / 2) / a

L '= D'

Equation (17) is an equation for generating a predictive reference pixel for the 4x4 image block according to the vertical down mode, which is the seventh mode in the 4x4 luminance mode.

[Equation 17]

temp1 = (B + 2 * C + D + 2) / 4

A '= ((temp1) * a + V0 + V1 + V2 + 1/2 * V3 + a / 2) / a

temp2 = (C + 2 * D + E + 2) / 4

B = ((temp2) * a + V0 + 1/2 * V1-V2-V3 + a / 2) / a

temp3 = (D + 2 * E + F + 2) / 4

C '= ((temp3) * a + V0-1/2 * V1-V2 + V3 + a / 2) / a

temp4 = (E + 2 * F + G + 2) / 4

D '= ((temp4) * a + H0-H1 + H2-1/2 * H3 + a / 2) / a

temp5 = (D + E + 1) / 2

I '= ((temp5) * a + H0 + H1 + H2 + 1/2 * H3 + a / 2) / a

temp6 = (D + 2 * E + F + 2) / 4

J '= ((temp6) * a + H0 + 1/2 * H1-H2-H3 + a / 2) / a

temp7 = (E + F + 1) / 2

K '= ((temp7) * a + H0-1/2 * H1 + H2 + H3 + a / 2) / a

L '= D'

Equation (18) is an equation for generating a predictive reference pixel for the 4x4 image block according to the horizontal up mode which is the eighth mode in the 4x4 luminance mode.

Equation 18

A '= ((L) * a + V0 + V1 + V2 + 1/2 * V3 + a / 2) / a

B '= ((L) * a + V0 + 1/2 * V1-V2-V3 + a / 2) / a

C '= ((L) * a + V0-1/2 * V1-V2 + V3 + a / 2) / a

D '= ((L) * a + H0-H1 + H2-1/2 * H3 + a / 2) / a

temp1 = (J + 2 * K + L + 2) / 4

I '= ((temp1) * a + H0 + H1 + H2 + 1/2 * H3 + a / 2) / a

temp2 = (K + 2 * L + L + 2) / 4

J '= ((temp2) * a + H0 + 1/2 * H1-H2-H3 + a / 2) / a

K '= ((L) * a + H0-1/2 * H1 + H2 + H3 + a / 2) / a

L '= D'

The four 16 × 16 luminance prediction modes and the four 8 × 8 color difference prediction modes are determined as the luminance prediction mode or the color difference prediction mode of the input 16 × 16 image block or 8 × 8 image block. A 4x4 image block is divided into 4x4 image blocks, and each of the divided 4x4 image blocks is generated by using the above-described equations (10) to (18) according to the determined intra prediction mode. .

8 is a flowchart illustrating a method of generating a predictive reference pixel according to an embodiment of the present invention.

Referring to FIG. 8, when a 4 × 4 image block constituting an image frame is input (S1), an intra prediction mode of the input 4 × 4 image block is determined (S3). Prediction of the 4x4 image block input after the 4x4 image block from the prediction reference pixels of the prestored 4x4 image block and the error pixels of the 4x4 image block according to the intra prediction mode of the input 4x4 image block A reference pixel is generated (S5). A prediction reference pixel for the 4x4 image block input after the 4x4 image block is generated differently according to the determined intra prediction mode of the 4x4 image block. The prediction reference pixels A ', B', C ', D', I ', J', K ', and L' generated in step S3 are stored in the storage unit (S7) and the stored prediction reference pixels A ' , B ', C', D ', I', J ', K', L ') are used to calculate the DC coefficients of the predictive block for the 4x4 image block that is input after the 4x4 image block.

9 is a flowchart illustrating a method of generating a thumbnail image of an image frame of the H.264 standard according to an embodiment of the present invention.

Predicting reference pixels for the N × N image blocks are extracted (S11), and constituting each N × N image block according to the prediction prediction pixels for the extracted N × N image blocks and the intra prediction mode of the determined N × N image blocks. The DC coefficient E of the prediction block for the unit image block is calculated (S13). A prediction reference pixel for each unit image block constituting the N × N image block is generated and stored in the same manner as described above.

Comprising an N × N image block by using the DC coefficient (E) of the prediction block for each unit image block constituting the N × N image block and the DC coefficient (W ₀₀ ) of the error coefficient block for each unit image block. The DC coefficient of each unit image block is calculated (S15). The DC coefficient for each unit image block constituting the N × N image block is calculated as shown in Equation (19) below.

[Equation 19]

DC coefficient of N × N picture block = E + W ₀₀

A method of calculating DC coefficients for 4 × 4 image blocks according to nine 4 × 4 luminance prediction modes will be described in detail below.

Equation (20) is an equation for calculating DC coefficients for a 4x4 image block according to the vertical mode which is the 0th mode.

[Equation 20]

DC = 16 * A + 16 * B + 16 * C + 16 * D + W ₀₀

Equation (21) is an equation for calculating a DC coefficient for a 4x4 image block according to a horizontal mode that is a first mode.

[Equation 21]

DC = 16 * I + 16 * J + 16 * K + 16 * L + W ₀₀

Equation (22) is an equation for calculating a DC coefficient for a 4x4 image block according to the DC mode, which is the second mode.

[Equation 22]

DC = (A + B + C + D + I + J + K + L) / 8 + W ₀₀

Equation (23) is an equation for calculating a DC coefficient for a 4x4 image block according to the third mode, the down-left mode.

&Quot; (23) "

DC = A + 4 * B + 8 * C + 12 * D + 14 * E + 12 * F + 8 * G + 5 * H + W ₀₀

Equation (24) is an equation for calculating DC coefficients for 4 × 4 image blocks according to the fourth mode, Diagonal down-right mode.

[Equation 24]

DC = 14 * M + 12 * A + 8 * B + 4 * C + D + 12 * I + 8 * J + 4 * K + L + W ₀₀

Equation (25) is an equation for calculating DC coefficients for a 4x4 image block according to a vertical right mode which is a fifth mode.

[Equation 25]

DC = 11 * M + 16 * A + 15 * B + 10 * C + 3 * D + 5 * I + 3 * J + K + W ₀₀

Equation (26) is an equation for calculating DC coefficients for a 4x4 image block according to a horizontal down mode, which is a sixth mode.

[Equation 26]

DC = 11 * M + 5 * A + 3 * B + C + 16 * I + 15 * J + 10 * K + 3 * L + W ₀₀

Equation (27) is an equation for calculating a DC coefficient for a 4x4 image block according to the vertical left mode, which is the seventh mode.

[Equation 27]

DC = 3 * A + 10 * B + 15 * C + 16 * D + 13 * E + 6 * F + G + W ₀₀

Equation 28 is an equation for calculating DC coefficients for a 4x4 image block according to a horizontal up mode which is an eighth mode.

[Equation 28]

DC = 3 * I + 10 * J + 15 * K + 36 * L + W ₀₀

Meanwhile, the DC coefficients of 16 × 16 image blocks according to four 16 × 16 luminance prediction modes will be described in more detail as follows. The DC coefficients for the 16 × 16 image block are calculated by adding the DC coefficients and the error coefficients of the prediction block to each unit image block constituting the 16 × 16 image block, and calculating the DC coefficients of each unit image block. Calculate the DC coefficient of.

Equation (29) calculates DC coefficients for the unit image blocks constituting the 16x16 image block, that is, the first 4x4 image block to the 16th 4x4 image block according to the vertical mode which is the 0th mode. Expression

[Equation 29]

DC = 16 * A _i + 16 * B _i + 16 * C _i + 16 * D _i + W _00ij i = 1, 2, 3, 4 j = 1, 2, 3, 4

Equation (30) is an equation for calculating DC coefficients for the first 4x4 image blocks to the sixteenth 4x4 image blocks constituting the 16x16 image block according to the horizontal mode which is the first mode.

Equation 30

DC = 16 * I _j + 16 * J _j + 16 * K _j + 16 * L _j + W _00ij i = 1, 2, 3, 4 j = 1, 2, 3, 4

Equation 31 calculates DC coefficients for the first 4x4 image blocks to the 16th 4x4 image blocks constituting the 16x16 image block according to the DC mode which is the second mode.

Equation 31

DC = (A _i + B _i + C _i + D _i + I _j + J _j + K _j + L _j ) / 8 + W _00ij

i = 1, 2, 3, 4 j = 1, 2, 3, 4

Equation (32) is a formula for calculating DC coefficients for the first 4x4 image blocks to the sixteenth 4x4 image blocks constituting the 16x16 image block according to the third mode.

Equation 32

ih = (A ₃ -C ₂ ) + 2 * (B ₃ -B ₂ ) + 3 * (C ₃ -A ₂ ) + 4 * (D ₃ -D ₁ ) + 5 * (A ₄ -C ₁ ) + 6 * (B ₄ -B ₁ ) + 7 * (C ₄ -A ₁ ) + 8 * (D ₄ -M)

iv = (I ₃ -K ₂ ) + 2 * (J ₃ -J ₂ ) + 3 * (J ₃ -I ₂ ) + 4 * (J ₃ -L ₁ ) + 5 * (I ₄ -K ₁ ) + 6 * (J ₄ -J ₁ ) + 7 * (J ₄ -I ₁ ) + 8 * (J ₄ -M)

a = 64

ib = (5 * ih + a / 2) / a

ic = (5 * iv + a / 2) / a

iaa = 16 * (D ₄ + L ₄ )

This

4 × 4 first block is DC = 2 * iaa-11 * ib -11 * ic + a / 2 + W ₀₀ of the first block

The 4x4 second block is DC = 2 * iaa-3 * ib -11 * ic + a / 2 + W ₀₀ of the second block

The 4x4 third block is DC = 2 * iaa + 5 * ib -11 * ic + a / 2 + W ₀₀ of the third block

4 × 4 fourth block is DC = 2 * iaa + 13 * ib -11 * ic + a / 2 + W ₀₀ of the fourth block

4x4 fifth block is DC = 2 * iaa-11 * ib -3 * ic + a / 2 + W ₀₀ of fifth block

4 × 4 6th block is DC = 2 * iaa-3 * ib -3 * ic + a / 2 + W ₀₀ of 6th block

4x4 seventh block is DC = 2 * iaa + 5 * ib -3 * ic + a / 2 + W ₀₀ of the seventh block

4x4 eighth block is DC = 2 * iaa + 13 * ib -3 * ic + a / 2 + W ₀₀ of the eighth block

4x4 ninth block is DC = 2 * iaa-11 * ib + 5 * ic + a / 2 + W ₀₀ of ninth block

4x4 10th block is DC = 2 * iaa-3 * ib + 5 * ic + a / 2 + W ₀₀ of 10th block

4 × 4 11th block is W ₀₀ of DC = 2 * iaa + 5 * ib + 5 * ic + a / 2 + 11th block

4x4 12th block is DC = 2 * iaa + 13 * ib + 5 * ic + a / 2 + W ₀₀ of 12th block

4x4 13th block is DC = 2 * iaa-11 * ib + 13 * ic + a / 2 + W ₀₀ of the 13th block

4x4 14th block is DC = 2 * iaa-3 * ib + 13 * ic + a / 2 + W ₀₀ of 14th block

4x4 15th block is DC = 2 * iaa + 5 * ib + 13 * ic + a / 2 + W ₀₀ of the 15th block

4 × 4 16th block is DC = 2 * iaa + 13 * ib + 13 * ic + a / 2 + W ₀₀ of the 16th block

Meanwhile, the DC coefficients of 8x8 image blocks according to four 8x8 color difference prediction modes will be described in detail below. The DC coefficients for the 8x8 image block are calculated by adding the DC coefficients and the error coefficients of the prediction block to each unit image block constituting the 8x8 image block to calculate the DC coefficient of each unit image block. Calculate the DC coefficient of.

Equation (33) calculates DC coefficients for the first 4x4 image blocks to the fourth 4x4 image blocks constituting the 8x8 image block according to the vertical mode, which is the 0th mode.

[Equation 33]

DC = 16 * A _i + 16 * B _i + 16 * C _i +16 _i * D + W _00ij i = 1, 2 j = 1, 2

Equation (34) calculates DC coefficients for the first 4x4 image blocks to the fourth 4x4 image blocks constituting the 8x8 image block according to the horizontal mode, which is the first mode.

[Equation 34]

DC = 16 * I _j + 16 * J _j + 16 * K _j + 16 * L _j + W _00ij i = 1, 2 j = 1, 2

Equation (35) calculates DC coefficients for the first 4x4 image blocks to the fourth 4x4 image blocks constituting the 8x8 image block according to the DC mode which is the second mode.

[Equation 35]

DC = (A _i + B _i + C _i + D _i + I _j + J _j + K _j + L _j ) / 8 + W _00ij i = 1, 2 j = 1, 2

Equation (36) calculates DC coefficients for the first 4x4 image blocks to the fourth 4x4 image blocks constituting the 8x8 image block according to the plane mode, which is the third mode.

[Equation 36]

ih = (A ₂ -C ₁ ) + 2 * (B ₂ -B ₁ ) + 3 * (C ₂ -A ₁ ) + 4 * (D ₂ -M)

iv = (I2-K ₁ ) + 2 * (J ₂ -J ₁ ) + 3 * (J ₂ -I ₁ ) + 4 * (J ₂ -M)

a = 64

ib = (17 * ih + a / 2) / a

ic = (17 * iv + a / 2) / a

iaa = 16 * (D ₂ + L ₂ )

This

4 × 4 first block is DC = 2 * iaa-3 * ib -3 * ic + a / 2 + W ₀₀ of the first block

The 4 × 4 second block is DC = 2 * iaa + 5 * ib -3 * ic + a / 2 + W ₀₀ of the second block

The 4x4 third block is DC = 2 * iaa -3 * ib + 5 * ic + a / 2 + W ₀₀ of the third block

4 × 4 fourth block is DC = 2 * iaa + 5 * ib + 5 * ic + a / 2 + W ₀₀ of the fourth block

Obtained as

Prediction reference pixels A ₁ , B ₁ , C ₁ , D ₁ , A ₂ , B ₂ , C ₂ , D in the spatial domain for the 16 × 16 image blocks input in equations (29)-(32). ₂ , A ₃ , B ₃ , C ₃ , D ₃ , A ₄ , B ₄ , C ₄ , D ₄ , MI ₁ , J ₁ , K ₁ , L ₁ , I ₂ , J ₂ , K ₂ , L ₂ , I ₃ , J ₃ , K ₃ , L ₃ , I ₄ , J ₄ , K ₄ , L ₄ ) and the first 4 × 4 blocks to 16th 16 × 4 blocks constituting the 16 × 16 image block are shown in FIG. It's like it's in 11.

Prediction reference pixels (A ₁ , B ₁ , C ₁ , D ₁ , A ₂ , B ₂ , C ₂ , D in the spatial domain for the 8 × 8 image block input in equations (33) to (36) ₂ , M, I ₁ , J ₁ , K ₁ , L ₁ , I ₂ , J ₂ , K ₂ , L ₂ ) and the first 4 × 4 blocks to fourth 4 × 4 blocks constituting an 8 × 8 image block. The order of is shown in FIG.

As shown in equations (20) to (36), it is not necessary to reconstruct all the reference pixel blocks for the 4x4 image block, and among the pixels constituting the reference pixel block for the 4x4 image block. The prediction pixels of the 4x4 image block are calculated by reconstructing only the reference pixels, that is, the prediction reference pixels used to calculate the prediction block of the 4x4 image block. The DC coefficient of the prediction block for each unit image block constituting the N × N image block is calculated directly from the prediction reference pixel of the N × N image block, and the DC coefficient of the prediction block for each calculated unit image block and The DC coefficients of the unit image blocks constituting the N × N image block are simply calculated by adding the DC coefficients of the error coefficient blocks to the unit image blocks.

Preferably, in the case of the first image block among the image blocks constituting one image frame, the prediction reference data of the first image block is set to 128, and the DC coefficient is calculated as shown in Equation (37) below.

[Equation 37]

DC coefficient = (A + B + C + D + I + J + K + L) / 8 + W ₀₀

It is determined whether the DC coefficients are calculated for all N × N image blocks constituting the image frame (S17). Steps S11 to S15 are repeated to calculate DC coefficients of all N × N image blocks. If it is determined in step S17 that the DC coefficients of all N × N image blocks constituting the image frame are calculated, a thumbnail image of the image frame is generated using the DC coefficients of all N × N image blocks (S19).

10 is required for generating a prediction reference pixel according to a method of generating a prediction reference pixel using a conventional fast inverse DCT (IDCT), a method of generating a prediction reference pixel using a direct inverse DCT, and a method of generating a prediction reference pixel according to the present invention. This chart compares the number of calculations and the number of memory accesses. 10 is a result showing the number of calculations and the number of memory accesses required to generate a predictive reference pixel from a 4x4 error coefficient block.

Referring to FIG. 10, the number of addition operations required to recover all the reference pixels required in the method of generating the predictive block using fast inverse DCT is 96 times, the number of shifts is 48, and the memory accesses. The number is 192 times. Meanwhile, in the method of generating the predictive reference pixel using the direct inverse DCT, the number of addition calculations required to recover all the reference pixels required is 240, the number of shifts is 112, and the number of memory accesses is 64. Meanwhile, in the method of generating the predictive reference pixel according to the present invention, the number of addition calculations required to generate only the predictive reference pixel required to generate the predictive block is 59, the number of shifts is 26, and the number of memory accesses is 8. As can be seen in Figure 10, the method of generating a prediction block according to the present invention is faster than the case of generating a prediction block using a conventional fast inverse DCT or direct inverse DCCT image block faster due to less computation amount and fewer memory access times Can be encoded.

Meanwhile, the above-described embodiments of the present invention can be written as a program that can be executed in a computer, and can be implemented in a general-purpose digital computer that operates the program using a computer-readable recording medium.

The computer-readable recording medium may be a magnetic storage medium (for example, a ROM, a floppy disk, a hard disk, etc.), an optical reading medium (for example, a CD-ROM, DVD, etc.) and a carrier wave (for example, the Internet). Storage medium).

Although the present invention has been described with reference to the embodiments shown in the drawings, this is merely exemplary, and it will be understood by those skilled in the art that various modifications and equivalent other embodiments are possible. Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

1 shows a functional block diagram for explaining intra prediction performed in a conventional H.264 / AVC encoder.

FIG. 2 is a diagram for explaining nine intra prediction modes used in a 4x4 image block.

3 illustrates an example of a thumbnail image used in a mobile phone.

FIG. 4 is a functional block diagram of an apparatus for generating a thumbnail image from an image frame input or stored in an H.264 standard image encoding apparatus according to an embodiment of the present invention.

FIG. 5 is a functional block diagram of an apparatus for generating a thumbnail image from an image frame received or stored in an H.264 standard image decoding apparatus according to an embodiment of the present invention.

6 is a functional block diagram for explaining in more detail the prediction reference pixel generator according to the present invention.

FIG. 7 illustrates predictive reference pixels of a 4x4 image block calculated according to an intra prediction mode of the 4x4 image block.

8 is a flowchart illustrating a method of generating a predictive reference pixel according to an embodiment of the present invention.

9 is a flowchart illustrating a method of generating a thumbnail image of an image frame of the H.264 standard according to an embodiment of the present invention.

10 is a chart comparing the number of calculations and the number of memory accesses of a method of generating a prediction block using a conventional fast inverse DCT (IDCT), a method of generating a prediction block using a direct inverse DCT, and a method of generating a prediction block according to the present invention. .

11 shows an example of a 16x16 image block.

12 shows an example of an 8x8 image block.

Claims (9)

In the method for generating a thumbnail image from the H.264-based N × N image blocks constituting the image frame, (a) extracting a predictive reference pixel for the N × N image block; (b) calculating DC coefficients of a prediction block for each unit image block constituting the N × N image block with the predicted reference pixels of the extracted N × N image block based on the intra prediction mode of the N × N image block; Doing; And (c) calculating the DC coefficient of the N × N image block by adding the calculated DC coefficients of the prediction block for each unit image block and the DC coefficients of the error block for each unit image block. How to produce. The method of claim 1, wherein the steps (a) to (c) are repeated for the N × N image blocks constituting the image frame. And generating a thumbnail image of the image frame by extracting DC coefficients for the N × N image blocks constituting the image frame. The DC coefficient of the prediction block for each unit image block is And calculating a combination of predicted reference pixels for the N × N image block. The method of claim 2, wherein the prediction reference pixel is And a reference pixel around the N × N image block used for intra prediction of the N × N image block. The method of claim 2, wherein the prediction reference pixel is (a1) calculating a prediction pixel from a predicted reference pixel of the unit image block according to an intra prediction mode of the unit image block; (a2) calculating an error pixel of the unit image block; (a3) generating a prediction reference pixel for the unit image block after the unit image block by adding the calculated prediction pixel and the error pixel; And (a4) generating a thumbnail image based on a unit image block constituting the N × N image block, including storing the predicted reference pixel of the generated unit image block. The method of claim 5, wherein the error pixel of the unit image block is And an error pixel for a pixel located at a boundary of the unit image block. The method of claim 6, wherein the error pixel of the unit image block And an error pixel for pixels located at right and bottom boundaries of the unit image block. 8. The method of claim 7, wherein the size of the N × N picture block is a 4 × 4 picture block. The predictive reference pixel of the N × N image block is 13 reference pixels around the N × N image block used for intra prediction of the N × N image block. The method of claim 5, wherein the error pixel of the unit image block is Calculated using a transform element value calculated from an error coefficient block for the unit image block, The conversion element values H ₀ , H ₁ , H ₂ , H ₃ , V ₀ , V ₁ , V ₂ , V ₃ are It is generated by the following equations (1) to (8). [Equation 1]

[Equation 2]

[Equation 3]

[Equation 4]

[Equation 5]

[Equation 6]

[Equation 7]

[Equation 8]

Here, the error coefficient block (W) of the transform region for the unit image block is [Equation 9]

Method of generating a thumbnail image, characterized in that the equation (9).

Images