JP2013017128A - Intra-prediction mode estimation device, image encoder, image decoder and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an intra-prediction mode estimation device for estimating a prediction mode in intra-prediction, and an image encoder, an image decoder and a program.SOLUTION: An intra-prediction mode estimation device 1 includes: a reference region determining section 2 for dividing an image of a prescribed reference region into an image of a first region and an image of a second region, which are previously determined within the reference region in accordance with a plurality of types of intra-prediction modes; a first region pixel value predicting section 3 for generating a predicted value for a pixel value of the image in the first region from a pixel value of the image in the second region; a comparing section 4 for generating a comparison value by comparing the predicted value generated in accordance with the intra-prediction mode for the pixel value of the image in the second region and the pixel value of the image of the first region; and an intra-prediction mode estimating section 5 for totaling the comparison values of respective pixels, which are generated by comparison between the first region and the second region for each intra-prediction mode, determining an evaluation value for each intra-prediction mode and estimating the intra-prediction mode optimum for intra-prediction.

Description

  The present invention relates to an intra prediction mode estimation device, an image encoding device, an image decoding device, and a program for estimating a prediction mode in intra prediction for predicting an image pattern of a prediction target region from an image pattern of a reference region.

  In an image encoding technique for performing intra prediction, an input image (or a residual image after motion compensation) is divided into a plurality of blocks, and surrounding pixel value groups that have already been encoded are decoded and stored as reference pixels. Prediction is performed by interpolating the pixel value of a certain target block (prediction block) using this reference pixel (for example, see Patent Documents 1 to 3). Thereafter, the image coding technique performs entropy coding on the residual between the prediction result of the intra prediction and the input image (or the residual image after motion compensation).

  FIG. 11 is a block diagram of a conventional image coding apparatus such as AVC / H.264 (ISO / IEC 14496-10 / ITU_T Rec. H.264). The image encoding device 50 includes a subtraction unit 51, an orthogonal transformation / quantization unit 52, an inverse quantization / inverse orthogonal transformation unit 53, an addition unit 54, an image memory 55, an intra prediction unit 56, and an entropy code. And a conversion unit 57. The description regarding inter prediction is omitted since it is not directly related to the subject matter of the present invention.

  The subtraction unit 51 takes the difference between the input image signal and the prediction image signal, and outputs a prediction residual that is the result of this difference. The orthogonal transform / quantization unit 52 obtains a coefficient by transforming the prediction residual using orthogonal transform such as discrete cosine transform, quantizes the coefficient with a predetermined quantization width, and uses the quantization result as a result. Output a certain conversion coefficient. The inverse quantization / inverse orthogonal transform unit 53 performs inverse transform of quantization and orthogonal transform on the transform coefficient output from the orthogonal transform / quantization unit 52, and outputs a prediction residual as a result of the inverse transform. To do. The adder 54 adds the prediction residual output from the inverse quantization / inverse orthogonal transform unit 53 and the predicted image, and stores the decoded image as a result of the addition in the image memory 55. The intra prediction unit 56 uses a decoded pixel of a block that has been stored in the image memory 55 and has already been encoded among the input image signals to be encoded as a reference pixel (for example, a block to be encoded (for example, Intra prediction is performed on each pixel of a block obtained by dividing the input image every 4 pixels in the vertical direction and every 4 pixels in the horizontal direction, and a prediction image as a result of the intra prediction is output. The entropy encoding unit 57 performs entropy encoding (for example, variable length encoding such as run length encoding, Huffman encoding, or arithmetic encoding) on the transform coefficient output from the orthogonal transform / quantization unit 52. Generate an encoded stream.

  The intra prediction unit 56 outputs intra prediction mode information indicating a prediction method of intra prediction to the decoding side to the entropy encoding unit 57, and the entropy encoding unit 57 outputs the orthogonal transform / quantization unit 52. Entropy coding is performed on the intra prediction mode information in conjunction with the transformed coefficients. In the intra prediction mode information, for example, as illustrated in FIG. 13 in AVC / H.264, a prediction mode number can be assigned to a prediction method of intra prediction as UDI (unrestricted digital information). Therefore, by transmitting a predetermined prediction mode number to the decoding side as intra prediction mode information, the decoding side can specify a prediction mode that defines a prediction method for intra prediction.

  FIG. 12 is a block diagram of a conventional image decoding apparatus such as AVC / H.264. The image decoding device 70 includes an entropy decoding unit 71, an inverse quantization / inverse orthogonal transform unit 72, an addition unit 73, an intra prediction unit 74, and an image memory 75. The description regarding inter prediction is omitted since it is not directly related to the subject matter of the present invention.

  The entropy decoding unit 71 performs entropy decoding on the input encoded stream, and extracts transform coefficients and intra prediction mode information. The inverse quantization / inverse orthogonal transform unit 72 performs inverse quantization and inverse orthogonal transform on the transform coefficient, and outputs a prediction residual. The addition unit 73 adds the prediction residual output from the inverse quantization / inverse orthogonal transform unit 72 and the prediction image generated by the intra prediction unit 74 to output a decoded image signal to the outside, and the image memory 75. To store. The intra prediction unit 74 acquires the prediction mode from the intra prediction mode information, and among the decoding target images, the decoded image of the block stored in the image memory 75 and already decoded is used as a reference pixel, and the target block is determined. Intra prediction is performed in the prediction mode acquired for each pixel, and a prediction image as a result of the intra prediction is output.

  Thus, for example, in intra prediction of the AVC / H.264 standard, prediction is performed with 0th-order hold from neighboring pixels of a prediction block, but in general, a plurality of trials are performed with an optimal prediction direction in terms of rate versus distortion. It is specified from the results. Thereby, optimal prediction is performed in the intra prediction on the encoding side, but it is necessary to transmit information on the intra prediction mode indicating the prediction method to the decoding side.

  In order to reduce the transmission of the information amount of the intra prediction mode indicating the prediction direction to the decoding side, it is possible to analyze the texture of the decoded reference pixels and reference blocks and predict the prediction mode in the intra prediction process. The amount of code for entropy coding can be reduced by calculating the difference between the prediction value obtained by predicting the prediction mode itself and the prediction mode, and an improvement in coding efficiency regarding information transmission in the prediction mode can be expected.

  For example, in HEVC (High Efficiency Video Coding), after applying image analysis based on the Sobel operator to a reference block including a group of surrounding pixel values that have been decoded, a prediction mode that seems to be optimal is obtained by obtaining an eigenvalue. Has been proposed as DCIM (Differential Coding of Intra Modes) (see, for example, Non-Patent Document 1).

JP 2009-49969 A JP 2008-245088 A JP 2008-271371 A

E. Maani, W. Liu, “Differential Coding of Intra Modes (DCIM)”, Doc. JCTVC-B109, Geneva, Switzerland, Jul 2010

  As intra prediction processing such as AVC / H.264, there is processing for predicting a prediction block from the image feature of a reference block including peripheral pixels of the prediction block by using a zero-order hold or an average value of pixel values indicating the directionality. Can be mentioned. However, in DCIM proposed in Non-Patent Document 1, the prediction direction in the intra prediction process is estimated by using arithmetic means (Sobel operator) that is not directly related to the intra prediction process. ing. Therefore, since there is a correlation between the intra prediction process and the DCIM (between operations having different behaviors), the prediction mode itself is used for prediction. There is room for improvement. In addition, in DCIM, it is necessary to introduce a different operation separately from the intra prediction process, and there is room for improvement from the viewpoint of processing cost.

  In view of the above-described problems, an object of the present invention is to provide an intra prediction mode estimation apparatus, an image encoding apparatus, and an image decoding apparatus that estimate a prediction mode in intra prediction that predicts an image pattern of a prediction target area from an image pattern of a reference area. And providing a program.

  In order to solve the above-described problem, an intra prediction mode estimation device according to the present invention is an intra prediction mode estimation device that estimates a prediction mode in intra prediction for predicting an image pattern of a prediction target region from an image pattern of a reference region, In accordance with each of a plurality of predetermined intra prediction modes, an image of a predetermined reference area is converted into an image of a first area predetermined in the reference area, and the reference area different from the first area. A reference area determination unit that divides the image into the second area image determined in advance, and a prediction value for the pixel value of the image in the first area associated according to the intra prediction mode from the pixel value of the image in the second area. A first region pixel value prediction unit to be generated; a prediction value generated according to an intra prediction mode for a pixel value of an image of the second region; and an image of the first region. And a comparison unit that generates a comparison value with the pixel value of each pixel, and the comparison value of each pixel compared between the first region and the second region for each intra prediction mode, and an evaluation value for each intra prediction mode And an intra prediction mode estimation unit for estimating an intra prediction mode optimal for intra prediction.

  Moreover, in the intra prediction mode estimation apparatus according to the present invention, prediction that generates a prediction image in which each pixel of the prediction target region is predicted using a pixel value of the image of the first region based on the estimated intra prediction mode. An image generation unit is further provided.

  According to such a configuration, the intra prediction mode can be estimated by a process highly correlated with the process used for the intra prediction. Therefore, it is possible to share hardware and software for executing the processing of the intra prediction mode estimation apparatus according to the present invention with respect to the processing used for the existing intra prediction, thereby reducing the scale and processing cost of the hardware and software. be able to. Moreover, since the process of the intra prediction mode estimation apparatus according to the present invention can be executed using the same operator having high correlation with the process used for the existing intra prediction, the estimation accuracy of the estimation of the intra prediction mode is improved. . The intra prediction mode estimated by the intra prediction mode estimation device according to the present invention can be used for switching the mode of intra prediction as it is, and mode switching information of the intra prediction mode determined by any other technique. In other words, it can be used for prior information (intra prediction mode estimation information that functions as a prediction value of intra prediction mode information) for instructing (intra prediction mode information).

  In the intra prediction mode estimation apparatus according to the present invention, the intra prediction includes direction prediction, and when estimating the intra prediction mode related to the direction prediction, the first region pixel value prediction unit includes the image of the second region. The prediction value of the image of the first region associated by extrapolation in the direction opposite to the prediction direction of the direction prediction of the intra prediction is generated from the pixel value of.

  In the intra prediction mode estimation device according to the present invention, the intra prediction includes an average value prediction, and when the intra prediction mode is estimated by the average value prediction, the comparison unit is configured to calculate a pixel value of the image in the second region. And a comparison value between the average value of pixel values of the image in the first region.

  According to such a configuration, it is possible to estimate an appropriate mode reflecting a pattern (texture directionality) from intra prediction modes based on direction prediction and average value prediction.

  An image encoding apparatus according to the present invention is an image encoding apparatus that encodes an image by performing intra prediction for predicting an image pattern of a prediction target region from an image pattern of a reference region, and includes an input image signal and a predicted image A subtraction unit that generates a prediction residual signal that is a result of the difference, transforms the prediction residual signal using orthogonal transformation, obtains a coefficient, An orthogonal transform / quantization unit that quantizes with a quantization width and generates a transform coefficient that is a result of the quantization; an entropy encoding unit that performs entropy coding on the transform coefficient and generates an encoded stream; An inverse quantization / inverse orthogonal transform unit that performs inverse transform of quantization and orthogonal transform on the transform coefficient, and an output signal of the inverse quantization / inverse orthogonal transform unit and a signal of the prediction image are added. Of the addition An addition unit that stores a decoded image that is a result in an image memory; and intra prediction that predicts an image pattern in a prediction target region from an image pattern in a reference region that is stored in the image memory and that has been decoded. The intra prediction mode estimation apparatus of this invention which estimates the prediction mode in prediction and produces | generates the said estimated image which is the result of the said intra prediction is characterized by the above-mentioned.

  An image encoding apparatus according to the present invention is an image encoding apparatus that encodes an image by performing intra prediction for predicting an image pattern of a prediction target region from an image pattern of a reference region, and includes an input image signal and intra prediction The first subtraction unit that takes a difference from the predicted image signal and generates a prediction residual signal as a result of the difference, and obtains a coefficient by converting the prediction residual signal using orthogonal transformation An orthogonal transform / quantization unit that quantizes the coefficient with a predetermined quantization width and generates a transform coefficient that is a result of the quantization, and performs entropy coding on the transform coefficient to generate an encoded stream An entropy encoding unit, an inverse quantization / inverse orthogonal transform unit that performs inverse transform of quantization and orthogonal transform on the transform coefficient, an output signal of the inverse quantization / inverse orthogonal transform unit, and the predicted image And the signal An addition unit that stores a decoded image that is a result of the addition in an image memory; and the intra that predicts an image pattern of a prediction target region from a decoded pixel group that is stored in the image memory and is close to the prediction target region An intra prediction unit that performs prediction and generates the prediction image that is a result of the intra prediction, and generates information on a prediction mode in the intra prediction as intra prediction mode information, and a decoded image stored in the image memory The intra prediction mode estimation apparatus of the present invention to be input, and information related to the intra prediction mode optimal for the intra prediction estimated by the intra prediction mode estimation apparatus is acquired as intra prediction mode estimation information, and the intra prediction mode estimation information and the intra prediction mode estimation information The difference between the intra prediction mode information and the intra prediction mode And a second subtraction unit which generates a difference information, the entropy encoding unit and generates the encoded stream by adding the intra prediction mode difference information.

  An image decoding apparatus according to the present invention is an image decoding apparatus that decodes an image by performing intra prediction for predicting an image pattern of a prediction target area from an image pattern of a reference area, and for an input encoded stream Entropy decoding, an entropy decoding unit that extracts a transform coefficient obtained by performing orthogonal transform and quantization on a prediction residual signal in intra prediction on the image coding side, and dequantizing and transforming the transform coefficient Performing inverse orthogonal transform, generating a signal of the prediction residual, an inverse quantization / inverse orthogonal transform unit, adding the signal of the prediction residual and the signal of the predicted image to generate a signal of the decoded image; An adder that stores the decoded image in an image memory; and an intra prediction mode estimation device of the present invention that generates a predicted image from the decoded image stored in the image memory, The intra prediction mode estimation device performs intra prediction that predicts an image pattern of a prediction target area from an image pattern of a reference area that is stored in the image memory and includes a decoded pixel group, and estimates a prediction mode in the intra prediction. Then, the prediction image as a result of the intra prediction is generated.

  An image decoding apparatus according to the present invention is an image decoding apparatus that decodes an image by performing intra prediction for predicting an image pattern of a prediction target area from an image pattern of a reference area, and for an input encoded stream Performs entropy decoding, extracts transform coefficients obtained by performing orthogonal transform and quantization on the prediction residual signal in intra prediction on the image encoding side, and information on prediction modes in intra prediction on the image encoding side An entropy decoding unit that extracts predetermined intra prediction mode difference information indicating a difference with respect to the intra prediction mode information, and performs inverse quantization and inverse orthogonal transform on the transform coefficient to generate a signal of the prediction residual The inverse quantization / inverse orthogonal transform unit, the prediction residual signal, and the intra prediction image signal are added to obtain a decoded image A first adder that stores the decoded image in an image memory, the intra prediction mode estimation device of the present invention that inputs the decoded image stored in the image memory, and the intra prediction mode estimation Information regarding the intra prediction mode optimal for the intra prediction estimated by the apparatus is acquired as intra prediction mode estimation information, and the intra prediction mode estimation information is added to the intra prediction mode difference information to restore the intra prediction mode information. A second adder that performs the intra prediction that predicts the image pattern of the prediction target region using the decoded pixel group that is stored in the image memory and close to the prediction target region, and the restored intra prediction mode information. An intra prediction unit that generates the prediction image that is a result of the intra prediction. And wherein the Rukoto.

  Moreover, the program of this invention can be comprised as a program for making a computer perform each component of the intra prediction mode estimation apparatus of this invention.

  According to the present invention, it is possible to estimate an optimal intra prediction mode for intra prediction direction prediction and average value prediction.

It is a block diagram of the intra prediction mode estimation apparatus of one Embodiment by this invention. (A), (B), (C), and (D) are prediction blocks made up of a plurality of pixels to be predicted that can be used in the intra prediction mode estimation apparatus according to an embodiment of the present invention, and decoded blocks. It is a figure which illustrates each reference block which consists of a plurality of pixels. (A), (B) is a figure explaining calculation of the evaluation value regarding an intra prediction mode about an example of direction prediction in the intra prediction mode estimation apparatus of one Embodiment by this invention. It is a block diagram of the intra prediction mode estimation apparatus of 1st Example by this invention. It is a block diagram of the intra prediction mode estimation apparatus of 2nd Example by this invention. It is a flowchart which shows operation | movement of the intra prediction mode estimation apparatus of one Embodiment by this invention. It is a block diagram which shows the image coding apparatus of 1st Embodiment provided with the intra prediction mode estimation apparatus of one Embodiment by this invention. It is a block diagram which shows the image decoding apparatus of 1st Embodiment provided with the intra prediction mode estimation apparatus of one Embodiment by this invention. It is a block diagram which shows the image coding apparatus of 2nd Embodiment provided with the intra prediction mode estimation apparatus of one Embodiment by this invention. It is a block diagram which shows the image decoding apparatus of 2nd Embodiment provided with the intra prediction mode estimation apparatus of one Embodiment by this invention. It is a block diagram which shows the conventional image coding apparatus. It is a block diagram which shows the conventional image decoding apparatus. It is explanatory drawing regarding the prediction mode number of the intra prediction utilized conventionally.

  First, “intra prediction” and “intra prediction mode” and “reference region”, “reference pixel”, and “target pixel” according to the present invention will be described.

  Intra prediction means predicting a pixel value sequence in an undecoded image region (prediction block) based on a pixel value sequence in a decoded image region (reference block) in image encoding (or video encoding). Say technology.

  FIG. 2 shows an example of a prediction block made up of a plurality of pixels to be predicted and a reference block made up of a plurality of decoded pixels. 2A to 2D, the prediction block 101 is predicted by extrapolation or average value calculation using the reference pixels 103 (pixel rows indicated by hatching) in a partial area in the reference block 102. Often. In FIG. 2, the shape of the prediction block 101 is illustrated as a square of 8 pixels × 8 pixels, but it does not necessarily have to be 8 pixels square and does not have to be a square. Further, the reference pixel 103 does not need to be a pixel column adjacent to the prediction block 101, and can be a pixel column of two or more adjacent rows (or two columns), and includes a reference close to the prediction block 101 comprehensively. Pixels in a partial area in the block 102 can be used as the reference pixels 103.

  In addition, since it is not necessary to limit the reference block 102 to a block shape, an image region including a plurality of decoded pixels referred to predict a pixel of the prediction block 101 is referred to as a “reference region”. Called. In particular, as will be described later, a pixel used for prediction of a prediction block by extrapolation or average value calculation is referred to as a “reference pixel”, and an area composed of a plurality of reference pixels is referred to as a “reference image”. A region other than the reference pixel among the pixels in the reference region is referred to as a “target pixel”, and a region including a plurality of target pixels is referred to as a “target image region (second region)”. Called.

  As an intra prediction method, for example, a pixel value sequence of a prediction block is obtained by extrapolating a pixel value at a predetermined position in a reference block in a predetermined direction, or a plurality of pixel values at a predetermined position in a reference block are obtained. There is a method of determining a pixel value sequence of a prediction block based on a result of calculation (for example, an average value).

  The “intra prediction mode” is an identifier for distinguishing a plurality of types of intra prediction prediction methods. In this specification, as illustrated in FIG. It can be identified by a number (m = 0, 1, 2,..., M). 13 exemplifies nine typical types used in AVC / H.264. For example, as proposed in DCIM (Differential Coding of Intra Modes) of Non-Patent Document 2, there are 33 different types. It is also possible to define up to 34 types of intra prediction modes consisting of direction prediction and one type of average value prediction. In the present specification, the intra prediction mode is also simply referred to as “prediction mode”.

  Hereinafter, an intra prediction mode estimation apparatus according to an embodiment of the present invention will be described.

(Intra prediction mode estimation device)
FIG. 1 is a block diagram of an intra prediction mode estimation apparatus according to an embodiment of the present invention. The intra prediction mode estimation device 1 receives an image signal of a decoded reference region (for example, the reference block 102 illustrated in FIG. 2), and predicts a prediction target region (for example, the prediction illustrated in FIG. 2) from the image pattern of the reference region. This is a device for estimating a prediction mode in intra prediction for predicting an image pattern of block 101). Although it can be defined in advance to change the area of the reference block 102 to be different depending on the intra prediction mode, in the example described below, the image of the input reference area is the reference shown in FIG. An example including the image of the block 102 will be described representatively. The intra prediction mode estimation device 1 estimates one or more optimal prediction modes from among a plurality of intra prediction modes, or arranges a plurality of intra prediction modes in order from those estimated to be optimal, The result is generated as intra prediction mode estimation information.

  As shown in FIG. 1, an intra prediction mode estimation apparatus 1 according to an embodiment of the present invention includes a reference region determination unit 2, a first region pixel value prediction unit 3, a comparison unit 4, and an intra prediction mode estimation unit 5. And a prediction mode / pixel scanning control unit 6 and a predicted image generation unit 7.

  The reference area determining unit 2 inputs an image of a predetermined reference area, and determines a predetermined reference area image in the reference area in accordance with each of a plurality of predetermined intra prediction modes. The image of the region (reference image region) and the image of the second region (target image region) determined in advance within the reference region different from the first region are divided, and the image of the first region is sent to the comparison unit 4; The image of the second area is output to the first area pixel value prediction unit 3.

  Information on a plurality of types of intra prediction modes and information on the first region and the second region are stored in a memory (not shown) held by the prediction mode / pixel scanning control unit 6, and the prediction mode / pixel scanning control unit 6 controls the reference region determination unit 2 to scan each of a plurality of types of intra prediction modes. Further, as will be apparent from the description of the embodiment with reference to FIG. 4 and FIG. 5 described later, the prediction mode / pixel scanning control unit 6 determines the reference region determination unit 2 according to each intra prediction mode. The pixels constituting the image of the second region (target image region) are respectively scanned, and the pixel of the image of the first region corresponding to a certain pixel of the image of the second region corresponding to the intra prediction mode is determined. Thereby, the reference area determination unit 2 includes a certain pixel and a second area (target image area) of the image of the first area (reference image area) divided according to each of a plurality of types of predetermined intra prediction modes. Can be associated according to the intra prediction mode to be scanned.

  The first region pixel value prediction unit 3 predicts the pixel value of the image of the first region (reference image region) associated according to the intra prediction mode to be scanned from the pixel value of the image of the second region (target image region). A value is generated and sent to the comparison unit 4. For example, when estimating the intra prediction mode related to the direction prediction, the first region pixel value prediction unit 3 reverses the prediction direction of the direction prediction of the intra prediction from the pixel value of the image of the second region (target image region). A predicted value is generated for the pixel value of the image in the first region associated by extrapolation of the direction. In generating the predicted value, a weight corresponding to the Euclidean distance between the pixel in the second region (target image region) and the pixel in the first region is set to the pixel value of the image in the second region (target image region). The prediction value for the pixel value of the image in the first area can be generated, and the pixel value of the image in the second area (target image area) is simply used as the prediction value for the pixel value of the image in the first area. Also good.

  The comparison unit 4 includes, for all pixels scanned in the second region (target image region) for each intra prediction mode, a prediction value generated according to the intra prediction mode for the pixel value of the image in the second region, and the first region The pixel value of the image in the (reference image area) is compared, and a comparison value indicating the comparison result is sent to the intra prediction mode estimation unit 5. For example, for all pixels scanned in the second region (target image region), the comparison unit 4 sets the pixel value of the image in the second region (target image region) and the pixel value of the image in the first region (reference image region). Is sent to the intra prediction mode estimation unit 5 as a comparison value.

  Further, when estimating the intra prediction mode by the average value prediction, the reference area determination unit 2 can generate an average value based on the pixel value of each pixel in the first area and send it to the comparison unit 4. The comparison unit 4 compares the pixel value of the image of the second region with the average value of the pixel values of each pixel in the first region for all the pixels scanned in the second region, and shows the comparison result Generate a value.

  The intra-prediction mode estimation unit 5 compares the comparison values for all the pixels scanned in the second region (target image region) for each intra-prediction mode (that is, for each pixel compared between the first region and the second region). (Comparison values) are aggregated to determine an evaluation value for each intra prediction mode, and an optimal intra prediction mode for intra prediction is estimated based on the evaluation values of a plurality of types of intra prediction modes. The evaluation value only needs to indicate the similarity between the pixels in the first region and the pixels in the second region associated with each intra prediction mode. That is, when the intra prediction mode estimation unit 5 estimates the optimal intra prediction mode for intra prediction based on the evaluation values of a plurality of types of intra prediction modes, the intra prediction mode is optimized for the intra prediction. It estimates as an intra prediction mode, and this estimated intra prediction mode is produced | generated as intra prediction mode estimation information. As an example of a prediction mode that maximizes the similarity, a prediction mode that minimizes an evaluation value obtained by a difference, as described later, or a prediction mode that maximizes an evaluation value obtained by cross-correlation may be used. it can.

  Each time the intra prediction mode estimation unit 5 determines an evaluation value for each intra prediction mode to be scanned, the intra prediction mode estimation unit 5 sends the fact to the prediction mode / pixel scanning control unit 6. Further, the intra prediction mode estimation unit 5 obtains the intra prediction mode by the prediction mode / pixel scanning control unit 6 and the scan completion information regarding the pixels in the second region (target image region), so that each intra prediction mode to be scanned is acquired. The evaluation value can be determined.

  That is, the intra prediction mode estimation device 1 compares the predicted value predicted from the image of the second region (target image region) with the image of the first region (reference image region) according to a plurality of types of intra prediction modes. The most suitable intra prediction mode for intra prediction is estimated by selecting the most similar one. In the examples described later, the images in the first region (reference image region) are scanned and compared pixel by pixel. However, since it is sufficient that an evaluation value for each intra prediction mode can be determined, every other pixel is used. Intermittent scanning may be performed.

  In particular, in the intra prediction mode estimation device 1, the intra prediction mode estimation unit 5 predicts the first region (reference image region) from the image of the second region (target image region) and the original first region (reference). The prediction mode that maximizes the similarity between the image in the image region) is estimated as the intra prediction mode optimal for intra prediction. More specifically, the intra prediction mode estimation unit 5 includes an image obtained by predicting the first region (reference image region) from the image of the second region (target image region) and the original first region (reference image region). Estimating the optimal intra prediction mode for intra prediction by minimizing the evaluation value based on the difference from the image, or estimating the optimal intra prediction mode for intra prediction by maximizing the evaluation value obtained by cross-correlation It can be performed. In addition, regarding “evaluation value based on difference” and “evaluation value obtained by cross-correlation”, the evaluation value is determined by substantially the same calculation. However, “evaluation value based on difference” is most minimized. The “evaluation value obtained by cross-correlation” is a standard indicating a high degree of similarity, and the maximum value is a standard indicating the highest degree of similarity.

  In the intra prediction mode estimation apparatus 1, the evaluation value for each intra prediction mode includes an image obtained by predicting the first region (reference image region) from the image of the second region (target image region) and the original first region ( The sum of absolute values of differences from the image in the reference image area) can be obtained.

  In the intra prediction mode estimation apparatus 1, the evaluation value for each intra prediction mode includes an image obtained by predicting the first region (reference image region) from the image of the second region (target image region) and the original first region ( The sum of the absolute value of the difference from the image in the reference image area) to the nth power (n is a positive integer) can be used.

  Note that the evaluation value for each intra prediction mode is not only the sum of the absolute value of the difference but the nth power (n is a positive integer), and the sum of the simple difference is the nth power (n is a positive integer). In other words, the evaluation value for each intra prediction mode includes an image obtained by predicting the first region (reference image region) from the image of the second region (target image region) and the original first region (reference image region). The sum of the nth power of the difference from the image of.

  In the intra prediction mode estimation apparatus 1, the evaluation value for each intra prediction mode includes an image obtained by predicting the first region (reference image region) from the image of the second region (target image region) and the original first region ( It can be the maximum absolute value of the difference from the image in the reference image area.

  The calculation of the evaluation value for each intra prediction mode enables estimation of the intra prediction mode by a simple arithmetic operation such as a difference calculation or a sum calculation, and is a process highly correlated with the existing intra prediction. Therefore, for example, it is possible to estimate with high reliability while reducing the processing cost as compared with DCIM proposed in Non-Patent Document 1.

  The predicted image generation unit 7 performs extrapolation or average value calculation using the pixel values of the image of the first region (reference image region) based on the intra prediction mode estimated by the intra prediction mode estimation unit 5. A prediction image in which each pixel of the prediction block is predicted is generated.

  The intra prediction mode estimation apparatus 1 according to the present embodiment calculates the evaluation value of each intra prediction mode illustrated in FIG. 13 from the pixel value sequence in the reference block 102 predetermined as the reference region as illustrated in FIG. When selecting the optimum “one” from a plurality of intra prediction modes as the mode estimation value (prediction mode number estimated as optimum), the prediction mode that maximizes the similarity based on the evaluation value is used. A mode estimation value is generated, and this mode estimation value is generated as intra prediction mode estimation information.

  Further, when the intra prediction mode estimation device 1a selects a plurality (N) of optimum ones from a plurality of intra prediction modes as the mode estimation value, this evaluation value is the nth minimum (or The prediction mode to be maximum) is obtained for n∈ {1, 2,..., N}, and a set of these prediction modes is set as a mode estimation value (a set of prediction mode numbers with an identifier estimated to be the nth optimal), This mode estimation value is generated as intra prediction mode estimation information.

  Further, when the intra prediction mode estimation device 1a arranges a plurality of intra prediction modes in order from those estimated to be optimum, and the result is set as a mode estimation value, this evaluation value is the nth smallest (or The prediction mode to be maximized) is obtained for n∈ {1, 2,..., N}, and the mode estimation values (orders estimated as optimal) are indicated by a sequence obtained by arranging these prediction modes in ascending order (or descending order) of n. The mode estimation value is generated as intra prediction mode estimation information.

  FIG. 3 is a diagram illustrating calculation of an evaluation value related to a prediction mode for an example of direction prediction in the intra prediction mode estimation device 1 of the present embodiment. FIGS. 3A and 3B show a calculation method for calculating the evaluation value of the intra prediction mode by referring to the reference block 102 in FIG. 2A in each prediction direction 104.

  In normal intra prediction, in order to obtain a prediction value of a certain pixel of the prediction block 101, a half line is extended in the direction of the prediction direction 104 from that pixel, and which reference pixel 103 the vicinity of this half line passes through. And extrapolation along the prediction direction 104 is performed by using the calculated pixel value of the reference pixel 103 as a predicted pixel value.

  As shown in FIG. 3, in the intra prediction mode estimation apparatus 1 of the present embodiment, a half line 106 opposite to the prediction direction 104 is drawn for each pixel (target pixel) 105 in the reference block 102. Through which reference pixel 103 passes, and the pixel value of the obtained reference pixel 107 is compared with the pixel value of the target pixel 105. The intra prediction mode estimation apparatus 1 according to the present embodiment scans the target pixel 105 in the reference block 102, obtains a comparison value, determines an evaluation value based on the comparison value, and determines an optimum intra based on the evaluation value. Estimate the prediction mode of prediction.

  For example, the evaluation value E (m) for the prediction mode number m is obtained by the following procedure. Let L be the set of reference pixels 103. An area obtained by excluding the area L from the area of the reference block 102 (denoted as area Q) is denoted as R. A line segment 109 that passes through the center of the reference pixel group horizontally aligned among the reference pixels 103 and belongs to the area of the reference pixel group, and a region of the reference pixel group that passes through the center of the reference pixel group vertically aligned among the reference pixels 103 Let d be the two line segments that combine the line segment 110 belonging to the inside. Let the pixel value of the image coordinate (position vector) x be I (x). It is assumed that the image coordinate of the target pixel 105 is p and the image coordinate r (p; m) of the corresponding reference pixel.

  A half line 106 starting from the target pixel 105 and parallel to the prediction direction 104 and going in the opposite direction is set as h (p; m). At this time, the image coordinate r (p; m) of the reference pixel 107 is obtained from Expression (1).

  When there is no pixel that satisfies the condition on the right side of Equation (1), an exceptional value that means “no reference pixel” is set in r (p; m). A set of pixels where r (p; m) is not “no reference pixel” (there is a reference pixel) (a region surrounded by a thick dotted line 108 in the example of FIG. 3) is S.

  As the evaluation value E (m) for the prediction mode number m, for example, Expression (2) can be used.

  Here, n is a real number. n is preferably a positive real number. For example, n = 1 or n = 2 can be used. When n = 1, Expression (1) represents the average value of the absolute values of the prediction residuals in the region S.

  Further, for example, the evaluation value E (m) may be obtained by the equation (3).

  The prediction mode may include average value prediction (for example, prediction mode number m = 2 illustrated in FIG. 13). In the case of the average value prediction, the evaluation value E (m) can be determined by Expression (4) or Expression (5). Here, μ is an average value of pixel values in the set L of reference pixels. The region S can be set to S = R, for example.

  An example of how to implement average prediction is described. First, an extra storage area sufficient to store pixel value data is secured in the memory that stores the image I (x). An address that refers to this storage area (here, “exceptional image coordinate value” that refers to the storage area) is set to Aav. Prior to obtaining the evaluation value E (m = average value prediction mode), the average value μ of the pixel values included in the set L is obtained and stored in the memory as I (Aav) = μ. By setting r (p; m = average value prediction mode) to r (p; m = average value prediction mode) = Aav regardless of the image coordinate p, equation (4) can be calculated by equation (2). It becomes. Also, equation (5) can be calculated by equation (3).

  The intra prediction mode estimation device 1 corresponds to a plurality of different modes m (a set of prediction mode numbers m to be evaluated is W = {0, 1, 2,..., M}, and the number of elements of the set W is T). Based on the evaluation value E, the mode estimation value P is output as intra prediction mode estimation information.

  For example, the intra prediction mode estimation device 1 has a prediction mode number m that minimizes the evaluation value E (m) (or a prediction mode number m that is estimated based on a prediction mode that maximizes the evaluation value E (m)). The mode estimation value P is assumed. In order to obtain the minimum and maximum of the evaluation value E (m), they can be defined as shown in Equation (6) and Equation (7), respectively.

  Also, for example, the intra prediction mode estimation device 1 includes a set of prediction mode numbers m that sets the evaluation value E (m) to the 1st to kth (2 ≦ k ≦ T, k is an integer) minimum (or maximum), The mode estimated value P may be used.

  Further, for example, the intra prediction mode estimation device 1 starts from the prediction mode number m (or the prediction mode that maximizes) that minimizes the evaluation value E (m) to the 1st to kth (2 ≦ k ≦ T, where k is an integer). The mode estimation value P may be determined by a numerical sequence in which the optimal prediction mode numbers m) to be estimated are arranged in ascending order (or descending order) with respect to k.

  The intra prediction mode estimation device 1 of the present embodiment can be incorporated into an image encoding device or an image decoding device, as will be described later. The intra prediction mode information and the prediction image generated by the intra prediction mode estimation device 1 according to the present embodiment are the intra prediction mode information and the prediction image, or the intra prediction mode depending on the use of the image encoding device and the image decoding device to be incorporated. Note that only one of the information and the predicted image can be used. Therefore, the predicted image generation unit 7 is not necessarily provided depending on the application.

  Hereinafter, based on the intra prediction mode estimation apparatus of one Embodiment by this invention, a more specific Example is described. In the following embodiments, an example in which the pixel value of the pixel in the second area (target image area) is used as the predicted value of the pixel in the first area (reference image area) will be described.

(Intra prediction mode estimation apparatus of the first embodiment)
FIG. 4 is a block diagram of the intra prediction mode estimation apparatus according to the first embodiment of the present invention. The intra prediction mode estimation device 1a according to the first embodiment is a block diagram illustrating an example of the configuration of the intra prediction mode estimation device related to direction prediction, and includes a pixel scanning unit 11, a target pixel reading unit 12, a mode scanning unit 13, and the like. , A reference pixel determining unit 14, a reference pixel reading unit 15, a comparing unit 16, a totaling unit 17, a prediction mode determining unit 18, and an image memory 20. The image memory 20 is a temporary storage memory that functions as a reference area image storage unit. The image memory 20 that holds the input reference area image may be included in the constituent elements of the intra prediction mode estimation apparatus 1 or may be configured to be connected to the outside of the intra prediction mode estimation apparatus 1.

  First, in the intra prediction mode estimation device 1a of the first embodiment, the correspondence relationship with the intra prediction mode estimation device 1 shown in FIG. 1 will be described. An image of a predetermined reference area input to the intra prediction mode estimation device 1a of the first embodiment is stored in the image memory 20. The reference region determination unit 2 in FIG. 1 determines the image of the first region (reference image region) predetermined in the reference region and the first region in accordance with each of a plurality of predetermined intra prediction modes. The image is divided into images of a predetermined second area (target image area) in different reference areas. This corresponds to the pixel scanning unit 11, the mode scanning unit 13, the reference pixel determining unit 14, and the reference pixel reading unit 15 in FIG. The first region pixel value prediction unit 3 in FIG. 1 uses the pixel values of the image of the second region (target image region) and the pixels of the image of the first region (reference image region) associated according to the intra prediction mode to be scanned. Generate a predicted value for the value. This corresponds to the target pixel reading unit 12 in FIG. The comparison unit 4 in FIG. 1 corresponds to the comparison unit 16 in FIG. The intra prediction mode estimation unit 5 in FIG. 1 corresponds to the counting unit 17 and the prediction mode determination unit 18 in FIG. The prediction mode / pixel scanning control unit 6 in FIG. 1 shows an example in which the pixel scanning unit 11, the mode scanning unit 13, and the totaling unit 17 are controlled by a clock in FIG. In FIG. 4, the description of the predicted image generation unit 7 in FIG. 1 is omitted.

  The reference area image held in the image memory 20 may be an entire frame of the image to be encoded, or may be a part of this frame. However, the image memory 20 includes each pixel value of the reference block 102.

  Since the pixel scanning unit 11 scans pixels included in the region R excluding the reference pixel group L in the reference block Q, the image coordinates of the pixels included in the region R are set in a predetermined order in synchronization with the clock signal. Output sequentially. As this predetermined order, for example, as in the case of pixel scanning by a general progressive scanning line, the order in which the pixels on the scanning line moving from top to bottom in the vertical direction are sequentially selected from left to right in the horizontal direction is selected. Can be used. This order is, for example, arranged on a memory such as ROM (Read Only Memory) or RAM (Random Access Memory), and the image coordinates of the pixels to be scanned in order from the smallest address of this array are stored in order. This can be realized by giving an address to the memory by a counter that counts in synchronization with the timing of the clock and reading the image coordinates stored in the address. The pixel scanning unit 11 outputs a pixel scanning completion signal when all the pixels included in the region R have been scanned.

  The target pixel reading unit 12 reads the pixel value I (p) of the pixel (target pixel) at the designated image coordinate p from the image I (x) of the reference area stored in the image memory 20. A memory address corresponding to the coordinates is output to the image memory 20, data is read from the image memory 20, and the read data is output to the comparison unit 16 as a target pixel value.

  The mode scanning unit 13 sequentially changes the prediction mode number m to be output every time a pixel scanning completion signal is input. The prediction mode number is an identifier for distinguishing M (M is an integer of 2 or more) prediction modes. For example, the mode scanning unit 13 sequentially outputs all M types of prediction mode numbers by inputting M image operation completion signals. Further, for example, the mode scanning unit 13 may be operated to scan only a part of all the prediction modes by inputting a predetermined number (two or more and less than M) of image operation completion signals.

  The mode scanning unit 13 outputs a mode scanning completion signal when all the predetermined prediction mode numbers to be scanned are scanned.

  The reference pixel determination unit 14 obtains an image coordinate r (p; m) of the reference pixel from the input image coordinate (image coordinate of the target pixel) p and the input prediction mode number m, and outputs this. . For example, in the example of FIG. 3, the image coordinates of the reference pixel 107 are obtained based on the image coordinates 105 of the target pixel and the prediction mode number m, and the obtained image coordinates are output.

  When the prediction method is directional mode, the image coordinate r (p; m) of the reference pixel is uniquely determined from the image coordinate p of the target pixel and the prediction mode number m. For this reason, the reference pixel determination unit 14 can be implemented as a look-up table that takes the image coordinates p of the target pixel and the prediction mode number m as arguments, for example.

  The reference pixel reading unit 15 out of the image I (x) stored in the image memory 20 has a pixel value I (r (p; m) of a pixel (reference pixel) at a specified image coordinate r (p; m). )) Is read, the memory address corresponding to the image coordinates is output to the image memory 20, the data is read from the image memory 20, and the read data is output to the comparison unit 16 as a reference pixel value.

  The comparison unit 16 compares the target pixel value I (p) input from the target pixel reading unit 12 with the reference pixel value I (r (p; m)) input from the reference pixel reading unit 15, and compares the result. A comparison value d (p; m) indicating is output.

  For example, the comparison unit 16 sets the difference between the target pixel value I (p) and the reference pixel value I (r (p; m)) as the comparison value d (p; m).

  The tabulation unit 17 tabulates the comparison value d (p; m) input from the comparison unit 16 for each prediction mode number m with respect to the image coordinates p of the target pixel. The aggregation unit 17 can store the aggregation value t in a predetermined memory (not shown). When the pixel scanning completion signal is input, the totaling unit 17 clears the internal storage means (for example, t = 0). The totaling unit 17 updates the total value t based on the input comparison value d (p; m) at the timing of the input clock.

  For example, the totaling unit 17 updates the total value t according to Expression (9).

  Here, n is a positive real constant. For example, when n = 1, the counting unit 17 operates to obtain an L1 norm of a difference between the target pixel value I (p) and the reference pixel value I (r (p; m)). For example, when n = 2, the totaling unit 17 calculates the square of the L2 norm (Euclidean norm) of the difference between the target pixel value I (p) and the reference pixel value I (r (p; m)). Operate. When n is an even number, the absolute value calculation of equation (9) can be omitted.

  Further, for example, the totaling unit 17 updates the total value t according to Expression (10).

  According to Expression (10), the totalization unit 17 operates to obtain the L∞ norm of the difference between the target pixel value I (p) and the reference pixel value I (r (p; m)).

  The prediction mode determination unit 18 receives the input of the total value t at the timing of the pixel scanning completion signal, and uses the received total value as t (m) for calculation of the evaluation value (the prediction mode at the moment when the total value t is received) Suppose the number was m). That is, the comparison value d (p; m) in the prediction mode number m regarding the target pixel p is set as a total value t (m) when the totals for the region R are completed.

  The prediction mode determination unit 18 sets the total value t (m) of the prediction mode number m as the evaluation value E (m) in the equation (2) or the equation (3) at every interval when the mode scanning completion signal is input. An optimal intra prediction mode is estimated based on the equation (6) or the equation (7), and a mode estimation value P indicating the estimation result is output as intra prediction mode estimation information.

  For example, as shown in the equation (11), the prediction mode determination unit 18 uses the prediction mode number m having the smallest total value t (m) within the interval at which the mode scanning completion signal is input, as a mode estimation value. P. Note that, similarly to the description of Expression (7), the prediction mode determination unit 18 may use the prediction mode number m determined based on the prediction mode having the largest total value t (m) as the mode estimation value P.

  In addition, for example, the prediction mode determination unit 18 may determine a predetermined number of prediction mode number sets as the mode estimation value P from the smaller (or larger) value of the total value t (m), or the total value t A number sequence in which a predetermined number of prediction mode numbers m are arranged in order from (m) in ascending (or increasing) order may be determined as the mode estimation value P.

(Intra prediction mode estimation device of the second embodiment)
FIG. 5 is a block diagram of an intra prediction mode estimation apparatus according to the second embodiment of the present invention. The intra prediction mode estimation apparatus 1b of 2nd Example is a block diagram which shows an example of a structure of the intra prediction mode estimation apparatus regarding direction prediction and average value prediction, The pixel scanning part 11, the object pixel read-out part 12, and mode A scanning unit 13, a reference pixel determining unit 14, a reference pixel reading unit 15, a comparing unit 16, a totaling unit 17, a prediction mode determining unit 18, an average value setting unit 19, and an image memory 20 are provided. The image memory 20 is a temporary storage memory that functions as a reference area image storage unit. The image memory 20 that holds the input reference area image may be included in the constituent elements of the intra prediction mode estimation apparatus 1 or may be configured to be connected to the outside of the intra prediction mode estimation apparatus 1.

  The intra prediction mode estimation device 1b of the second embodiment is further provided with an average value setting unit 19 in order to include the average value prediction in the prediction mode, as compared with the intra prediction mode estimation device 1a of the first embodiment. Is different. Therefore, the intra prediction mode estimation device 1b of the second embodiment is the same as the intra prediction mode estimation device 1a of the first embodiment except for the processing of the average value setting unit 19, and the same components are the same. A reference number is attached. Hereinafter, the point which determines the evaluation value at the time of "average value prediction" different from 1st Example is demonstrated.

  The average value setting unit 19 reads out the pixel values of the pixel group of the image I (x) in the reference region (the pixel group in the region L of the reference pixel 103) from the image memory 20, calculates the average value, and calculates the image memory. 20. In calculating the average value, rounding may be performed for less than a predetermined bit (for example, less than 1). The obtained average value is stored in the memory address Aav that does not interfere with the image I (x) in the image memory 20.

  When the prediction mode number m is an average value prediction (for example, the prediction mode number m = 2 in FIG. 13), the reference pixel determination unit 14 does not depend on the image coordinates of the target pixel input from the pixel scanning unit 11. A special image coordinate (for example, (-1, -1)) different from the image coordinate in the image I (x) is always output.

  When this “special image coordinate” is input, the reference pixel reading unit 15 stores the pixel value stored in the memory address Aav from the image memory 20 (that is, the average value shown in Expression (4) or Expression (5)). μ) is read out and outputted to the comparison unit 16 as a reference pixel value.

  The comparison unit 16 sets a difference between the target pixel value I (p) and the average value μ as a comparison value d (p; m).

  The aggregation unit 17 aggregates the comparison value d (p; m) input from the comparison unit 16 with respect to the image coordinates p of the target pixel, similarly to the intra prediction mode estimation device 1a of the first embodiment.

  The prediction mode determination unit 18 sets the total value t (m) when the prediction mode number m indicates average value prediction as the evaluation value E (m) in the equation (4) or the equation (5).

  Accordingly, the prediction mode determination unit 18 determines the evaluation value E (m) from the total value t (m) of the prediction mode number m including the direction prediction and the average value prediction at every interval at which the mode scanning completion signal is input. The optimal intra prediction mode is estimated based on the equation (6) or the equation (7), and the mode estimation value P indicating the estimation result is output as the intra prediction mode estimation information.

  For example, as with the intra prediction mode estimation device 1a of the first embodiment, the prediction mode determination unit 18 calculates the total value t (within the interval at which the mode scanning completion signal is input, as shown by the equation (11). The mode estimation value P is determined by the prediction mode number m in which m) is the smallest. Note that, similarly to the description of Expression (7), the prediction mode determination unit 18 may use the prediction mode number m determined based on the prediction mode having the largest total value t (m) as the mode estimation value P.

  As described in the first and second embodiments, the intra prediction mode estimation device 1 can be operated even when the intra prediction mode includes average value prediction.

  Next, an example of the operation of the intra prediction mode estimation device 1 according to the present embodiment will be described with reference to FIG. 1, FIG. 3, and FIG.

(Device operation)
First, an image of a predetermined reference area is input by the reference area determination unit 2, and a specific prediction mode number (for example, m = 0) among a plurality of predetermined intra prediction modes is previously stored in the reference area. The image is divided into a predetermined first region (reference image region) and a predetermined second region (target image region) within a reference region different from the first region (step S11).

  Next, as described with reference to FIG. 3, the reference region determination unit 2 determines the reference pixel r corresponding to the target pixel p for the prediction method of the prediction mode number m (step S12).

  Next, the first region pixel value prediction unit 3 generates a prediction value for the pixel value of the reference pixel r from the pixel value of the target pixel p associated in accordance with the prediction method of the prediction mode number m (step S13). .

  Next, the comparison unit 4 compares the predicted value from the target pixel p associated according to the prediction method of the prediction mode number m with the pixel value of the reference pixel r, and calculates a comparison value indicating the comparison result. (Step S14).

  Next, the prediction mode / pixel scanning control unit 6 controls the prediction method of the prediction mode number m to repeat steps S12 to S14 until all target pixels scanned in the target image region are compared (step S15). ).

  Next, the intra prediction mode estimation unit 5 adds up all comparison values related to the target image region in the prediction mode number m to determine an evaluation value in the prediction mode number m (step S16).

  Next, the prediction mode / pixel scanning control unit 6 controls to repeat Steps S12 to S16 until the comparison values are totaled for all prediction modes (Step S17). For example, the prediction mode / pixel scanning control unit 6 starts from m = 0 when the prediction mode is given by an intra prediction number mε {0, 1, 2,..., M} as shown in FIG. When m = m + 1, control is performed so that steps S12 to S16 are repeated (step S18).

  Finally, the intra prediction mode estimation unit 5 determines an evaluation value based on the comparison values aggregated in all prediction modes, determines an optimal prediction mode, and estimates the determined prediction mode as a prediction mode for intra prediction. (Step S19).

  As described above, the intra prediction mode estimation device 1 according to the present embodiment can determine an optimal intra prediction mode for intra prediction in direction prediction and average value prediction. The intra prediction mode can be estimated by processing having high consistency with processing used for existing intra prediction such as H.264 (that is, processing having high correlation). For this reason, it is possible to share hardware and software for executing the intra prediction process, and it is possible to estimate the intra prediction mode by simple arithmetic operations such as difference calculation and summation calculation. Processing costs can be reduced. In addition, since the intra prediction mode is estimated by a process having high consistency with the existing intra prediction process (that is, a process having high correlation), the estimation accuracy for the prediction mode predicted by the existing intra prediction is improved. be able to. As will be described later, the intra prediction mode estimated by the intra prediction mode estimation apparatus 1 can be used as it is for switching modes of intra prediction as it is, and the mode of the intra prediction mode determined by any other technique. It can be used for prior information (predicted value of intra prediction mode information) for instructing switching information (intra prediction mode information).

  Moreover, when estimating the prediction mode regarding direction prediction, the intra prediction mode estimation apparatus 1 uses a pixel value of a pixel in the second region (target image region) and has a direction opposite to the prediction direction of the direction prediction of the intra prediction. When a prediction value of a pixel in the first region (reference image region) associated by extrapolation is generated and the intra prediction mode is estimated by average value prediction, the first region in the reference region (for example, the reference block 102) Since an average value based on the pixel value of each pixel in the (reference image area) is used, an appropriate mode reflecting the pattern (texture directionality) is estimated from the direction prediction and the intra prediction mode based on the average value prediction. Will be able to.

  Next, an example in which the intra prediction mode estimation device 1 of the present embodiment is applied to an image encoding device and an image decoding device will be described.

[Image Encoding Device and Image Decoding Device of First Embodiment]
FIG. 7 is a block diagram showing an image coding apparatus according to the first embodiment including the intra prediction mode estimation apparatus according to an embodiment of the present invention. FIG. 8 is a block diagram illustrating an image decoding apparatus according to the first embodiment including the intra prediction mode estimation apparatus according to an embodiment of the present invention.

(Image coding device)
The image encoding device 50a of the present embodiment includes a subtraction unit 51, an orthogonal transform / quantization unit 52, an inverse quantization / inverse orthogonal transform unit 53, an adder 54, an image memory 55, and an intra prediction mode estimation. The apparatus 1 and the entropy encoding part 57 are provided. The description regarding inter prediction is omitted since it is not directly related to the subject matter of the present invention.

  Specifically, the image encoding device 50a of the present embodiment is a device that encodes an image by performing intra prediction that predicts the image pattern of the prediction target region from the image pattern of the reference region, and the prediction according to the present invention. An intra prediction mode estimation device 1 for generating an image is provided.

  The subtraction unit 51 takes a difference between the input image signal and the signal of the prediction image generated by the intra prediction mode estimation device 1, generates a prediction residual signal as a result of the difference, and generates an orthogonal transform / quantization unit 52. To send.

  The orthogonal transform / quantization unit 52 transforms the prediction residual signal obtained from the subtraction unit 51 using orthogonal transform to obtain a coefficient, quantizes the coefficient with a predetermined quantization width, and the result of the quantization Is generated and sent to the inverse quantization / inverse orthogonal transform unit 53 and the entropy coding unit 57.

  The entropy encoding unit 57 performs entropy encoding on the transform coefficient obtained from the orthogonal transform / quantization unit 52 to generate an encoded stream.

  The inverse quantization / inverse orthogonal transform unit 53 performs inverse transform of quantization and orthogonal transform on the transform coefficient obtained from the orthogonal transform / quantization unit 52, and outputs a prediction residual signal as a result of the inverse transform. It is generated and sent to the adder 54.

  The adder 54 adds the prediction residual signal and the prediction image signal, and stores the decoded image as a result of the addition in the image memory 55.

  The intra prediction mode estimation device 1 performs intra prediction that predicts the image pattern of the prediction target region from the image pattern of the reference region that is stored in the image memory 55 and includes the decoded pixel group, and estimates the prediction mode in the intra prediction. A prediction image as a result of the intra prediction is generated and sent to the subtraction unit 51 and the addition unit 54.

  Here, the entropy encoding unit 57 generates an encoded stream so that information regarding the prediction mode in intra prediction is not output to the outside.

(Image decoding device)
As shown in FIG. 8, the image decoding device 70a of the present embodiment includes an entropy decoding unit 71, an inverse quantization / inverse orthogonal transform unit 72, an addition unit 73, an intra prediction mode estimation device 1, and an image. And a memory 75.

  Specifically, the image decoding device 70a of the present embodiment is a device that decodes an image by performing intra prediction that predicts the image pattern of the prediction target region from the image pattern of the reference region, and the predicted image according to the present invention is decoded. An intra prediction mode estimation device 1 for generation is provided.

  The entropy decoding unit 71 performs entropy decoding on the input encoded stream, performs orthogonal transform on the prediction residual signal in intra prediction on the image encoding side (in this example, the image encoding device 50a), and The quantized transform coefficient is extracted and sent to the inverse quantization / inverse orthogonal transform unit 72.

  The inverse quantization / inverse orthogonal transform unit 72 performs inverse quantization and inverse orthogonal transform on the transform coefficient obtained from the entropy decoding unit 71, generates a prediction residual signal, and sends the signal to the adder 73.

  The addition unit 73 adds the prediction residual signal obtained from the inverse quantization / inverse orthogonal transformation unit 72 and the prediction image signal generated by the intra prediction mode estimation device 1 to generate a decoded image signal. At the same time, the decoded image is stored in the image memory 75.

  The intra prediction mode estimation device 1 performs intra prediction that predicts an image pattern of a prediction target area from an image pattern of a reference area that is stored in the image memory 75 and includes decoded pixels, and estimates a prediction mode in intra prediction. A predicted image as a result of the intra prediction is generated and sent to the adding unit 73.

  7 and 8, the same reference numerals are given to the same components as those of the conventional image encoding device 50 shown in FIG. 11 and the image decoding device 70 shown in FIG. In the comparison between the image encoding device 50a and the image decoding device 70a of FIG. 5 and the conventional image encoding device 50 and the image decoding device 70, the image encoding device 50a and the image decoding device 70a of the present embodiment include an intra prediction unit 56. , 74 in that an intra prediction mode estimation device 1 is used. This difference will be described in detail below.

  The intra prediction unit 56 in the conventional image encoding device 50 assigns a prediction mode number to a prediction method of intra prediction as UDI (unrestricted digital information), and uses a predetermined prediction mode number as intra prediction mode information. By transmitting to the decoding device 70, the conventional image decoding device 70 can identify a prediction mode that defines a prediction method for intra prediction. For this reason, it is necessary to transmit intra prediction mode information in addition to image information.

  The image encoding device 50a and the image decoding device 70a of the present embodiment include an intra prediction mode estimation device 1, and the intra prediction mode estimation device 1 refers to decoded pixels stored in the image memories 55 and 75. As a pixel of a region, the image of each first region (reference image region) predicted according to a plurality of types of intra prediction modes is most similar in comparison with the image of the second region (target image region) in the reference region. By selecting one, an intra prediction mode in intra prediction is estimated, and a predicted image is generated.

  Therefore, according to the image encoding device 50a and the image decoding device 70a of the first embodiment, transmission of intra prediction mode information can be made unnecessary, and transmission efficiency can be improved. Furthermore, the image decoding apparatus 70a of this embodiment can also be used as an apparatus for decoding an encoded stream by the conventional image encoding apparatus 50. Intra prediction is possible without using any intra prediction mode information included in the encoded stream by the conventional image encoding device 50, or, alternatively, an intra prediction mode included in the encoded stream by the conventional image encoding device 50 Information may be used as a reference value. That is, the image decoding device 70a of this embodiment can be used without distinction between the image encoding device 50a of this embodiment and the conventional image encoding device 50.

[Image Encoding Device and Image Decoding Device of Second Embodiment]
FIG. 9 is a block diagram illustrating an image encoding device according to the second embodiment including the intra prediction mode estimation device according to an embodiment of the present invention. FIG. 10 is a block diagram illustrating an image decoding apparatus according to a second embodiment including the intra prediction mode estimation apparatus according to an embodiment of the present invention.

(Image coding device)
The image encoding device 50b of the present embodiment includes a first subtraction unit 51, an orthogonal transform / quantization unit 52, an inverse quantization / inverse orthogonal transform unit 53, an adder 54, an image memory 55, an intra A prediction unit 56, an intra prediction mode estimation device 1, an entropy encoding unit 57, and a second subtraction unit 58 are provided. The description regarding inter prediction is omitted since it is not directly related to the subject matter of the present invention.

  Specifically, the image encoding device 50b of the present embodiment is a device that encodes an image by performing intra prediction that predicts the image pattern of the prediction target region from the image pattern of the reference region, and the intra according to the present invention. An intra prediction mode estimation apparatus 1 that generates prediction mode estimation information is provided.

  The first subtraction unit 51 takes a difference between the input image signal and the predicted image signal generated by the intra prediction unit 56, generates a prediction residual signal as a result of the difference, and performs orthogonal transform / quantization. The data is sent to the unit 52.

  The orthogonal transform / quantization unit 52 transforms the prediction residual signal obtained from the first subtraction unit 51 using orthogonal transform to obtain a coefficient, quantizes the coefficient with a predetermined quantization width, A transform coefficient that is a result of the quantization is generated and sent to the inverse quantization / inverse orthogonal transform unit 53 and the entropy coding unit 57.

  The entropy encoding unit 57 performs entropy encoding on the transform coefficient obtained from the orthogonal transform / quantization unit 52 to generate an encoded stream.

  The inverse quantization / inverse orthogonal transform unit 53 performs inverse transform of quantization and orthogonal transform on the transform coefficient obtained from the orthogonal transform / quantization unit 52, and outputs a prediction residual signal as a result of the inverse transform. It is generated and sent to the adder 54.

  The adder 54 adds the prediction residual signal obtained from the inverse quantization / inverse orthogonal transform unit 53 and the signal of the prediction image, and stores the decoded image as a result of the addition in the image memory 55.

  The intra prediction unit 56 performs intra prediction that predicts the image pattern of the prediction target region from the decoded pixel group that is stored in the image memory 55 and close to the prediction target region, and generates a prediction image that is a result of the intra prediction. To the first subtraction unit 51 and the addition unit 54, and information on the prediction mode in the intra prediction is generated as intra prediction mode information and sent to the second subtraction unit 58.

  The second subtraction unit 58 acquires information about the intra prediction mode optimal for the intra prediction estimated by the intra prediction mode estimation device 1 as the intra prediction mode estimation information, and the intra prediction mode estimation information and the intra prediction unit 56 A difference with the intra prediction mode information is generated as intra prediction mode difference information and sent to the entropy encoding unit 57.

  The entropy encoding unit 57 generates an encoded stream so as to output the intra prediction mode difference information to the outside instead of outputting the intra prediction mode information to the outside.

(Image decoding device)
As shown in FIG. 10, the image decoding device 70b of the present embodiment includes an entropy decoding unit 71, an inverse quantization / inverse orthogonal transform unit 72, a first addition unit 73, an intra prediction unit 74, The intra prediction mode estimation apparatus 1, the image memory 75, and the 2nd addition part 76 are provided.

  Specifically, the image decoding device 70b according to the present embodiment is a device that decodes an image by performing intra prediction that predicts an image pattern of a prediction target region from an image pattern of a reference region, and the intra prediction mode according to the present invention. An intra prediction mode estimation device 1 that generates estimation information is provided.

  The entropy decoding unit 71 performs entropy decoding on the input encoded stream, performs orthogonal transform on the prediction residual signal in the intra prediction on the image encoding side (in this example, the image encoding device 50b), and As the information about the prediction mode in the intra prediction on the image coding side (in this example, the image coding device 50b), the quantized transform coefficient is extracted and sent to the inverse quantization / inverse orthogonal transform unit 72. Predetermined intra prediction mode difference information indicating a difference with respect to the intra prediction mode information is extracted and sent to the second addition unit 76.

  The inverse quantization / inverse orthogonal transform unit 72 performs inverse quantization and inverse orthogonal transform on the transform coefficient obtained from the entropy decoding unit 71, generates a signal of the prediction residual, and sends the signal to the first addition unit 73. Send it out.

  The first adding unit 73 adds the prediction residual signal obtained from the inverse quantization / inverse orthogonal transform unit 72 and the prediction image signal generated by the intra prediction unit 74 to generate a decoded image signal. At the same time, the decoded image is stored in the image memory 75.

  The intra prediction unit 74 performs intra prediction that predicts the image pattern of the prediction target region from the decoded pixel group that is stored in the image memory 75 and close to the prediction target region, and generates a prediction image that is a result of the intra prediction. And sent to the first adder 73.

  The second addition unit 76 acquires information on the intra prediction mode optimal for the intra prediction estimated by the intra prediction mode estimation device 1 as intra prediction mode estimation information, and obtains the intra prediction mode estimation information from the entropy decoding unit 71. By adding to the obtained intra prediction mode difference information, the intra prediction mode information used in the intra prediction unit 74 is restored.

  Thereby, the intra prediction unit 74 can generate a prediction image using the restored intra prediction mode information.

  9 and 10, the same reference numerals are given to the same components as those of the conventional image encoding device 50 shown in FIG. 11 and the image decoding device 70 shown in FIG. 12. In the comparison between the image encoding device 50b and the image decoding device 70b of FIG. 5 and the conventional image encoding device 50 and the image decoding device 70, the image encoding device 50b and the image decoding device 70b of the present embodiment include an intra prediction unit 56. , 74 in that the intra prediction mode estimation device 1 is provided, and the image encoding device 50b of the present embodiment further includes a second subtraction unit 58, and the image decoding device 70b of the present embodiment. Is different in that a second addition unit 76 is provided. This difference will be described in detail below.

  As described above, the intra prediction unit 56 in the conventional image encoding device 50 assigns a prediction mode number to a prediction method for intra prediction as UDI (unrestricted digital information), and assigns a predetermined prediction mode number to the intra. By transmitting the prediction mode information to the image decoding device 70, the conventional image decoding device 70 can specify a prediction mode that defines a prediction method for intra prediction. For this reason, in addition to the image information, an intra prediction number for identifying M types of prediction modes as exemplified in FIG. 13 needs to be transmitted as intra prediction mode information.

  The image encoding device 50b and the image decoding device 70b of the present embodiment include an intra prediction mode estimation device 1, and the intra prediction mode estimation device 1 refers to decoded pixels stored in the image memories 55 and 75. As a pixel of a region, the image of each first region (reference image region) predicted according to a plurality of types of intra prediction modes is most similar in comparison with the image of the second region (target image region) in the reference region. By selecting one, an intra prediction mode in intra prediction is estimated, and a mode estimated value (an estimated value of an intra prediction number) is generated. The intra prediction mode estimation device 1 in the image encoding device 50b and the image decoding device 70b of the present embodiment outputs this mode estimation value as intra prediction mode estimation information.

  In the image encoding device 50b of the present embodiment, the second subtraction unit 58 uses the intra prediction mode information (intra prediction number) for identifying the M types of prediction modes output from the intra prediction unit 56, and the intra prediction mode estimation device. 1 generates intra prediction mode difference information obtained by subtracting the intra prediction mode estimation information (intra prediction number estimation value) output by 1, and sends the generated information to the entropy encoding unit 57. The entropy encoding unit 57 generates an encoded stream that transmits the intra prediction mode difference information in addition to the image information.

  On the other hand, in the image decoding device 70b of the present embodiment, the second addition unit 76 extracts the intra prediction mode difference information extracted from the encoded stream and the intra prediction mode estimation information (intra prediction) output from the intra prediction mode estimation device 1. The estimated number value) is added to restore the intra prediction mode information (intra prediction number) for identifying the M types of prediction modes, and the information is sent to the intra prediction unit 74. The intra prediction unit 74 generates a prediction image by the same processing as that of the conventional art.

  Therefore, according to the image encoding device 50b and the image decoding device 70b of the second embodiment, the transmission efficiency can be increased by transmitting the intra prediction mode difference information that is information with many zeros, and the same as the existing intra prediction. Performance can be maintained, and encoding efficiency can be improved.

  The intra-prediction mode estimation device 1, the image encoding devices 50a and 50b, or the image decoding devices 70a and 70b of the present embodiment can each function as a computer, and the computer can realize each component according to the present invention. These programs are stored in a memory provided inside or outside the computer. Intra prediction mode estimation according to the present embodiment is appropriately read from a memory in which a program describing processing contents for realizing the function of each component is controlled by control of a central processing unit (CPU) provided in the computer. The function of each component of the device 1, the image encoding devices 50a and 50b, or the image decoding devices 70a and 70b can be realized by a computer. Here, the function of each component may be realized by a part of hardware.

  According to the present invention, it is possible to estimate an optimal intra prediction mode in intra prediction direction prediction and average value prediction, which is useful for any application using intra prediction.

1, 1a, 1b Intra prediction mode estimation device 2 Reference region determination unit 3 First region pixel value prediction unit 4 Comparison unit 5 Intra prediction mode estimation unit 6 Prediction mode / pixel scanning control unit 7 Predicted image generation unit 11 Pixel scanning unit 12 Target pixel reading unit 13 Mode scanning unit 14 Reference pixel determining unit 15 Reference pixel reading unit 16 Comparison unit 17 Totaling unit 18 Prediction mode determining unit 19 Average value setting unit 20 Image memory 50, 50a, 50b Image encoding device 51 Subtracting unit ( First subtraction unit)
52 Orthogonal Transform / Quantizer 53 Inverse Quantization / Inverse Orthogonal Transformer 54 Adder (First Adder)
55 Image memory 56 Intra prediction unit 57 Entropy coding unit 58 Second subtraction unit 70, 70a, 70b Image decoding device 71 Entropy decoding unit 72 Inverse quantization / inverse orthogonal transform unit 73 Addition unit (first addition unit)
74 Intra-prediction unit 75 Image memory 76 Second addition unit 101 Prediction block 102 Reference block 103 Reference pixel 104 Prediction direction 105 Target pixel 106 Half line in the direction opposite to the prediction direction 107 Reference pixel obtained by a half line in the direction opposite to the prediction direction 108 A region of the target pixel in which there is a reference pixel to be compared by directional prediction 109 A line segment that passes through the center of the reference pixel group aligned horizontally and belongs to the region of the reference pixel group 110 passes through the center of the reference pixel group aligned vertically and Line segments belonging to the reference pixel group area

Claims (9)

An intra prediction mode estimation device for estimating a prediction mode in intra prediction for predicting an image pattern of a prediction target region from an image pattern of a reference region,
In accordance with each of a plurality of predetermined intra prediction modes, an image of a predetermined reference area is converted into an image of a first area predetermined in the reference area, and the reference area different from the first area. A reference area determination unit for dividing the image into a second area image determined in advance,
A first region pixel value prediction unit that generates a prediction value for the pixel value of the image of the first region associated according to the intra prediction mode from the pixel value of the image of the second region;
A comparison unit that generates a comparison value between the prediction value generated according to the intra prediction mode for the pixel value of the image of the second region and the pixel value of the image of the first region;
For each intra prediction mode, the comparison value of each pixel compared between the first region and the second region is aggregated to determine an evaluation value for each intra prediction mode, and an optimal intra prediction mode for intra prediction is estimated. An intra prediction mode estimation unit;
An intra prediction mode estimation apparatus comprising:   A prediction image generation unit configured to generate a prediction image in which each pixel of the prediction target region is predicted using a pixel value of the image of the first region based on the estimated intra prediction mode; The intra prediction mode estimation apparatus according to claim 1.   The intra prediction includes direction prediction, and when estimating the intra prediction mode related to the direction prediction, the first region pixel value prediction unit predicts the direction prediction of the intra prediction from the pixel value of the image of the second region. The intra prediction mode estimation apparatus according to claim 1, wherein a prediction value of the image of the first region associated by extrapolation in a direction opposite to the direction is generated.   The intra prediction includes average value prediction, and when the intra prediction mode is estimated by the average value prediction, the comparison unit calculates an average based on a pixel value of the image in the second region and a pixel value of the image in the first region. The intra prediction mode estimation apparatus according to any one of claims 1 to 3, wherein a comparison value with a value is generated. An image encoding device that encodes an image by performing intra prediction for predicting an image pattern of a prediction target region from an image pattern of a reference region,
A subtraction unit that takes a difference between the input image signal and the prediction image signal and generates a prediction residual signal that is a result of the difference;
An orthogonal transform / quantization unit that transforms the signal of the prediction residual using orthogonal transform to obtain a coefficient, quantizes the coefficient with a predetermined quantization width, and generates a transform coefficient that is a result of the quantization; ,
An entropy encoding unit that performs entropy encoding on the transform coefficient and generates an encoded stream;
An inverse quantization / inverse orthogonal transform unit that performs inverse transform of quantization and orthogonal transform on the transform coefficient;
An addition unit that adds the output signal of the inverse quantization / inverse orthogonal transform unit and the signal of the predicted image, and stores a decoded image as a result of the addition in an image memory;
It is the result of the intra prediction by performing intra prediction for predicting the image pattern of the prediction target region from the image pattern of the reference region made up of the decoded pixel group stored in the image memory and estimating the prediction mode in the intra prediction. The intra prediction mode estimation device according to claim 2, wherein the prediction image is generated;
An image encoding device comprising: An image encoding device that encodes an image by performing intra prediction for predicting an image pattern of a prediction target region from an image pattern of a reference region,
A first subtraction unit that takes a difference between an input image signal and a signal of a prediction image of intra prediction, and generates a prediction residual signal that is a result of the difference;
An orthogonal transform / quantization unit that transforms the signal of the prediction residual using orthogonal transform to obtain a coefficient, quantizes the coefficient with a predetermined quantization width, and generates a transform coefficient that is a result of the quantization; ,
An entropy encoding unit that performs entropy encoding on the transform coefficient and generates an encoded stream;
An inverse quantization / inverse orthogonal transform unit that performs inverse transform of quantization and orthogonal transform on the transform coefficient;
An addition unit that adds the output signal of the inverse quantization / inverse orthogonal transform unit and the signal of the predicted image, and stores a decoded image as a result of the addition in an image memory;
The intra prediction that predicts the image pattern of the prediction target region from the decoded pixel group that is stored in the image memory and close to the prediction target region is performed, the prediction image that is a result of the intra prediction is generated, and the intra An intra prediction unit that generates information about a prediction mode in prediction as intra prediction mode information;
The intra prediction mode estimation device according to claim 1, wherein the decoded image stored in the image memory is input.
Information about the intra prediction mode optimal for the intra prediction estimated by the intra prediction mode estimation device is acquired as intra prediction mode estimation information, and the difference between the intra prediction mode estimation information and the intra prediction mode information is calculated as an intra prediction mode difference. A second subtracting unit that generates information,
The said entropy encoding part adds the said intra prediction mode difference information, and produces | generates the said encoding stream, The image encoding apparatus characterized by the above-mentioned. An image decoding device that decodes an image by performing intra prediction for predicting an image pattern of a prediction target region from an image pattern of a reference region,
An entropy decoding unit that performs entropy decoding on an input encoded stream and extracts a transform coefficient obtained by performing orthogonal transform and quantization on a prediction residual signal in intra prediction on the image encoding side;
An inverse quantization / inverse orthogonal transform unit that performs inverse quantization and inverse orthogonal transform on the transform coefficient to generate a signal of the prediction residual;
An addition unit that adds the prediction residual signal and the prediction image signal to generate a decoded image signal, and stores the decoded image in an image memory;
The intra prediction mode estimation device according to claim 2, wherein a prediction image is generated from a decoded image stored in the image memory.
The intra prediction mode estimation device performs intra prediction that predicts an image pattern of a prediction target region from an image pattern of a reference region including a decoded pixel group stored in the image memory, and estimates a prediction mode in the intra prediction. An image decoding apparatus that generates the prediction image that is a result of the intra prediction. An image decoding device that decodes an image by performing intra prediction for predicting an image pattern of a prediction target region from an image pattern of a reference region,
Entropy decoding is performed on the input encoded stream, and the transform coefficient obtained by performing orthogonal transform and quantization on the prediction residual signal in the intra prediction on the image encoding side is extracted. An entropy decoding unit for extracting predetermined intra prediction mode difference information indicating a difference with respect to the intra prediction mode information as information on the prediction mode in the intra prediction;
An inverse quantization / inverse orthogonal transform unit that performs inverse quantization and inverse orthogonal transform on the transform coefficient to generate a signal of the prediction residual;
A first addition unit that generates a decoded image signal by adding the prediction residual signal and the intra prediction image signal, and stores the decoded image in an image memory;
The intra prediction mode estimation device according to claim 1, wherein the decoded image stored in the image memory is input.
Information about the intra prediction mode optimal for the intra prediction estimated by the intra prediction mode estimation device is acquired as intra prediction mode estimation information, and the intra prediction mode estimation information is added to the intra prediction mode difference information, and the intra prediction mode difference information is added. A second adder for restoring prediction mode information;
It is the result of the intra prediction performed by performing intra prediction that predicts the image pattern of the prediction target region using the decoded pixel group that is stored in the image memory and close to the prediction target region and the restored intra prediction mode information. An intra prediction unit for generating the predicted image;
An image decoding apparatus comprising: To a computer functioning as an intra prediction mode estimation device for estimating a prediction mode in intra prediction for predicting an image pattern of a prediction target region from an image pattern of a reference region,
In accordance with each of a plurality of predetermined intra prediction modes, an image of a predetermined reference area is converted into an image of a first area predetermined in the reference area, and the reference area different from the first area. Dividing into an image of the second region determined in advance,
Generating a prediction value for the pixel value of the image of the first region associated according to the intra prediction mode from the pixel value of the image of the second region;
Generating a comparison value between the prediction value generated according to the intra prediction mode for the pixel value of the image of the second region and the pixel value of the image of the first region;
For each intra prediction mode, the comparison value of each pixel compared between the first region and the second region is aggregated to determine an evaluation value for each intra prediction mode, and an optimal intra prediction mode for intra prediction is estimated. Steps,
A program for running