JP2014027371A - Intra-prediction processor and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform determination processing in a prediction direction at high speed without deteriorating prediction efficiency and to reduce processing time and power consumption even if there are a plurality of prediction directions.SOLUTION: An intra-prediction processor 1 determines an optimum prediction direction by using a character of prediction efficiency, in which prediction efficiency in a direction orthogonal to one optimum direction is remarkably low and the prediction efficiency becomes high as a direction is nearer to the one optimum direction. Specifically, a comparing section 3 of an optimum prediction direction determining section 2 calculates an evaluation value on two orthogonal directions, compares two evaluation values and specifies the direction where the evaluation value is larger. A comparing section 4 selects a direction region according to the prediction direction specified by the comparing section 3, calculates the evaluation value in a direction where the evaluation value is not calculated in two directions where correlation in the direction region is the lowest, compares the two evaluation values, specifies the direction where the evaluation value is larger, narrows down the direction by repeating such processing and determines the optimum prediction direction.

Description

  The present invention relates to an intra prediction processing apparatus and program for determining an efficient prediction direction at a high speed from a plurality of prediction directions when performing intra prediction (intra prediction) using a plurality of prediction directions.

  In recent years, with the advance of multimedia, there are many opportunities to handle moving images. Image data has a large amount of information, and moving image data, which is a set of images, has an information amount exceeding 1 Gbit / sec in high-definition (registered trademark) (1920 samples × 1080 lines) broadcasting, which is the mainstream of television broadcasting. Therefore, the moving image data is H.264. Data is compressed and transmitted and stored by an encoding standard technique represented by H.264 / AVC (Advanced Video Coding), HEVC (High Efficiency Video Coding) being standardized, and the like.

  The HEVC standard does not simply divide a screen from the upper left into blocks that are encoding units as in the conventional encoding technique, but newly adopts hierarchical division to enable block division of a plurality of hierarchies. A coding unit divided into a plurality of blocks is called a CU (Coding Unit) or a coding block.

  FIG. 10 is a diagram illustrating an example of block division according to the HEVC standard. As shown in FIG. 10, in the HEVC standard, it is possible to divide an image into pixel sizes such as 4 × 4, 8 × 8, 16 × 16, and 32 × 32 and perform encoding. Further, in the HEVC standard, in order to efficiently express the prediction error signal, the CU can be divided into layers and divided into conversion units. This transform unit is called a TU (Transform Unit) or transform block.

  FIG. 11 is a diagram illustrating TU hierarchy division. As shown in FIG. 11, for example, a TU with 0 layer division for a 32 × 32 CU is the same block as a CU with 32 × 32 pixels, and a TU with 1 layer division is 16 × 16 pixels. It is a divided block. In addition, a TU with two hierarchical divisions is a block divided into 8 × 8, and a TU with three hierarchical divisions is a block divided into 4 × 4 pixels. Since the TU is divided into layers according to the size of the CU, the size of the TU in each layer is different according to the size of the CU.

  In the HEVC standard, a CU is divided into prediction units that are a plurality of types of rectangular areas, and prediction processing is performed in each prediction unit. This prediction unit is referred to as a PU (Prediction Unit) or a prediction block.

  Here, in the intra-frame coding (intra coding) processing for the divided blocks, spatial signal prediction within the screen is performed. This technique is called intra prediction (intra prediction). There are a plurality of methods for intra prediction that are being studied in the HEVC standard.

  As one method of a plurality of intra predictions, a technique for predicting a signal of an encoded block by referring to an already encoded and decoded signal located around the encoded block to be encoded It has been known. The encoder uses a signal known by the decoder as a signal used for prediction. Therefore, the encoder includes a local decoder that performs the same processing as the decoder, and sequentially decodes the encoded signals.

  Therefore, the signal located around the encoding block to be encoded has already been decoded by the local decoder. In the case of the HEVC standard, as in the conventional encoding process, the blocks located on the left and upper side of the encoding block to be encoded have already been decoded, and the encoded block signal is converted using this decoded signal. Predict.

  The basic principle of this prediction is H.264. Signal prediction is performed by a method similar to the intra prediction defined in H.264 / AVC (see Non-Patent Document 1). In the HEVC standard, the signal prediction direction is H.264. This is an increase over the H.264 / AVC standard, and signal prediction in 33 directions is performed at the maximum.

  FIG. 12 is a diagram illustrating an intra prediction mode in the HEVC standard. As shown in FIG. 12, the prediction directions of intra prediction are 33 directions of indexes 2-34. The encoder uniquely evaluates a direction with high prediction efficiency from these prediction directions, encodes data indicating the prediction direction, and transmits the code to the decoder. As a result, the decoder can efficiently generate a prediction signal using the decoded neighboring pixels based on the received code.

  In order to determine which direction has the highest prediction efficiency in the process of determining the prediction direction of this intra prediction, in general, encoding is actually performed in each prediction direction comprehensively, or the amount of processing In order to reduce the above, encoding is performed virtually and the encoding performance is compared. Note that the method for evaluating the prediction efficiency from each direction and the method for the order are not defined in the encoding standard, and are left to the discretion of the developer of the encoding device. This is an important part where performance is affected.

  In the HEVC standard intra prediction in the coding model used in the HEVC standardization work stage, all prediction directions shown in FIG. 12 are evaluated in the order of the indexes, and the direction with the highest prediction efficiency is determined. . Specifically, starting from plane prediction (Planar) of index 0 and DC prediction (DC) of index 1, from the lower left direction of index 2 to the horizontal left direction of index 10, and to the upper left direction of index 18, The prediction efficiency in each direction is evaluated in the order of the vertical upward direction and the upper right direction of the index 34, and one direction having the highest prediction efficiency is determined from the evaluation efficiency. The index shown in FIG. 12 is attached to the intra prediction mode in the HEVC standardization work stage. In the official HEVC standard after the standardization work, the contents of the index may be changed. . The same applies to FIG. 2 described later.

  In addition, as an example of processing for determining the direction of intra prediction, a technique for reducing the amount of calculation has been disclosed (see, for example, Patent Document 1). In this technique, H.C. In the H.264 intra prediction, evaluation values (sum of absolute values of pixel difference values between the predicted image and the input image: SAD) when the prediction direction is vertical and horizontal are respectively calculated, and the evaluation value is small. A direction is specified, an evaluation value in a direction near the specified direction is calculated, and a direction having a smaller evaluation value is specified. Then, by repeating such comparison of the evaluation values, the prediction direction is narrowed down and the direction with the highest prediction efficiency is determined.

JP 2009-177357 A

Supervised by Satoshi Okubo, “Revised Third Edition H.264 / AVC Textbook”, Impress R & D, p.264-268, January 1, 2009

  However, in the intra prediction at the above-described HEVC standardization work stage, the prediction efficiency is evaluated in all directions. Therefore, if the prediction efficiency in the vertical direction is high, the evaluation of the prediction efficiency in the horizontal direction is useless. Further, since evaluation in all directions is necessary, an increase in power consumption is caused.

  Moreover, in the intra prediction described in the above-mentioned Patent Document 1, the direction is specified based on the evaluation of the prediction efficiency, and the direction is narrowed down by evaluating the direction near the specified direction. Compared to the evaluation, the amount of calculation can be reduced. For example, according to the example of Patent Document 1, when the horizontal prediction direction is specified as a result of comparing the evaluation values in the vertical prediction direction (prediction mode 0) and the horizontal prediction direction (prediction mode 1), The prediction directions to be compared with each other are the horizontal prediction direction, the horizontal downward direction (prediction mode 6), and the horizontal upward direction (prediction mode 8) (Patent Document 1, paragraphs 31 to 33, FIGS. 5 and 7-1). .

  However, this technique is described in H.C. This method is applied when the number of prediction directions is 8 in the H.264 method, and when the number of prediction directions is large as in the HEVC method, the processing is not always efficient. For example, as illustrated in FIG. 12, a case is assumed where the number of intra prediction directions is 33. As described above, when the horizontal prediction direction is specified as a result of comparing the evaluation values in the vertical prediction direction (the direction of the index 26) and the horizontal prediction direction (the direction of the index 10), the next comparison is performed. Although the prediction direction to perform is not clear from Patent Document 1, it is considered to be 17 directions from the lower left direction (the direction of index 2) to the upper left direction (the direction of index 18).

  This eliminates the need for evaluation in all directions, but it cannot be said that the amount of calculation can be reduced sufficiently. In this case, it is desirable that the amount of calculation can be sufficiently reduced and efficient processing can be realized even when the number of prediction directions is large as in the HEVC method.

  Therefore, the present invention has been made to solve the above problems, and the purpose thereof is to perform a prediction direction determination process at high speed without lowering the prediction efficiency even when the prediction directions are many. An object of the present invention is to provide an intra prediction processing apparatus and program capable of reducing processing time and power consumption.

  To achieve the above object, the intra prediction processing apparatus according to claim 1, wherein an image is divided into a plurality of blocks, and an intra prediction process for determining a prediction direction of intra prediction when performing intra coding on the block. In the apparatus, the prediction efficiency is evaluated for each of the two directions having the lowest correlation included in the predetermined direction area, the direction area including the direction having the high prediction efficiency is selected, and the correlation included in the selected direction area is the lowest. Specifying two directions, evaluating the prediction efficiency in the two directions, selecting a direction region including a direction with high prediction efficiency, specifying the two directions, processing for evaluating the prediction efficiency, and the direction region The optimum prediction direction determination unit that narrows down the direction with the highest prediction efficiency and determines the prediction direction with the highest prediction efficiency among the plurality of prediction directions. Characterized by comprising.

  The intra prediction processing apparatus according to claim 2 is the intra prediction processing apparatus according to claim 1, wherein the optimum prediction direction determining unit evaluates prediction efficiency for two orthogonal directions among the plurality of prediction directions. In accordance with the comparison result obtained by the first comparison unit and the first comparison unit, the direction region including the direction with high prediction efficiency is selected, and the two directions with the lowest correlation included in the selected direction region are selected. A second comparison unit that evaluates and compares the respective prediction efficiencies, and the second comparison unit further includes a direction including a direction in which the prediction efficiency is high according to a comparison result by the second comparison unit. A region is selected, and the process of evaluating and comparing the prediction efficiencies of the two directions with the lowest correlation included in the selected direction region is repeated, and the most predictive of the plurality of prediction directions is repeated. Determining a high prediction direction efficiency, characterized in that.

  The intra prediction processing device according to claim 3 is the intra prediction processing device according to claim 2, wherein the processing in the two directions in the first comparison unit or the second comparison unit is performed in the processing. It is performed recursively.

  A program according to claim 4 causes a computer to function as the intra-prediction processing device according to any one of claims 1 to 3.

  As described above, according to the present invention, even when there are a large number of intra prediction directions, prediction direction determination processing can be performed at high speed without lowering prediction efficiency, and processing time and power consumption can be reduced. Can be reduced.

It is a block diagram which shows schematic structure of the intra prediction processing apparatus by embodiment of this invention. It is a figure explaining the direction of intra prediction mode. It is a flowchart which shows the outline of a process of the optimal prediction direction determination part. It is a flowchart which shows the detail of a process of the optimal prediction direction determination part. It is an example of the program list which showed the process of the optimal prediction direction determination part in the format of C language. It is an example of the program list which showed the orthogonal direction comparison process ((alpha)) in the format of C language. It is an example of the program list which showed narrowing-down comparison processing (β) in a C language format. It is a block diagram which shows the structure of an image coding apparatus. It is a block diagram which shows the structure of an image decoding apparatus. It is a figure which shows an example of the block division | segmentation by HEVC specification. It is a figure which shows the hierarchy division | segmentation of TU. It is a figure which shows the intra prediction mode in HEVC specification.

  Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the drawings. In general, in the prediction efficiency of intra prediction, the prediction efficiency in the direction orthogonal to the optimal one direction is extremely low, and the prediction efficiency is higher in the direction closer to the optimal one direction. In consideration of the nature of the prediction efficiency, it is not necessary to comprehensively evaluate the prediction efficiency in all directions. The prediction efficiency is evaluated only in a predetermined sampled direction, and the range of the direction to be evaluated is determined. By narrowing, it is expected that the optimum prediction direction can be determined at high speed. The present invention utilizes such a property of prediction efficiency.

  When performing intra prediction using a plurality of prediction directions, the intra prediction processing apparatus according to the embodiment of the present invention evaluates prediction efficiency for two orthogonal directions, and then performs a direction with high prediction efficiency ( For example, the prediction efficiency is evaluated for two directions (for example, the B direction and the C direction) having the lowest correlation in a predetermined region including the A direction. If the prediction efficiency is lower in the B direction than in the C direction, the prediction efficiency is not evaluated for the area between the BAs (the area between the B direction and the A direction). Only the region (the region between the C direction and the A direction) is an evaluation target of the prediction efficiency. In this way, by sequentially evaluating the prediction efficiency for the two directions, the direction area to be evaluated can be halved, and the direction area including the optimum prediction direction can be narrowed down. Hereinafter, a case where intra prediction according to the HEVC standard shown in FIG. 12 is performed will be described as an example.

[Intra prediction processor]
First, an intra prediction processing apparatus according to an embodiment of the present invention will be described. FIG. 1 is a block diagram showing a schematic configuration of an intra prediction processing apparatus according to an embodiment of the present invention. This intra prediction processing device 1 is a device that performs intra prediction using a plurality of prediction directions in the intra prediction mode in the HEVC standard shown in FIG. 12, and sequentially evaluates the prediction efficiency in two predetermined directions. An optimum prediction direction determination unit 2 that compares and determines an optimum prediction direction (optimum prediction direction) and outputs the result is provided.

  The optimal prediction direction determination unit 2 compares the comparison unit 3 that evaluates and compares the prediction efficiency with respect to a pair of orthogonal (two directions) prediction directions from among candidates for which the prediction efficiency is to be evaluated, and the comparison unit 3. A comparison unit 4 that selects a candidate direction area to be evaluated based on the result, and evaluates and compares the prediction efficiency for one set (two directions) of prediction directions with the lowest correlation included in the selected direction area. And. After the process of the comparison unit 3, the process of the comparison unit 4 is repeated as many times as necessary, so that the comparison unit 4 determines and outputs an optimum prediction direction. In addition, one set orthogonal among the predetermined candidates is also one set having the lowest correlation among the predetermined candidates.

  Before describing the process of the optimum prediction direction determination unit 2, the prediction direction shown in FIG. 12 will be described. FIG. 2 is a diagram for explaining the direction of the intra prediction mode, and shows the prediction directions and areas of indexes 2-34 in the HEVC standard shown in FIG. As shown in FIG. 2, the prediction direction of index 2 is the lower left direction, the prediction direction of index 10 is the horizontal left direction, the prediction direction of index 18 is the upper left direction, and the prediction direction of index 26 is the vertical upper direction. Let 34 prediction direction be the upper right direction. An area from the lower left direction of index 2 to the upper left direction of index 18 is a direction area belonging to the horizontal left, and an area from the upper left direction of index 18 to the upper right direction of index 34 is a direction area belonging to the upper vertical direction. Further, an area from the lower left direction of the index 2 to the horizontal left direction of the index 10 is a direction area belonging to the horizontal lower left, and an area from the horizontal left direction of the index 10 to the upper left direction of the index 18 is a direction area belonging to the horizontal upper left. The area from the upper left direction of the index 18 to the vertical upper direction of the index 26 is defined as a direction area belonging to the vertical upper left, and the area from the vertical upper direction of the index 26 to the upper right direction of the index 34 is defined as a direction area belonging to the vertical upper right.

  Here, the horizontal left direction of the index 10 and the vertical upper direction of the index 26 are orthogonal to each other. Similarly, the lower left direction of the index 2 and the upper left direction of the index 18 are also orthogonal to each other. 34 is also orthogonal to the upper right direction. Further, for example, in the direction region belonging to the vertical upper right of the indexes 26 to 34, the two directions having the lowest correlation are the vertical upper direction of the index 26 and the upper right direction of the index 34.

[Process of optimum prediction direction determination unit]
FIG. 3 is a flowchart showing an outline of processing of the optimum prediction direction determination unit 2 shown in FIG. First, the comparison unit 3 of the optimum prediction direction determination unit 2 calculates evaluation values for two orthogonal directions, compares the two evaluation values, and identifies a direction in which the evaluation value is large (a direction in which prediction efficiency is high) (step). S301). Note that the larger the evaluation value, the higher the prediction efficiency. Details of the orthogonal direction comparison processing (α) in step S301 will be described later with reference to FIG. And the comparison part 4 selects the direction area | region containing the prediction direction determined by the orthogonal direction comparison process ((alpha)) of step S301, and the direction which has not calculated evaluation value about two directions with the lowest correlation in the direction area | region. The evaluation value is calculated, the two evaluation values are compared, the direction in which the evaluation value is large is identified, the direction area of the next evaluation target is selected, and the direction is narrowed down by repeating such processing, and the optimal prediction A direction (direction with the largest evaluation value) is determined (step S302). Details of the narrowing comparison process (β) in step S302 will be described later with reference to FIG.

  FIG. 4 is a flowchart showing details of the process of the optimum prediction direction determination unit 2 shown in FIG. 4, steps S401 to S407 show details of the orthogonal direction comparison process (α) of step S301 shown in FIG. 3, and are performed by the comparison unit 3 of the optimum prediction direction determination unit 2. Steps S408 to S417 and the like show details of the narrowing comparison process (β) in step S302 shown in FIG. 3 and are performed by the comparison unit 4 of the optimum prediction direction determination unit 2.

(Orthogonal direction comparison processing (α))
First, the comparison unit 3 of the optimal prediction direction determination unit 2 sets the index 18 as an initial index (step S401). Then, the comparison unit 3 specifies the indexes 26 and 10 in the orthogonal direction from all the regions of the indexes 2 to 34 with the initial index 18 as a reference, and calculates the evaluation value b [26] in the vertical upward direction of the index 26. At the same time, an evaluation value b [10] in the horizontal left direction of the index 10 is calculated (step S402).

  Here, the evaluation value is a value obtained by quantifying the prediction efficiency in the direction. When the evaluation value is large, the prediction efficiency is high, and when the evaluation value is small, the prediction efficiency is low. For example, the evaluation value is calculated by RD cost determination or the like. Further, as the evaluation value, the reciprocal of the sum of absolute values of the pixel difference values between the predicted image in the prediction direction and the actual input image is calculated. When the difference between the predicted image and the input image is small (when the reciprocal of the difference is large), that is, when the evaluation value is large, the prediction efficiency is high, and when the difference is large (when the reciprocal of the difference is small), that is, the evaluation value When is small, the prediction efficiency is low.

  The comparison unit 3 compares the evaluation value b [26] with the evaluation value b [10], and determines whether or not the evaluation value b [26] is larger than the evaluation value b [10] (step S403). If the comparison unit 3 determines in step S403 that the evaluation value b [26] is greater than the evaluation value b [10] (step S403: Y), the comparison unit 3 uses the index 26 as a reference and the indexes 18 to 34 are compared. Select a directional area that belongs vertically. In this case, the comparison process is advanced for the direction region belonging to the vertical direction of the selected indexes 18 to 34, and the direction of the direction region belonging to the horizontal left of the unselected index 2 to 18 is excluded from the optimum prediction direction. Thereby, the amount of calculation can be reduced. The comparison unit 3 identifies the orthogonal indexes 34 and 18 as the two directions having the lowest correlation among the selected indexes 18 to 34. Then, the comparison unit 3 calculates the evaluation value b [34] in the upper right direction of the index 34 and calculates the evaluation value b [18] in the upper left direction of the index 18 (step S404).

  The comparison unit 3 compares the evaluation value b [34] with the evaluation value b [18], and determines whether or not the evaluation value b [34] is larger than the evaluation value b [18] (step S405). If the comparison unit 3 determines in step S405 that the evaluation value b [34] is greater than the evaluation value b [18] (step S405: Y), the comparison unit 3 proceeds to step S408. In this case, the comparison unit 3 sets the index 34 and the flag 0 (best = 34, flag = 0). Note that the processing in the case where it is determined in step S405 that the evaluation value b [34] is not larger than the evaluation value b [18] (step S405: N) is omitted. In this case, the comparison unit 3 sets the index 18 and the flag 1 (best = 18, flag = 1). The best and flag will be described later with reference to FIG.

  When the comparison unit 3 determines in step S403 that the evaluation value b [26] is not greater than the evaluation value b [10] (step S403: N), the comparison unit 3 uses the index 10 as a reference and the horizontal index 2-18. Select the direction area belonging to the left. In this case, the comparison process proceeds with respect to the direction region belonging to the horizontal left of the selected indexes 2 to 18, and the direction of the direction region belonging to the vertical direction of the unselected indexes 18 to 34 is excluded from the optimum prediction direction. Thereby, the amount of calculation can be reduced. The comparison unit 3 identifies the orthogonal indexes 18 and 2 as the two directions having the lowest correlation among the selected indexes 2 to 18. The comparison unit 3 calculates the evaluation value b [18] in the upper left direction of the index 18 and calculates the evaluation value b [2] in the lower left direction of the index 2 (step S406).

  The comparison unit 3 compares the evaluation value b [18] with the evaluation value b [2], and determines whether or not the evaluation value b [18] is larger than the evaluation value b [2] (step S407). When it is determined in step S407 that the evaluation value b [18] is larger than the evaluation value b [2] (step S407: Y), and the evaluation value b [18] is not larger than the evaluation value b [2]. (Step S407: N) is omitted. In this case, the comparison unit 3 sets the index 18 and the flag 0 (best = 18, flag = 0), and the index 2 and the flag 1 (best = 2, flag = 1). The best and flag will be described later with reference to FIG.

(Refinement comparison process (β))
When the evaluation value b [34] is larger than the evaluation value b [18] (step S405: Y), the comparison unit 4 of the optimal prediction direction determination unit 2 moves from step S405 and uses the index 34 as a reference. A direction area belonging to the vertical upper right of the indexes 26 to 34 having a bias to the right is selected from the indexes 18 to 34. In this case, the comparison process proceeds with respect to the direction area belonging to the upper right of the selected indexes 26 to 34, and the direction of the direction area belonging to the upper left of the unselected index 18 to 26 is excluded from the optimum prediction direction. Thereby, the amount of calculation can be reduced. The comparison unit 4 identifies the indexes 34 and 26 as the two directions having the lowest correlation among the selected indexes 26 to 34, and calculates the evaluation value b [26] in the vertical upward direction of the index 26 (step S408). Note that the evaluation value b [26] has not been calculated in step S408 since it has already been calculated in step S402. The reason why the evaluation value b [26] is calculated in step S408 is that the common function CompL is used as shown in FIG.

  The comparison unit 4 compares the evaluation value b [34] with the evaluation value b [26], and determines whether or not the evaluation value b [34] is larger than the evaluation value b [26] (step S409). When the comparison unit 4 determines in step S409 that the evaluation value b [34] is larger than the evaluation value b [26] (step S409: Y), the comparison unit 4 uses the index 34 as a reference and sets the indexes 26 to 34. Among them, the direction area of the indexes 30 to 34 having a bias to the right is selected. In this case, the comparison process is advanced for the selected direction area of the indexes 30 to 34, and the direction of the direction area of the unselected index 26 to 30 is excluded from the optimum prediction direction. Thereby, the amount of calculation can be reduced. The comparison unit 4 identifies the indexes 34 and 30 as the two directions having the lowest correlation among the selected indexes 30 to 34, and calculates the evaluation value b [30] of the index 30 (step S410). In this case, the evaluation value b [34] of the index 34 has been calculated.

  On the other hand, when the comparison unit 4 determines in step S409 that the evaluation value b [34] is not larger than the evaluation value b [26] (step S409: N), the comparison unit 4 proceeds from step S409 and uses the index 26 as a reference. Thus, the direction area of the indexes 26 to 30 having a bias to the left is selected from the indexes 26 to 34. In this case, the comparison process is advanced for the selected direction areas of the indexes 26 to 30, and the direction areas of the unselected index areas 30 to 34 are excluded from the optimum prediction direction. Thereby, the amount of calculation can be reduced. The description of the subsequent processing is omitted.

  The comparison unit 4 proceeds from step S410, compares the evaluation value b [34] with the evaluation value b [30], and determines whether the evaluation value b [34] is greater than the evaluation value b [30]. Determination is made (step S411). If the comparison unit 4 determines in step S411 that the evaluation value b [34] is larger than the evaluation value b [30] (step S411: Y), the comparison unit 4 uses the index 34 as a reference. Among them, the direction area of the indexes 32-34 having a bias on the right is selected. In this case, the comparison process is advanced for the selected direction areas of the indexes 32 to 34, and the directions of the direction areas of the unselected indexes 30 to 32 are excluded from the optimum prediction directions. Thereby, the amount of calculation can be reduced. The comparison unit 4 identifies the indexes 34 and 32 as the two directions having the lowest correlation among the selected indexes 32 to 34, and calculates the evaluation value b [32] of the index 32 (step S412). In this case, the evaluation value b [34] of the index 34 has been calculated.

  The comparison unit 4 performs the same processing in steps S413 to S415, and finally determines in step S415 that the evaluation value b [34] is larger than the evaluation value b [33] (step S415: Y) The index 34 is determined as the optimum prediction direction (step S416). On the other hand, if the comparison unit 4 determines in step S415 that the evaluation value b [34] is not larger than the evaluation value b [33] (step S415: N), the comparison unit 4 determines the index 33 as an optimal prediction direction ( Step S417).

(Program list)
Next, a program list when the process of the optimum prediction direction determination unit 2 shown in FIGS. 3 and 4 is executed by a C language program will be described. FIG. 5 is an example of a program list showing the process of the optimal prediction direction determination unit 2 in the C language format. FIG. 6 shows the orthogonal direction comparison process (α) in step S301 shown in FIG. 7 is an example of a program list in which the narrowing down comparison process (β) in step S302 is shown in a C language format.

  In FIG. 5, the Comp function is a function for executing the above-described orthogonal direction comparison process (α) by the comparison unit 3 of the optimum prediction direction determination unit 2, and among the prediction directions of the indexes 2 to 34 shown in FIG. 2, An intermediate index 18 is designated as the initial index. By this Comp function, best and flag described later are obtained. Then, the CompL function or the CompS function is executed according to the value of the flag. The CompL function is a function for executing the above-described narrowing comparison process (β) by the comparison unit 4 of the optimum prediction direction determination unit 2 for a prediction direction having a large index when flag = 0. On the other hand, the CompS function is a function for executing the above-described narrowing comparison process (β) by the comparison unit 4 of the optimum prediction direction determination unit 2 for a prediction direction with a small index when flag = 1.

  In FIG. 6, the Comp function is a function for executing the orthogonal direction comparison process (α) as described above, and the evaluation value evaluated as efficient among the set of prediction directions compared in the previous stage is as follows. An index x indicating a large one direction is input, and a layer that counts the number of comparisons is used to use recursive calls. The output signal of the Comp function is an index value stored in the memory best. Further, the end condition max_layer is 2 in the case of the direction number 33 shown in FIG. It should be noted that when the target coding block is square, it is always 2. The control flag flag is used for controlling the comparison unit 4 at the subsequent stage. Specifically, it is used to identify whether the index value evaluated to be efficient is larger or smaller from a set of prediction directions to be compared.

  E (x) in the program list is a function for calculating an evaluation value that is the prediction efficiency of the intra prediction of the index x. For example, RD cost determination or the like is used. Further, b [x] represents a memory that stores an evaluation value that is the prediction efficiency of the intra prediction of the index x.

  In FIG. 7, the CompS function and the CompL function are functions for executing the narrowing comparison process (β), as described above, and one of the functions is executed depending on the direction of the prediction direction according to the flag that is the input signal. That is, the comparison unit 4 of the optimum prediction direction determination unit 2 performs a narrowing comparison process (β) using these two functions. Specifically, the CompS function is a comparison unit to which the index value is input when the efficiency of the prediction direction with a small index value is high as a result of comparing a set of prediction directions, and the CompL function is On the contrary, when the efficiency in the prediction direction with a large index value is high, this is the comparison unit to which the index value is input.

  The CompS function and the CompL function are processing units that perform the next comparison based on the input index values, respectively, calculate the efficiency (evaluation value) of the next prediction candidate once, and input the prediction Compare with the efficiency (evaluation value) of the candidate. Then, the CompS function or the CompL function is executed again depending on the result. In addition, the CompS function and the CompL function have a layer that counts the number of comparisons in order to use recursive calls, similarly to the Comp function. In the example shown in FIG. 2, since the number of directions to be processed is 8, max_layer is 3.

  As described above, according to the intra prediction processing apparatus 1 according to the embodiment of the present invention, the prediction efficiency in the direction orthogonal to the optimal one direction is extremely low, and the prediction efficiency is higher in the direction closer to the optimal one direction. The optimal prediction direction was determined in consideration of the nature of prediction efficiency. Specifically, the comparison unit 3 of the optimum prediction direction determination unit 2 performs an orthogonal direction comparison process (α), calculates evaluation values for two orthogonal directions, compares the two evaluation values, and the evaluation value is large. The direction was specified. Then, the comparison unit 4 performs a narrowing comparison process (β), selects a direction region based on the prediction direction specified by the comparison unit 3, and calculates evaluation values for the two directions having the lowest correlation in the direction region. Calculate the evaluation value of the direction that has not been completed, compare the two evaluation values, specify the direction in which the evaluation value is large, select the direction area of the next evaluation target, narrow down the direction by repeating such processing, The optimal prediction direction was determined. This eliminates the need for evaluating the prediction efficiency in all directions and reduces the amount of calculation.

  Also, in the example of the program list representing the process of the optimum prediction direction determination unit 2 shown in FIGS. 5 to 7, the Comp function shown in a and b of FIG. 6, and the CompS function shown in cf of FIG. In the CompL function, the same function is executed in the program that executes the function, and the process is performed recursively. Thereby, since the process of the optimal prediction direction determination part 2 can be performed by the program by a recursive expression, it becomes possible to mount the said program compactly in the computer which comprises the intra prediction processing apparatus 1. FIG. .

  As described above, according to the intra prediction processing device 1 according to the embodiment of the present invention, the evaluation efficiency evaluation for two directions is sequentially executed by the program based on the recursive expression, thereby reducing the direction area to be evaluated by half. The direction area including the optimal prediction direction can be narrowed down at high speed. Therefore, even when there are many prediction directions for intra prediction, prediction direction determination processing can be performed at high speed without lowering prediction efficiency, and processing time and power consumption can be reduced.

[Image coding device]
Next, a case where the optimal prediction direction determination unit 2 of the intra prediction processing apparatus 1 shown in FIG. 1 is provided in an image encoding apparatus will be described. FIG. 8 is a block diagram showing the configuration of the image coding apparatus. The image encoding device 10 includes a preprocessing unit 11, a subtraction unit 12, an orthogonal transformation unit 13, a quantization unit 14, an inverse quantization unit 15, an inverse orthogonal transformation unit 16, an addition unit 17, a frame memory 18, and a signal prediction unit. 19 and an entropy encoding unit 20. The signal prediction unit 19 includes a prediction signal generation unit 21 and an optimal prediction direction determination unit 2 shown in FIG.

  The preprocessing unit 11 receives a signal of an image to be encoded, and performs predetermined preprocessing necessary for encoding on the input signal. The subtraction unit 12 receives the preprocessed signal from the preprocessing unit 11 and the prediction signal from the signal prediction unit 19 described later, and subtracts the prediction signal from the preprocessed signal to generate a difference signal. . The orthogonal transform unit 13 receives the difference signal as a subtraction result from the subtraction unit 12, performs orthogonal transform, and generates an orthogonal transform coefficient. The quantization unit 14 receives the orthogonal transform coefficient from the orthogonal transform unit 13 and performs quantization. The inverse quantization unit 15 receives the quantized signal from the quantization unit 14, performs inverse quantization, and outputs the inversely quantized signal to the inverse orthogonal transform unit 16. The inverse orthogonal transform unit 16 receives the inversely quantized signal from the inverse quantization unit 15, performs inverse orthogonal transform, and generates an inverse orthogonal transform coefficient to reproduce the difference signal. The adding unit 17 receives the inverse orthogonal transform coefficient from the inverse orthogonal transform unit 16 and also receives the prediction signal from the signal prediction unit 19 described later, and adds the prediction signal to the inverse orthogonal transform coefficient. The frame memory 18 receives the addition result signal from the adder 17 and stores it as a decoded signal. The decoded signal stored in the frame memory 18 is read out as a reference signal when generating a prediction signal by a signal prediction unit 19 described later.

  The signal prediction unit 19 reads the reference signal from the frame memory 18, generates a prediction signal by a predetermined prediction method based on the reference signal, and outputs the prediction signal to the subtraction unit 12 and the addition unit 17. Here, as described above, the optimal prediction direction determination unit 2 of the signal prediction unit 19 has a function of performing intra-screen prediction using a plurality of prediction directions in the intra prediction mode, and prediction efficiency in two predetermined directions. Are sequentially evaluated and compared to determine an optimum prediction direction (optimum prediction direction) and output to the prediction signal generation unit 21. The prediction signal generation unit 21 receives the optimal prediction direction from the optimal prediction direction determination unit 2 and generates a prediction signal in the optimal prediction direction. The entropy encoding unit 20 receives the quantized signal from the quantization unit 14, performs entropy encoding, and outputs an encoded signal. This encoded signal is output to the image decoding device 30 described later.

  In FIG. 8, the preprocessing unit 11 to the frame memory 18, the prediction signal generation unit 21 included in the signal prediction unit 19, and the entropy encoding unit 20 are known, and thus detailed description thereof is omitted here.

  As described above, according to the image encoding device 10 illustrated in FIG. 8, the optimal prediction direction in the intra prediction mode is determined at high speed using the optimal prediction direction determination unit 2 of the intra prediction processing device 1 illustrated in FIG. 1. Therefore, the processing time and power consumption can be reduced.

[Image decoding device]
Next, the case where the image decoding apparatus includes the optimum prediction direction determination unit 2 of the intra prediction processing apparatus 1 illustrated in FIG. 1 will be described. For example, there is an application when a code indicating an optimal prediction direction is not transmitted from the image encoding device 10 shown in FIG. FIG. 9 is a block diagram showing the configuration of the image decoding apparatus. The image decoding device 30 includes an entropy decoding unit 31, an inverse quantization unit 32, an inverse orthogonal transform unit 33, an addition unit 34, a post-processing unit 35, a frame memory 36, and a signal prediction unit 37. The signal prediction unit 37 includes a prediction signal generation unit 38 and the optimum prediction direction determination unit 2 shown in FIG.

  The entropy decoding unit 31 receives the encoded signal output from the image encoding device 10 illustrated in FIG. 8, and performs entropy decoding corresponding to the entropy encoding of the entropy encoding unit 20 illustrated in FIG. The inverse quantization unit 32 receives the signal entropy-decoded by the entropy decoding unit 31 and performs inverse quantization similar to the inverse quantization unit 15 illustrated in FIG. To the unit 33. The inverse orthogonal transform unit 33 receives the inversely quantized signal from the inverse quantization unit 32 and performs inverse orthogonal transform in the same manner as the inverse orthogonal transform unit 16 illustrated in FIG. The adder 34 receives a signal from the inverse orthogonal transform unit 33 and also receives a prediction signal from a signal prediction unit 37 described later, and adds both signals. The post-processing unit 35 inputs a decoded signal that is a signal resulting from the addition from the adding unit 34, and performs predetermined post-processing corresponding to the pre-processing of the pre-processing unit 11 illustrated in FIG. Output as an image output signal. The frame memory 36 receives the addition result signal from the adder 34 and stores it as a decoded signal. The decoded signal stored in the frame memory 36 is read as a reference signal when generating a prediction signal by a signal prediction unit 37 described later.

  The signal prediction unit 37 reads the reference signal from the frame memory 36, generates a prediction signal by a predetermined prediction method based on the reference signal, and outputs the prediction signal to the addition unit 34. Here, as described above, the optimal prediction direction determination unit 2 of the signal prediction unit 37 has a function of performing intra-screen prediction using a plurality of prediction directions in the intra prediction mode, and prediction efficiency in two predetermined directions. Are sequentially evaluated and compared to determine an optimum prediction direction (optimum prediction direction), and output to the prediction signal generation unit 38. The prediction signal generation unit 38 receives the optimal prediction direction from the optimal prediction direction determination unit 2 and generates a prediction signal in the optimal prediction direction.

  In FIG. 9, since the entropy decoding unit 31 to the frame memory 36 and the prediction signal generation unit 38 included in the signal prediction unit 37 are known, detailed description thereof is omitted here.

  As described above, according to the image decoding device 30 illustrated in FIG. 9, the optimal prediction direction in the intra prediction mode is set at high speed using the optimal prediction direction determination unit 2 of the intra prediction processing device 1 illustrated in FIG. 1. Thus, the processing time and power consumption can be reduced.

  As mentioned above, although the intra prediction processing apparatus 1 by embodiment of this invention was demonstrated, the above-mentioned method is the direction of evaluation object by performing evaluation of prediction efficiency about two directions sequentially by the program by a recursive expression. It halves the area and is targeted for single thread processing. On the other hand, the intra prediction processing apparatus 1 according to another aspect does not sequentially discard the evaluation target direction areas by evaluating the prediction efficiency, but performs multi-threading that evaluates the prediction efficiency with brute force in parallel processing. Realize the process. Specifically, the intra-prediction processing device 1 performs the process of narrowing down the region by evaluating the prediction efficiency in two directions and specifying the direction in which the evaluation value is large in a brute force manner, and the optimal prediction direction To decide. In other words, the process (branch process) by the above-described method is used as thread allocation for multi-threads, and the process of truncating without branching (the process of “N” in the branch process shown in FIG. 4) is always assigned to another thread. assign.

  Thereby, even if the prediction direction of intra prediction is a large number, the same effect as the above-described method that can perform the prediction direction determination processing at high speed without lowering the prediction efficiency, Since the entire processing load is evenly distributed to each of the parallel processes, the thread use efficiency can be made uniform. Moreover, the processing effect suitable for the multithread for determining the optimal prediction direction can be obtained by the brute force processing. Therefore, the intra prediction processing apparatus 1 can be used efficiently and the maximum performance can be exhibited.

  In addition, as a hardware configuration of the intra prediction processing apparatus 1, the image encoding apparatus 10, and the image decoding apparatus 30 according to the embodiment of the present invention, a normal computer can be used. The intra prediction processing device 1, the image encoding device 10, and the image decoding device 30 are configured by a computer including a CPU, a volatile storage medium such as a RAM, a non-volatile storage medium such as a ROM, an interface, and the like. The function of the optimum prediction direction determination unit 2 provided in the intra prediction processing device 1 is realized by causing the CPU to execute a program describing this function (for example, the program shown in FIGS. 5 to 7). In addition, the preprocessing unit 11, the subtraction unit 12, the orthogonal transformation unit 13, the quantization unit 14, the inverse quantization unit 15, the inverse orthogonal transformation unit 16, the addition unit 17, the frame memory 18, and the signal included in the image encoding device 10. Each function of the prediction unit 19 and the entropy encoding unit 20 is realized by causing the CPU to execute a program describing these functions. The functions of the entropy decoding unit 31, the inverse quantization unit 32, the inverse orthogonal transform unit 33, the addition unit 34, the post-processing unit 35, the frame memory 36, and the signal prediction unit 37 included in the image decoding device 30 are as follows. Each is realized by causing the CPU to execute a program describing the function. These programs can also be stored and distributed in a storage medium such as a magnetic disk (floppy (registered trademark) disk, hard disk, etc.), optical disk (CD-ROM, DVD, etc.), semiconductor memory, etc. You can also send and receive.

DESCRIPTION OF SYMBOLS 1 Intra prediction processing apparatus 2 Optimal prediction direction determination part 3 and 4 Comparison part 10 Image coding apparatus 11 Preprocessing part 12 Subtraction part 13 Orthogonal transformation part 14 Quantization part 15 Inverse quantization part 16 Inverse orthogonal transformation parts 17 and 34 Addition Units 18 and 36 Frame memories 19 and 37 Signal prediction unit 20 Entropy encoding units 21 and 38 Prediction signal generation unit 30 Image decoding device 31 Entropy decoding unit 32 Inverse quantization unit 33 Inverse orthogonal transform unit 35 Post-processing unit

Claims (4)

In an intra prediction processing device for determining a prediction direction of intra prediction when an image is divided into a plurality of blocks and intra coding is performed on the block,
For each of the two directions with the lowest correlation included in the predetermined direction area, the prediction efficiency is evaluated, and the direction area including the direction with the high prediction efficiency is selected
Identify the two directions with the lowest correlation included in the selected direction area, evaluate the prediction efficiency in the two directions, and select the direction area including the direction with the high prediction efficiency;
The process of specifying the two directions, the process of evaluating the prediction efficiency, and the process of selecting the direction area are repeated to narrow down the direction with high prediction efficiency,
An intra prediction processing apparatus comprising an optimum prediction direction determination unit that determines a prediction direction having the highest prediction efficiency among the plurality of prediction directions. In the intra prediction processing apparatus of Claim 1,
The optimum prediction direction determination unit
A first comparison unit that evaluates and compares prediction efficiencies for two orthogonal directions among the plurality of prediction directions;
According to the comparison result by the first comparison unit, a direction region including a direction with high prediction efficiency is selected, and prediction efficiency is evaluated and compared for each of the two directions with the lowest correlation included in the selected direction region. A second comparison unit,
The second comparison unit further selects a direction region including a direction with high prediction efficiency according to a comparison result by the second comparison unit, and the two directions having the lowest correlation included in the selected direction region. An intra-prediction processing device, wherein the process of evaluating and comparing the prediction efficiencies is repeated to determine the prediction direction with the highest prediction efficiency among the plurality of prediction directions. In the intra prediction processing apparatus of Claim 2,
An intra prediction processing apparatus, wherein the processing in the two directions in the first comparison unit or the second comparison unit is recursively performed in the processing.   The program for functioning a computer as an intra prediction processing apparatus as described in any one of Claim 1 to 3.

Images