KR100912077B1 - Apparatus and method for adaptively reading image data for intra prediction, and system of adaptive intra prediction for decoding image using it - Google Patents

Abstract

The present invention relates to an adaptive image data read control device for intra prediction and a method thereof, and to an adaptive intra prediction system for image decoding using the same, wherein the image data stored in the memory for image decoding is read. By determining the reading order based on whether the corresponding prediction mode of the prediction target block corresponds to a horizontal / vertical mode or a non-horizontal / vertical mode, and performing reading of the corresponding image data according to the determined reading rank, We want to be able to perform intra prediction quickly with minimal computation.

To this end, the present invention, in the adaptive image data read control device, for the prediction target blocks that are the target of intra prediction, the prediction mode is any of the horizontal / vertical mode or non-horizontal / vertical mode Control means for determining a 'priority of reading of reference data necessary for the intra prediction' on the basis of whether Read address generation means for generating a read address for the reference data; And a read mode performing means for reading the image data corresponding to the read address generated by the read address generating means, according to the read rank determined by the control means.

Intra prediction, image data reading, reading rank, read control, horizontal / vertical prediction mode, non-horizontal / vertical prediction mode

Description

APPARATUS AND METHOD FOR ADAPTIVELY READING IMAGE DATA FOR INTRA PREDICTION, AND SYSTEM OF ADAPTIVE INTRA PREDICTION FOR DECODING IMAGE USING IT}

The present invention relates to an adaptive image data read control device for intra prediction and a method thereof, and to an adaptive intra prediction system for image decoding using the same, and more particularly, to reading image data stored in a memory for image decoding. In performing the operation, the reading order is determined based on whether the corresponding prediction mode of the block to be predicted corresponds to the horizontal / vertical mode, and the image data is read according to the determined reading order, thereby quickly performing a minimum computation amount. The present invention relates to an adaptive image data read control device for intra prediction and a method thereof, and to an adaptive intra prediction system for image decoding using the same.

The present invention is derived from a study conducted as part of the IT new growth engine core technology development project of the Ministry of Information and Communication and the Ministry of Information and Communication Research and Development. [Task Management Number: 2005-S-078-02, Title: Low-Power Graphic Processing SoC Platform] .

Conventional image compression algorithms include inter prediction and intra prediction, among which intra prediction uses high correlation between adjacent pictures of an image signal.

That is, 'intra prediction' predicts information of adjacent images in time and space in order to reconstruct the current image, and compresses / restores the current image closer to the original image using the prediction result.

Recently, H.264 was confirmed as a standard for high-definition video compression technology. Among them, intra prediction is adopted as a standard. However, when the compression algorithm is implemented according to such a standard, it takes a lot of computation time and a lot of hardware cost.

Conventional techniques for solving this problem include a fixed type intra prediction system as illustrated in FIG. 1, which will be described below.

As shown in FIG. 1, a conventional fixed type intra prediction system includes a memory 10, an intra prediction unit 11, and a prediction control unit 12. Here, the intra predictor 11 includes an input selector 111, a 16x16 luminance signal generator 112, a 4x4 luminance signal generator 113, a chromaticity signal generator 114, and an output selector 115, The prediction controller 12 includes a controller 121, a write address generator 122, a read address generator 123, and a write / read mode performer 124.

The input selector 11 receives the image data stored in the memory unit 10 and, based on the control of the prediction controller 12, receives the image data stored in the memory unit 10, or the 16x16 luminance signal generator 112, the 4x4 luminance signal generator 113, or the like. It is selectively output to the chroma signal generator (114).

Then, the 16x16 luminance signal generator 112, the 4x4 luminance signal generator 113, and the chromaticity signal generator 114 control the controller (FSM) 121 with respect to the data components input for each from the input selector 111. Intra-prediction is performed separately, and in particular, the 16x16 luminance signal generator 112 and the 4x4 luminance signal generator 113 may be operated in parallel.

That is, the 16x16 luminance signal generator 112 is configured with various preset prediction modes for the 16x16 luminance component (luminance signal) output from the input selector 111 under the control of the controller (FSM) 121 (see FIG. 7). Intra prediction is performed by selecting any one of the prediction modes (the corresponding intra prediction mode used in the image encoder). At the same time, the 4x4 luminance signal generator 113 is controlled by one of the various prediction modes (see FIG. 8) preset for the 4x4 luminance component output from the input selector 111 under the control of the controller (FSM) 121. Intra prediction is performed by selecting a prediction mode (a corresponding intra prediction mode used in the image encoder). The chroma signal generator 114 operates after the 16x16 luminance signal generator 112 and the 4x4 luminance signal generator 113 are operated in parallel, and is output from the input selector 111 under the control of the controller (FSM) 121. An intra prediction is performed by selecting one prediction mode (the corresponding intra prediction mode used in the image encoder) among various preset prediction modes with respect to the chromaticity component (chromatic signal).

Thereafter, the output selector 115 receives the prediction result from the 16x16 luminance signal generator 112, the 4x4 luminance signal generator 113, and the chromaticity signal generator 114, and then multiplexes and outputs the result according to the control of the controller 121. do.

Meanwhile, the prediction controller 12 writes the image data to the memory 10 or reads (writes) the image data stored in the memory 10 to the intra predictor 11 (write / read image data) and the intra predictor. Perform the function of controlling intra prediction in (11). Hereinafter, each component of the prediction controller 12 will be described.

The controller 121 performs an intra prediction process in the intra predictor 11, a write / read address generation process in the address generators 122 and 123, and an image data write / read process in the write / read mode performer 124. Perform the overall control process of the intra prediction system and so on.

The write address generator 122 or the read address generator 123 generates a write address or a read address under the control of the controller 121, and the write / read mode performer 124 is in control of the controller 121. Accordingly, in the write mode, the input image data is written to the memory 10 using the write address, and in the read mode, the image data stored in the memory is read and transferred to the input selector 111 using the read address.

The conventional intra prediction system as described above has the advantage of reducing the hardware cost and the execution time while maximizing the efficiency of the system through the parallel structure and the memory structure. In other words, by sharing one memory to perform 16x16 luma intra prediction, 4x4 luma intra prediction, and chroma intra prediction, the cost of the memory is reduced, and the processing speed of processing by processing 16x16 luma intra prediction and 4x4 luma intra prediction in parallel There is an advantage to improve.

However, in the conventional intra prediction system as described above, since the intra prediction is performed through the 'fixed memory read' which simply reads the prediction target block (decoding unit) to be decoded in the existing decoding order, it is unnecessary. There is a problem that the decoding processing performance is degraded due to the time for processing the data.

In other words, the conventional technique of the fixed method is difficult to implement a real-time video decoding in the video decoder because of the vast amount of operations due to the unnecessary amount of operation, and also increases the system size and power consumption due to the additional internal memory There is. In addition, the prior art also has a problem in that the use of the intra prediction algorithm is limited due to unnecessary computation amount depending on the type and application of the image.

The prior art as described above has a problem that due to the excessive amount of computation causes a decrease in intra prediction performance, an increase in power consumption, and a system miniaturization limit, and to solve such a problem is a problem of the present invention.

That is, according to the present invention, in performing the reading of the image data stored in the memory for image decoding, the reading rank is determined based on whether the corresponding prediction mode of the prediction target block corresponds to the horizontal / vertical mode, and the determined reading rank. Apparatus and method for adaptive video data reading control for intra prediction, which enables intra prediction to be performed quickly with a minimum amount of computation by reading the corresponding video data according to the method, and adaptive intra for video decoding using the same The purpose is to provide a prediction system.

The objects of the present invention are not limited to the above-mentioned objects, and other objects and advantages of the present invention which are not mentioned above can be understood by the following description, and will be more clearly understood by the embodiments of the present invention. Also, it will be readily appreciated that the objects and advantages of the present invention may be realized by the means and combinations thereof indicated in the claims.

In order to solve the above problem, the present invention determines a reading order based on which mode the prediction mode of the block to be predicted corresponds to 'horizontal / vertical mode' or 'non horizontal / vertical mode'. The image data may be read according to the determined reading rank.

More specifically, the present invention, in the adaptive image data read control device, for the prediction target blocks that are the target of intra prediction, the prediction mode is in the horizontal / vertical mode or non (horizontal / vertical mode) Control means for determining a 'priority of reading of reference data necessary for the intra prediction' based on which mode; Read address generation means for generating a read address for the reference data; And a read mode performing means for reading the image data corresponding to the read address generated by the read address generating means, according to the read rank determined by the control means.

In addition, according to the present invention, in the adaptive image data read control method, for a prediction target block that is an object of intra prediction, the prediction mode is set to any one of a horizontal / vertical mode and a non-horizontal / vertical mode. A ranking step of determining a 'read rank for reference data required for the intra prediction' based on the corresponding; An address generation step of generating a read address for the reference data; And a reading step of reading image data corresponding to the generated read address into an intra predicting unit according to the determined reading rank.

In addition, the present invention provides an adaptive intra prediction system for image decoding, comprising: storage means for separating and storing image data to be decoded into luminance and chroma components; Intra prediction means for performing intra prediction for image decoding on each of the luminance component and the chromaticity component of the stored image data; And controlling the process of writing (reading / reading image data) of the image data stored in the storage means or reading the image data stored in the storage means into the intra prediction means and performing the intra prediction, and predicting the prediction target blocks. And predictive control means for controlling the reading of the image data depending on whether the mode corresponds to the horizontal / vertical mode.

The present invention as described above, unlike the conventional intra prediction technology using a fixed memory read method, the amount of computation by adaptively determining the read rank (image data read rank) according to the prediction block size and the type of prediction mode Can be minimized, and the high performance intra prediction system with low power and low system area / size can be realized.

In addition, the present invention has a performance improvement of about 37% per macroblock compared to the conventional fixed method, and the improved performance enables real-time operation even on a large screen VGA (640x480) or D1 (720x480). It works.

The above objects, features and advantages will become more apparent from the following detailed description taken in conjunction with the accompanying drawings, whereby those skilled in the art may easily implement the technical idea of the present invention. There will be. In addition, in describing the present invention, when it is determined that the detailed description of the known technology related to the present invention may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted. Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

2 is a diagram illustrating an exemplary embodiment of an adaptive intra prediction system for image decoding according to the present invention.

Hereinafter, the adaptive image data read control apparatus for intra prediction according to the present invention (when the prediction controller 22 is related only to reading) and a method thereof will also be described.

As shown in FIG. 2, the adaptive intra prediction system according to the present invention includes a memory 20, an intra prediction unit 21, and a prediction controller 22.

First, the memory 20 and the intra predictor 21 will be described.

The memory 20 separates and stores the image data input from the outside into luminance and chroma components, and the intra predictor 21 decodes each of the luminance and chroma components of the image data stored in the memory 20. Intra prediction is performed for generating a 16x16 luminance signal, generating a 4x4 luminance signal, and generating a chroma signal.

Here, the intra predictor 21 includes an input selector 211, a 16x16 luminance signal generator 212, a 4x4 luminance signal generator 213, a chromaticity signal generator 214, and an output selector 215. The function of each component is as follows.

The input selector 211 receives the image data stored in the memory 20 under the control of the prediction controller 22, and receives the 16x16 luminance signal generator 212, the 4x4 luminance signal generator 213, or the chromaticity signal generator 214. Output to either

Then, the 16x16 luminance signal generator 212, the 4x4 luminance signal generator 213, and the chromaticity signal generator 214 separately perform intra prediction on the data components input to each from the input selector 211.

That is, the 16x16 luminance signal generator 212 is configured with various preset prediction modes for the 16x16 luminance component (luminance signal) output from the input selector 211 under the control of the controller (FSM) 221 (see FIG. 7). Intra prediction is performed by selecting any one of the prediction modes (the corresponding intra prediction mode used in the image encoder).

At the same time, the 4x4 luma signal generator 213 controls any one of various preset prediction modes (see FIG. 8) for the 4x4 luma component output from the input selector 211 under the control of the controller (FSM) 221. Intra prediction is performed by selecting a prediction mode (a corresponding intra prediction mode used in the image encoder).

The chromaticity signal generator 214 is a prediction mode (image encoder) of any one of various preset prediction modes for the chromaticity component (chromatic signal) output from the input selector 211 under the control of the controller (FSM) 221. Intra prediction is performed by selecting a corresponding intra prediction mode used in the Rx.

Thereafter, the output selector 215 receives the prediction result from the 16x16 luminance signal generator 212, the 4x4 luminance signal generator 213, and the chromaticity signal generator 214, and then, according to the control of the controller (FSM) 221. Output multiplexed.

Next, the prediction controller 22 will be described in detail.

The prediction controller 22 controls the process of writing image data to the memory 20 or reading image data stored in the memory 20 into the intra prediction unit 21 (writing / reading of image data) and also intra prediction unit 21. Control intra prediction, and control image data reading according to whether the prediction mode for the prediction target blocks corresponds to the horizontal / vertical mode or the non-horizontal / vertical mode. The controller 221 includes a controller 221, a write address generator 222, a read address generator 223, and a write / read mode performer 224. Each component will be described below. .

The controller 221 performs overall control of the intra prediction system, such as writing / reading image data and intra prediction, and in particular, the prediction mode is horizontal / vertical mode or non-horizontal for the prediction target blocks to be predicted. Based on which of the two vertical modes is used, the read rank of the reference data required for the intra prediction is determined. Here, whether the corresponding prediction mode corresponds to a horizontal / vertical mode or a non-horizontal / vertical mode may be checked through the 'prediction mode used in the image encoder'. Hereinafter, the reading rank determination in the controller 221 will be described in detail.

First, when the prediction target block is a block for the 'luminance component', priority is given to a block having a large size of the prediction target block, and blocks corresponding to the horizontal / vertical mode among the prediction target blocks having the same size. Give priority to. That is, the read rank determination process when the 16x16 luminance prediction mode and the 4x4 luminance prediction is performed in the intra predictor 21 gives priority to the 16x16 luminance prediction mode block over the 4x4 luminance prediction mode block, and has the same luminance. The prediction mode blocks may be given priority to blocks corresponding to the horizontal / vertical mode over blocks corresponding to the non-horizontal / vertical mode.

Next, when the prediction target block is a block for chromatic components (8x8 chroma prediction mode block), the read ranking process may include a non-horizontal / vertical mode for a block whose prediction mode corresponds to a horizontal / vertical mode. It is characterized in that the priority is given to the block corresponding to the.

In the write mode, on the other hand, the write address generator 222 is driven to generate a write address required for writing the input image data into the memory 20 under the control of the controller 221. In the read mode, the read address generator 223 is driven to generate a read address for 'reference data required for the intra prediction' under the control of the controller 221.

When the block to be predicted is a block for the luminance component, the read address generator 223 may include a 16x16 horizontal / vertical mode address 2231, a 16x16 non-horizontal / vertical mode address 2232, and 4x4 horizontal / Generated by dividing into a vertical mode address 2233 or a 4x4 non-horizontal / vertical mode address 2234, and an 8x8 horizontal / vertical mode address 2235 when the prediction target block is a block for 'chromatic components'. Or 8x8 non-horizontal / vertical mode addresses (2236).

Then, the write / read mode performing unit 224 may select one of the write / read modes under the control of the controller 221 using the address generated by the write address generator 222 or the read address generator 223. In particular, the reading of the image data is performed according to the reading order determined by the controller 221.

For example, when performing reading on the luminance component, the write / read mode performing unit 224 may execute a 16x16 horizontal / vertical mode address → 16x16 non-horizontal / vertical mode address → 4x4 horizontal / vertical mode address → 4x4. The image data is read in the order of non-horizontal / vertical mode addresses, but after the reading of the priorities is finished, the reading of the next ranking is started. When reading chroma components, the corresponding image data is read in the order of 8x8 horizontal / vertical mode address → 8x8 non-horizontal / vertical mode address.

3 is a flowchart illustrating an adaptive image data read control method for intra prediction according to an exemplary embodiment of the present invention, and illustrates a method of determining a read rank of a luminance signal.

First, it is checked whether the prediction mode used in the image encoder is the 16x16 luminance mode for the prediction target block (decoding unit) that is the target of intra prediction for image decoding (300). Check the recognition (302), generate the corresponding horizontal / vertical address with the read rank as the first priority for the horizontal / vertical prediction mode (304), and remove the read rank for the non-horizontal / vertical prediction mode (Non). The horizontal / vertical address is generated as the second priority (306). Here, there is no right or lower order of read rank between the vertical prediction mode (mode 0) and the vertical prediction mode (mode 1), which may be performed first. In addition, the non-horizontal / vertical prediction modes include a DC prediction mode (mode 2) and a plane prediction mode (mode 4).

On the other hand, even when the prediction mode is the 4x4 luminance mode (300), as in the case of the 16x16 luminance mode, it is again checked whether the horizontal / vertical mode (308), the read rank for the horizontal / vertical mode as the third priority A corresponding horizontal / vertical address is generated (310), and a non-horizontal / vertical mode generates a corresponding horizontal / vertical address with a read rank as a fourth priority (312).

Here, there is no right or lower order of read rank between the vertical prediction mode (mode 0) and the vertical prediction mode (mode 1), which may be performed first. Non-horizontal / vertical prediction modes also include DC prediction mode (mode 2), diagonal down-left prediction (mode 3), diagonal down-right prediction mode (mode 4), and horizontal-left prediction mode (mode 5). , A horizontal-down prediction mode (mode 6), a vertical-right prediction mode (mode 7), and a horizontal-up prediction mode (mode 8), and there is no superiority of read rank between these modes.

That is, in the 4x4 luminance mode, there are nine operation modes (prediction mode). Among these, the horizontal / vertical prediction mode is given priority in the read order.

In short, the present invention provides (1) processing (reading) 16x16 luminance mode in preference to 4x4 luminance mode so as to increase the efficiency of intra prediction, and (2) horizontal / vertical in the same 16x16 luminance mode or 4x4 luminance mode. It is characterized in that the mode is processed (read) in preference to the non-horizontal / vertical mode.

As described above, the reason why the reading of the horizontal / vertical prediction mode is preferentially performed is that the horizontal-vertical prediction mode occupies more than 50% of the total when the simulation is performed with the 'Foreman CIF' image. To deal with.

As described above, the method of determining the reading rank of the luminance component has been described with reference to FIG. 3, but the same method is applied to the chromaticity component.

4 is an explanatory diagram of a cycle gain when the adaptive intra prediction according to the present invention is applied.

Looking at the memory write / read for the 4x4 luminance component, it can be seen that a total of 241 cycles are performed in the case of using the conventional fixed method (400), but a total of 153 cycles are performed in the case of using the present invention (410).

Specifically, when writing the image data of the 4x4 luminance component to the memories 10 and 20, 33 cycles are performed in the conventional method or the present invention.However, while reading the image data (4x4 luminance component) stored in the memory, the intra When inputting to the predictors 11 and 21, about 208 cycles 401 are performed when the conventional fixed method is used. However, in the present invention, image data is read according to a specially set reading order. 411 is performed. The reason for this is that in the prior art, since the fixed method is used, all image data must be read sequentially, whereas in the present invention, the read cycle is reduced in the case of 4x4 because the read order is read out because of the high priority. Compared with the prior art and the present invention, it can be seen that on average, there is an improvement of about 37% (reduction from 241 cycles to 153 cycles).

5 is a definition diagram of a sample for intra spatial prediction applied to the present invention, and shows intra prediction for a 4x4 block.

a to p (50) are samples included in a 4x4 block, and A to Q are samples neighboring the block. Also, samples A to Q must already be decoded for prediction, which is used as reference data.

If the sample is outside of the picture or outside of the current slice, if one macroblock is included in the current macroblock in successive raster scan order, the block of motion compensation (16x16 to 4x4 blocks) In the case of being sent as if related to the current 4x4 block in order, intra prediction is not performed when the intra macroblock is not or the value of the constraint intra prediction value (constrained_intra_pred) is "1". That is, under the conditions listed above, no sample takes into account A-Q.

If the sample data E ~ H are not useful (not decoded yet), the E ~ H sample values are replaced with the sample values of D. Also, if sample data M to P is not useful (if not yet decoded), then the M to P sample values are replaced with L sample values.

FIG. 6 is an explanatory diagram of directions of intra prediction modes applied to the present invention and shows a prediction mode of 4 × 4 luma components.

That is, the arrow directions in FIG. 6 include Mode 0 (vertical prediction), Mode 1 (horizontal prediction), Mode 2 (DC), Mode 3 (diagonal down-left prediction), Mode 4 (diagonal down-right prediction), The prediction directions of the respective modes such as mode 5 (horizontal-left prediction), mode 6 (horizontal-down prediction), mode 7 (horizontal-right prediction), mode 8 (horizontal-up prediction), and the like are shown.

7 is a diagram illustrating a prediction operation for a 4x4 luminance prediction mode applied to the present invention, and illustrates an intra prediction process performed by the 4x4 luminance signal generator 213.

As for the luminance signal, a total of nine intra predictions are performed, such as mode 0, mode 1, mode 2, mode 3, mode 4, mode 5, mode 6, mode 7, and mode 8, Where mode 2 is DC prediction. Hereinafter, each mode will be described.

[ mode  0]: Vertical Prediction

Mode 0 is performed as follows when A, B, C, and D are valid.

      a, e, i, m are predicted as A

      b, f, j, n are predicted to be B

      c, g, k, o are predicted by C

      d, h, l, p are predicted by D

[ mode  1]: Horizontal Prediction

 In mode 1, when I, J, K, and L are valid, prediction is performed as follows.

     a, b, c, d are predicted by I

     e, f, g, h are predicted by J

     i, j, k, l are predicted by K

     m, n, o, p are predicted by L

[ mode  2]: DC prediction

First, if A, B, C, D, I, J, K, L are all valid, all samples are (A + B + C + D + I + J + K + L + 4) >> 3 It is predicted. Second, if A, B, C, and D are not valid and I, J, K, and L are valid, then (I + J + K + L + 2) >> 2 is expected. Third, if I, J, K, and L are not valid, and A, B, C, and D are valid, all samples are predicted to be (A + B + C + D + 2) >> 2. Fourth, if all eight samples are invalid, the value predicted in the 4x4 block is set to 128.

[ mode  3]: Diagonal down-left prediction Down-left  Prediction)

In this mode, when A, B, C, D, I, J, K, L, and Q values are valid, prediction is performed in the following manner. This is called diagonal prediction.

    -a is predicted by (A + 2B + C + I + 2J + K + 4) >> 3

    -b, e are predicted by (B + 2C + D + J + 2K + L + 4) >> 3

    -c, f, i are predicted by (C + 2D + E + K + 2L + M + 4) >> 3

    -d, g, j, m are predicted by (D + 2E + F + L + 2M + N + 4) >> 3

    -h, k, n are predicted by (E + 2F + G + M + 2N + O + 4) >> 3

    -l, o are predicted by (F + 2G + H + N + 2O + P + 4) >> 3

    -p is predicted by (G + H + O + P + 2) >> 2

Where >> represents the right shift operator.

[ mode  4]: Diagonal down-right prediction Down-right  Prediction)

This mode performs 'diagonal prediction' in the following manner when A, B, C, D, I, J, K, L and Q values are valid.

     -m is predicted by (J + 2K + L + 2) >> 2

     i, n are predicted by (I + 2J + K + 2) >> 2

     e, j, o are predicted by (Q + 2I + J + 2) >> 2

     -a, f, k, p are predicted by (A + 2Q + I + 2) >> 2

     -b, g, l are predicted by (Q + 2A + B + 2) >> 2

     -c, h are predicted by (A + 2B + C + 2) >> 2

     -d is predicted by (B + 2C + D + 2) >> 2

[ mode  5]: Vertical-Right Prediction

This mode is used when A, B, C, D, I, J, K, L, and Q are valid, and the prediction is performed as follows.

    -a is predicted by (2A + 2B + J + 2K + L + 4) >> 3

    -b, i are predicted by (B + C + 1) >> 1

    -c, j are predicted by (C + D + 1) >> 1

    -d, k are predicted by (D + E + 1) >> 1

    -l is predicted by (E + F + 1) >> 1

    -e is predicted by (A + 2B + C + K + 2L + M + 4) >> 3

    -f, m are predicted by (B + 2C + D + 2) >> 2

    -g, n are predicted by (C + 2D + E + 2) >> 2

    -h, o are predicted by (D + 2E + F + 2) >> 2

    -p is predicted by (E + 2F + G + 2) >> 2

[ mode  6]: Horizontal-Down Prediction

This mode performs 'diagonal prediction' in the following manner when A, B, C, D, I, J, K, L, and Q are valid.

     -a, g are predicted by (Q + I + 1) >> 1

     -b, h are predicted by (I + 2Q + A + 2) >> 2

     -c is predicted by (Q + 2A + B + 2) >> 2

     -d is predicted by (A + 2B + C + 2) >> 2

    -e, k are predicted by (I + J + 1) >> 1

    -f, l are predicted by (Q + 2I + J + 2) >> 2

    -i, o are predicted by (J + K + 1) >> 1

    -j, p are predicted by (I + 2J + K + 2) >> 2

    -m is predicted by (K + L + 1) >> 1

    -n is predicted by (J + 2K + L + 2) >> 2

[ mode  7]: Vertical-Left Prediction

This mode performs 'diagonal prediction' in the following manner when A, B, C, D, I, J, K, L, and Q are valid inside the slice.

     -a, j are predicted by (Q + A + 1) >> 1

     -b, k are predicted by (A + B + 1) >> 1

     -c, l are predicted by (B + C + 1) >> 1

     -d is predicted by (C + D + 1) >> 1

     e, n are predicted by (I + 2Q + A + 2) >> 2

     -f, o are predicted by (Q + 2A + B + 2) >> 2

     -g, p are predicted by (A + 2B + C + 2) >> 2

     -h is predicted by (B + 2C + D + 2) >> 2

     -i is predicted by (Q + 2I + J + 2) >> 2

     -m is predicted by (I + 2J + K + 2) >> 2

[ mode  8]: Horizontal-Up Prediction

This mode is used when A, B, C, D, I, J, K, L, and Q are valid and performs prediction as follows.

    -a is predicted by (B + 2C + D + 2I + 2J + 4) >> 3

    -b is predicted by (C + 2D + E + I + 2J + K + 4) >> 3

    -c, e are predicted by (J + K + 1) >> 1

    -d, f are predicted by (J + 2K + L + 2) >> 2

    -g, i are predicted by (K + L + 1) >> 1

    -h, j are predicted by (K + 2L + M + 2) >> 2

    -l, n are predicted by (L + 2M + N + 2) >> 2

    -k, m are predicted by (L + M + 1) >> 1

    -o is predicted by (M + N + 1) >> 1

    -p is predicted by (M + 2N + O + 2) >> 2

8 is a diagram illustrating a prediction operation for a 16x16 luminance prediction mode applied to the present invention, and illustrates an intra prediction process performed by the 16x16 luminance signal generator 212.

In the 16x16 luminance prediction mode, unlike the 4x4 luminance prediction mode (see FIG. 7), the 16x16 luminance component of the entire macroblock may be predicted in one operation, and there are four modes.

16x16 luma prediction performs each prediction mode using 0-15 samples horizontally and 0-15 sample values vertically. P (x, y) (or Pred (x, y)) can be expressed as x = 0, ..., 15 in the horizontal direction, and y = 0, in the vertical direction. Mark the position with ..., 15. The sample for P (x, -1), x = 0, ..., 15 refers to the neighboring upper block of the block, and if P (-1, y), y = 0, ..., 15 Predicts the neighboring left block. The sample P (x, -1) or P (-1, y) can be: ① If the sample is outside the picture or not the current slice, ② the sample is a non-intra macroblock and the value of 'contrained_intra_pred' is “ 1 ”is not considered

[ mode  0]: Vertical Prediction

Mode 0 is a case where all neighboring samples P (x, -1) (H) are valid, and an intra prediction value is obtained by Equation 1 below.

Pred (x, y) = P (x, -1), x = 0, ..., 15

[ mode  1]: Horizontal Prediction

Mode 1 is a case where all neighboring samples P (-1, y) (V) are valid, and an intra prediction value is obtained by Equation 2 below.

Pred (x, y) = P (-1, y), y = 0, ..., 15

[ mode  2]: DC prediction

Mode 2 is a case where all neighboring samples H and V are valid, and a prediction value is obtained as shown in Equation 3 below.

If the sample P (x, -1) (H) is not valid and the neighboring sample P (-1, y) (V) is valid, the prediction of all luminance samples is as shown in [Equation 4]. Is performed.

On the contrary, if the sample P (-1, y) is not valid and the neighboring sample P (x, -1) is valid, the prediction of all luminance samples is performed as shown in Equation 5 below.

If both neighbor samples P (x, -1) and P (-1, y) are not valid, the prediction value for the luminance samples is set to " 128 ".

[ mode  3]: Plain  Plane Prediction

Mode 3 is performed as shown in Equation 6 below if all neighboring samples P (x, -1) and P (-1, y) are valid.

Here, H (top sample) and V (left sample) are defined as shown in Equation 7 below.

Hereinafter, 'intra coding prediction for chroma blocks' performed by the chroma signal generator 214 will be described.

The processing method for chromaticity blocks for intra macroblocks is very similar to the luminance block processing method for an intra 16x16 macroblock using four prediction modes. The same prediction mode is applied to the two chroma blocks. But it is applied independently of what was applied to luminance.

At 8x8 luminance, the prediction mode is performed by using the values of the samples 0-7 vertically and the samples 0-7 horizontally. Looking at the method of expressing P (x, y), the position is represented by x = 0, ..., 7 in the horizontal direction and the position is represented by y = 0, ..., 7 in the vertical direction. The sample for P (x, -1), x = 0, ..., 7 refers to the neighboring top block of the block, where P (-1, y), y = 0, ..., 7 Refers to the neighboring left block for prediction.

Sample P (x, -1) or P (-1, y) is not considered in the following cases. That is, if (1) the sample is outside the picture or is not the current slice, (2) the sample is a non-intra macroblock, the value of the constrained intra prediction (contrained_intra_pred) is "1", and P (x, y) is invalid. If not, set it to "128".

[ mode  0]: vertical prediction

 Mode 0 is predicted by the following Equation 8.

[ mode  1]: Horizontal Prediction

Mode 1 is predicted by Equation 9 below.

[ mode  2]: DC prediction

If all samples, ie, P (-1, n) and P (n, -1), are valid, are represented by Equation 10 below.

If eight samples of P (-1, n) are not valid, the following equation (11) is used.

If eight samples P (n, -1) are not valid, the following equation (12) is predicted.

If all 16 sample values are not valid, the predicted Pred (x, y) value is equal to all samples x, y = 0. Has a value of 128 for 7

[ mode  3]: Plain  Plane Prediction

In the plane mode, it is predicted as in Equation 13 below.

Here, H (top sample) and V (left sample) are as shown in Equation 14 below.

As described above, the method of the present invention may be implemented as a program and stored in a recording medium (CD-ROM, RAM, ROM, floppy disk, hard disk, magneto-optical disk, etc.) in a computer-readable form. Since this process can be easily implemented by those skilled in the art will not be described in more detail.

The present invention described above is capable of various substitutions, modifications, and changes without departing from the technical spirit of the present invention for those skilled in the art to which the present invention pertains. It is not limited by the drawings.

1 is a block diagram of a conventional fixed prediction system;

2 is a block diagram of an embodiment of an adaptive intra prediction system for image decoding according to the present invention;

3 is a flow chart of an embodiment of an adaptive image data read control method for intra prediction according to the present invention;

4 is an explanatory diagram of a cycle gain in the case of applying the adaptive intra prediction according to the present invention;

5 is a definition diagram of a sample for intra spatial prediction applied to the present invention;

6 is an explanatory diagram of directions of intra prediction modes applied to the present invention;

7 is a prediction operation diagram for the 4x4 luminance prediction mode applied to the present invention,

8 is a diagram illustrating a prediction operation for a 16x16 luminance prediction mode applied to the present invention.

* Explanation of symbols for the main parts of the drawings

20: memory 21: intra prediction unit

22: prediction controller 221: controller

222: write address generator 223: read address generator

224: write / read mode execution unit

Claims (18)

In the adaptive image data read control device, For the blocks to be predicted for intra prediction, the reading order of the reference data required for the intra prediction based on whether the prediction mode corresponds to a horizontal / vertical mode or a non-horizontal / vertical mode. Control means for determining '; Read address generation means for generating a read address for the reference data; And Means for performing read mode for reading image data corresponding to the read address generated by said read address generation means according to the read rank determined by said control means Adaptive image data read control device comprising a. The method of claim 1, The control means, In determining the read rank, when the prediction target block is a block for a luminance component, a high priority is given to a block having a large size of the prediction target block, and the horizontal / And a block having a vertical mode higher priority than a block corresponding to the non-horizontal / vertical mode. The method of claim 2, The control means, When the intra predictor performs 16x16 luma prediction and 4x4 luma prediction, the 16x16 luma prediction mode block is given a higher priority than the 4x4 luma prediction mode block, and the luminance prediction mode blocks of the same size correspond to the horizontal / vertical mode. Adaptive image data read control device, characterized in that to give a higher priority to the block than the non-horizontal / vertical mode block. The method of claim 3, wherein The address generating means, If the prediction target block is a block for a luminance component, a 16x16 horizontal / vertical mode address, a 16x16 non-horizontal / vertical mode address, a 4x4 horizontal / vertical mode address, or a 4x4 non-horizontal / vertical mode address. Adaptive image data read and control device, characterized in that for generating separately. The method of claim 4, wherein The read mode performing means, The image data is read in the order of the 16x16 horizontal / vertical mode address, the 16x16 non-horizontal / vertical mode address, the 4x4 horizontal / vertical mode address, and the 4x4 non-horizontal / vertical mode address. Wherein, after the reading of the priority is finished, the adaptive image data read control device, characterized in that to start reading for the next rank. The method of claim 1, The control means, In determining the read rank, if the prediction target block is a block for chromatic components, the corresponding prediction mode is higher than a block corresponding to the non-horizontal / vertical mode for a block corresponding to the horizontal / vertical mode. Adaptive image data read control device, characterized in that to give a high priority. The method of claim 6, The address generating means, And when the block to be predicted is a block for chromatic components, generate the image by dividing it into an 8x8 horizontal / vertical mode address and an 8x8 non-horizontal / vertical mode address. The method of claim 7, wherein The read mode performing means, The image data is read in the order of the 8x8 horizontal / vertical mode address and the 8x8 non-horizontal / vertical mode address, but the reading of the next rank is started after the reading of the priority is finished. Adaptive image data read control device. The method according to any one of claims 1 to 8, The control means, And determining whether the prediction block corresponds to the horizontal / vertical mode or the non-horizontal / vertical mode through the prediction mode used in the image encoder. In the adaptive image data read control method, For the blocks to be predicted for intra prediction, the reading order of the reference data required for the intra prediction based on whether the prediction mode corresponds to a horizontal / vertical mode or a non-horizontal / vertical mode. Ranking step of determining '; An address generation step of generating a read address for the reference data; And A reading step of reading image data corresponding to the generated read address into an intra predicting unit according to the determined reading order; Adaptive image data read control method comprising a. The method of claim 10, The ranking step, When the prediction target block is a block for a luminance component, a high priority is given to a block having a large size of the prediction target block, and a block corresponding to the horizontal / vertical mode is included among prediction target blocks having the same size. Adaptive image data read control method characterized by giving a high priority. The method of claim 11, wherein The reading step, 16x16 horizontal / vertical mode address, 16x16 non horizontal / vertical mode address, 4x4 horizontal / vertical mode address, and 4x4 non horizontal / vertical mode address. Adaptive image data read control method. The method according to any one of claims 10 to 12, The ranking step, If the prediction target block is a block for chromatic components, the corresponding prediction mode gives higher priority to a block corresponding to the horizontal / vertical mode than a block corresponding to the non-horizontal / vertical mode. Adaptive image data read control method. The method of claim 13, The reading step, An adaptive image data read control method according to claim 8, wherein the image data is read in the order of an 8x8 horizontal / vertical mode address and an 8x8 non-horizontal / vertical mode address. In the adaptive intra prediction system for image decoding, Storage means for separating and storing the image data to be decoded into a luminance component and a chromaticity component; Intra prediction means for performing intra prediction for image decoding on each of the luminance component and the chromaticity component of the stored image data; And While controlling the process of writing the image data to the storage means or reading the image data stored in the storage means to the intra prediction means (write / read image data) and the process of performing the intra prediction, the prediction mode for the prediction target blocks Predictive control means for controlling the reading of the image data depending on whether the image corresponds to the horizontal / vertical mode Adaptive intra prediction system for image decoding comprising a. The method of claim 15, The prediction control means, The process of writing / reading the image data and performing the intra prediction is performed, and the prediction mode corresponds to any one of the horizontal / vertical mode or the non-horizontal / vertical mode for the prediction target blocks that are the targets of intra prediction. Control means for determining a 'priority of reading of reference data necessary for the intra prediction' on the basis of the information; Address generating means for generating a write address for writing the image data or a read address for the reference data; And Means for performing any one of write / read modes under the control of the control means by using the address generated by the address generating means, wherein reading of the image data is performed according to the determined reading order; Adaptive intra prediction system for image decoding comprising a. The method of claim 16, The control means, In determining the read rank, when the prediction target block is a block for a luminance component, a high priority is given to a block having a large size of the prediction target block, and the horizontal / Adaptive intra prediction system for image decoding, characterized in that high priority is given to blocks corresponding to the vertical mode. With respect to claim 16 or 17, The control means, In determining the read rank, if the prediction target block is a block for chromatic components, the corresponding prediction mode is higher than a block corresponding to the non-horizontal / vertical mode for a block corresponding to the horizontal / vertical mode. Adaptive intra prediction system for image decoding characterized by giving a high priority.

Images