JP2006295408A - Image encoding apparatus and image encoding program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image encoding device capable of suppressing the occurrence of linear noise in a decoded image without selecting a prediction mode in which linear noise may occur.
An image encoding apparatus that divides an input image into blocks and encodes a difference from a prediction image generated according to a prediction mode selected by performing intra prediction, and is adjacent to an encoding target block Output from an adjacent prediction mode reference unit 401 that refers to a prediction mode of a block to be performed, a cost function switching unit 402 that outputs an instruction to switch to a predetermined cost function according to the referenced adjacent prediction mode, and a cost function switching unit 402 An intra prediction unit 101 that switches to a predetermined cost function according to the cost function switching command and generates a predicted image using a prediction mode selected based on a result of calculation processing using the cost function.
[Selection] Figure 1

Description

  The present invention relates to H.264. The present invention relates to an image encoding device and an image encoding program for realizing H.264 / AVC intra encoding.

H. H.264 / AVC is a moving picture coding standard established by ITU-T and ISO / IEC, and is a conventional coding standard such as MPEG-4 and H.264. Compared with H.263, the compression efficiency is about twice as high. H. H.264 / AVC is a moving image coding method, but it has been reported in Non-Patent Documents 1 to 3 and the like that it has excellent performance as a still image coding method, that is, intra coding. An image called an intra frame generated by performing intra coding is compressed by a single image, and is excellent in terms of random accessibility and image editability in image search. Hereinafter, H.C. An outline of H.264 / AVC intra coding will be described.
Kotaro Asai, “Intra-only video coding”, MPEG2003 / M10075 Kotaro Asai, “Intra-only video coding”, MPEG2003 / M10246 Kotaro Asai, “Intra-only video coding”, MPEG 2003 / M 10320 FIG. 1 is a block diagram of an encoding device that realizes H.264 / AVC intra coding. In the figure, an intra prediction unit 101 inputs a decoded image stored in a frame memory 109 and performs intra prediction. A process that uses a high correlation between adjacent pixels is performed, and an H. One mode is selected from a plurality of prediction modes defined by H.264 / AVC, and an image called a predicted image is generated.

  H. The plurality of prediction modes defined by H.264 / AVC include 13 modes of 9 modes of 4 × 4 blocks and 4 modes of 16 × 16 blocks, and an optimal prediction mode is selected from these. 4 is a conceptual diagram showing the prediction direction of 4 × 4 blocks, and FIG. 5 is a prediction direction of 16 × 16 blocks. As shown in FIG. 4, the 4 × 4 block has a total of 9 modes including 8 modes of directed prediction and 1 mode of undirected prediction, and as shown in FIG. 5, the 16 × 16 block has 3 directed predictions. There are a total of four modes, one mode and one undirected prediction mode. As shown in the figure, the number assigned to each prediction direction indicates the mode number.

  FIG. 6 is a flowchart showing processing performed by the intra prediction unit 101. First, in step 21, the intra prediction unit 101 performs intra prediction by dividing a macroblock into 4 × 4 blocks. As described above, the 4 × 4 block intra prediction performs a process of selecting from nine modes defined by the 4 × 4 block. Processing related to determination of the prediction mode of 4 × 4 blocks will be described later.

  Next, in step 22, the intra prediction unit 101 performs intra prediction by dividing the macroblock into 16 × 16 blocks. As described above, in the 16 × 16 block intra prediction, a process of selecting from the four modes defined by the 16 × 16 block is performed. The processing operation relating to the determination of the prediction mode of 16 × 16 blocks will also be described later.

  Finally, in step 23, the intra prediction unit 101 uses the modes determined in the 4 × 4 block and 16 × 16 block intra prediction, and determines which one to use. The determination method is not defined by the standard, but generally, the smaller prediction mode is selected by comparing the generated code amounts of the 4 × 4 block and the 16 × 16 block.

  FIG. 7 is a flowchart showing the 4 × 4 block intra prediction process performed by the intra prediction unit 101 in step 21. In step 31, the intra prediction unit 101 generates a prediction image using each prediction mode defined by 4 × 4 blocks. FIG. 8 is a diagram for explaining generation of a predicted image. In the figure, P is a predicted image, and X and Y are adjacent blocks located above or to the left of the predicted image. The pixels a to p belonging to the predicted image P are determined by extrapolating the pixels A to M in the direction indicated by the prediction mode. Specific determination examples are shown in FIGS.

  FIG. 9 is a prediction image generated when mode 0 indicating prediction in the vertical direction is selected from the prediction modes shown in FIG. In this case, the predicted image is generated by extrapolating the pixel values of adjacent pixels A to D in the vertical direction. FIG. 10 is a prediction image generated when the mode 1 indicating the prediction in the horizontal direction is selected from the prediction modes shown in FIG. In this case, the predicted image is generated by extrapolating the pixel values of the adjacent pixels I to L in the horizontal direction.

Next, in step 32, the intra prediction unit 101 calculates the cost required for encoding the prediction image generated in step 31. The definition formula of the cost function J prescribed | regulated by the reference software shown by the nonpatent literature 4 is shown.
Karsten Suhring, “H.264 / AVC Software Coordination”, http: // bs. hhi. de / suehring / tml

式１において、Ｄは入力画像と予測画像の類似度を示す歪み量、Ｒは予測モードをヘッダに記述する際に要するヘッダ符号量、ＱＰは圧縮率を示す量子化パラメータである。歪み量は、入力画像と予測画像のＳＡＤ（Ｓｕｍ ｏｆ Ａｂｓｏｌｕｔｅ Ｄｉｆｆｅｒｅｎｃｅｓ）やＳＳＤ（Ｓｕｍ ｏｆ Ｓｑｕａｒｅｄ Ｄｉｆｆｅｒｅｎｃｅｓ）で計算することができる。ＳＡＤとＳＳＤの定義式を式２と式３に示す。式２と式３において、Ｏは入力画像、Ｐは予測画像、ｉは画像の位置を示すインデックスである。

In Equation 1, D is a distortion amount indicating the similarity between the input image and the predicted image, R is a header code amount required for describing the prediction mode in the header, and QP is a quantization parameter indicating the compression rate. The amount of distortion can be calculated by SAD (Sum of Absolute Differences) or SSD (Sum of Squared Differences) of the input image and the predicted image. Formulas 2 and 3 show definition formulas for SAD and SSD. In Expressions 2 and 3, O is an input image, P is a predicted image, and i is an index indicating the position of the image.

また、ヘッダ符号量は、隣接ブロックの予測モード番号との相対関係により長さが変わる。具体的には、上または左に位置する隣接ブロックのモード番号の内、小さい番号の予測モードと同じモードを選択すると１ビット、その他のモードを選択すると４ビットになる。

Also, the length of the header code amount varies depending on the relative relationship with the prediction mode number of the adjacent block. Specifically, among the mode numbers of adjacent blocks located above or to the left, 1 bit is selected when the same mode as the prediction mode with a lower number is selected, and 4 bits when other modes are selected.

  FIG. This is a header syntax defined by H.264 / AVC. prev_Intra4x4_pred_mode_flag is 1-bit information, and takes a value of 1 when the same mode as the one with the smaller mode number of the adjacent block located above or to the left is selected, and takes a value of 0 when a different mode is selected. When the same mode is selected, since the decoding apparatus can know the prediction mode of the prediction target block from the prediction mode of the adjacent block, one bit is sufficient. On the other hand, when a different mode is selected, the encoding apparatus describes 8 modes obtained by subtracting the prediction modes of adjacent blocks from all 9 modes by 3 bits of rem_prev_Intra4 × 4_pred_mode_flag. This is information compression using a high correlation of prediction modes between adjacent blocks. As a specific example, when the mode numbers of the adjacent blocks located on the upper and left sides are 4 and 8, respectively, if the mode number 4 is selected in the prediction target block, it becomes 1 bit, and if the other is selected, it becomes 4 bits. Summarizing the above, Equation 1 shows that both the similarity D between the input image and the predicted image and the header code amount R required for description of the prediction mode are considered, and the intra prediction unit 101 uses 4 × 4 blocks. The cost calculation is performed using all nine modes defined in the above, and the prediction mode having the minimum value is selected.

  That is, in step 33, the intra prediction unit 101 confirms whether the costs for all nine modes defined by 4 × 4 blocks have been calculated. When the end of the calculation is confirmed, in step 34, the intra prediction unit 101 selects a prediction mode that minimizes the cost. Thus, the 4 × 4 block intra prediction process performed in step 21 is completed. Note that the 16 × 16 block intra prediction process is determined using only the distortion amount D that does not include the header code amount R (J = D) in Equation 1. This is because there are four prediction modes defined by 16 × 16 blocks, and even if any prediction mode is selected, it can be described by 2 bits. Except for the above points, the block is the same as the 4 × 4 block, and the intra prediction unit 101 calculates the cost of all four modes defined by the 16 × 16 block, and selects the prediction mode that minimizes the cost. The above is H. 2 is an overview of H.264 / AVC intra coding.

However, in the case of a low bit rate, when the prediction mode is determined using Equation 1, horizontal and vertical linear noise that does not exist in the input image (see FIG. 12) may occur in the decoded image (see FIG. 13). Non-patent document 5 points out.
Yada et al., “Verification of linear noise suppression algorithm in H.264 / MPEG-4 AVC”, IEICE General Conference, D-11-38, 2004 There are two possible causes of linear noise. One is in the cost function of Equation 1. In Equation 1, in the case of a low bit rate with a large quantization parameter QP, the coefficient before the header code amount R becomes large, and the distortion amount D indicating the similarity between the input image and the predicted image is not emphasized as a cost. That is, Formula 1 selects a prediction mode with a small header code amount with priority even if the prediction image is significantly different from the input image. That is, as described above, in 4 × 4 block intra prediction, when a prediction mode with a small mode number of an adjacent block is selected, the header code amount is 1 bit, and when other prediction modes are selected, a header is used. The code amount is 3 bits. In the case of a low bit rate with a large quantization parameter QP, the distortion amount D indicating the similarity between the input image and the predicted image is not emphasized as a cost, and the frequency of selecting a prediction mode with a small mode number of an adjacent block increases.

  FIG. 14 shows an example when a prediction mode with a small header code amount is selected. O is an input image, P is a predicted image, D is a difference image, X and Y are adjacent blocks, and MX and MY are prediction modes of adjacent blocks X and Y, respectively. In the figure, the predicted image P is generated by selecting the same vertical prediction mode as the prediction mode MY of the adjacent block Y. However, as shown in the figure, the second column (R1) from the left of the predicted image P has a pixel value 80 that is distant from the surroundings. When a predicted image having pixel values that are significantly different from the surroundings is used, there is a problem that linear noise occurs in the decoded image.

  That is, in the intra prediction unit 101, when the cost function of Equation 1 with the header code amount taken into account is used, the cost function is minimized by using the prediction mode with a small mode number of an adjacent block in which the header code amount is small. The prediction mode is selected as a result although there is a possibility that noise is generated. This is considered as one of the causes of the generation of linear noise.

  The other is related to quantization processing. The model expression of the decoded image is expressed by Expression 5. In Equation 5, O is an input image, P is a predicted image, W is orthogonal transform / quantization, and R is a decoded image. In Expression 5, when the difference image corresponding to OP is large in the quantization step, the texture information is deleted.

そのため、図１４の差分画像Ｄは、量子化ステップの大きい低ビットレートの場合、左から２列目（Ｒ２）のテクスチャ情報を失い、予測画像Ｐ上の垂直方向の線がそのまま復号画像に反映され、結果として、線状ノイズが発生してしまう。つまり、式１を用いて、線状ノイズが発生するおそれがあるにもかかわらず選択された予測モードを用いて生成される予測画像は、そのまま復号画像に反映されてしまい、結果として線状ノイズが発生してしまう。

Therefore, the difference image D in FIG. 14 loses texture information in the second column (R2) from the left in the case of a low bit rate with a large quantization step, and the vertical line on the predicted image P is directly reflected in the decoded image. As a result, linear noise is generated. In other words, the prediction image generated using the selected prediction mode in spite of the possibility of occurrence of linear noise using Equation 1, is reflected in the decoded image as it is, and as a result, linear noise is generated. Will occur.

  The present invention pays attention to such a problem, and provides an image encoding device and an image encoding program capable of suppressing the generation of linear noise in a decoded image without selecting a prediction mode in which linear noise may occur. The purpose is to provide.

  The invention according to claim 1 of the present invention is an image encoding apparatus that divides an input image into blocks and encodes a difference from a predicted image generated according to a prediction mode selected by performing intra prediction. An adjacent prediction mode reference unit that refers to a prediction mode of a block adjacent to the encoding target block, a cost function switching unit that outputs a command to switch to a predetermined cost function according to the referenced adjacent prediction mode, and the cost An intra prediction unit that switches to a predetermined cost function according to a cost function switching command output from the function switching unit and generates a predicted image using a prediction mode selected based on a result of a calculation process using the cost function. An image encoding device characterized by the above. Thereby, based on the prediction mode of the adjacent block referred when performing intra prediction, it can switch to a predetermined | prescribed cost function and can prevent selection of the prediction mode which may generate a linear noise in a prediction image, and decoding Generation of linear noise in the image can be suppressed.

  According to a second aspect of the present invention, in the image coding device according to the first aspect, the cost function switching unit is configured to calculate a pixel value of a pixel whose reference prediction mode is adjacent to the encoding target block. When it is a mode for generating a predicted image by extrapolating in a predetermined direction, it outputs a command to switch to a cost function consisting only of the distortion amount indicating the similarity between the input image and the predicted image, otherwise An instruction for switching to a cost function including a header code amount required for describing a prediction mode in a header and the similarity is output. As a result, a cost function that does not consider the header code amount is used when a prediction mode that may cause linear noise in the predicted image is selected, and a cost function that considers the header code amount otherwise. By using, it is possible to suppress the generation of linear noise and achieve high coding efficiency.

  According to a third aspect of the present invention, in the image encoding device according to the second aspect, the predetermined direction is a horizontal or vertical direction. Thereby, the linear noise of the horizontal or vertical direction which is easy to generate statistically can be suppressed.

  According to a fourth aspect of the present invention, in the image coding device according to the first aspect, the adjacent prediction mode reference unit predicts a prediction mode of a block located adjacent to the encoding target block and located above or to the left. It is characterized by referring to. As a result, H.C. It can be realized within the H.264 / AVC standard.

  According to a fifth aspect of the present invention, in the image coding device according to the fourth aspect, the adjacent prediction mode reference unit selects a prediction mode having a smaller mode number from among the blocks located above or to the left. It is characterized by reference. As a result, H.C. It can be realized within the H.264 / AVC standard.

  The invention according to claim 6 of the present invention is executed by a computer, so that an adjacent prediction mode reference unit that refers to a prediction mode of a block adjacent to an encoding target block, and a predetermined number according to the referred adjacent prediction mode. A cost function switching unit that outputs an instruction to switch to a cost function, a prediction that is selected based on a result of a calculation process using the cost function by switching to a predetermined cost function according to the cost function switching command output from the cost function switching unit An image encoding program that functions as an intra prediction unit that generates a predicted image using a mode. By causing the computer of the image encoding apparatus to execute this program, switching to a predetermined cost function is performed based on the prediction mode of an adjacent block referred to when performing intra prediction, and linear noise is generated in the predicted image. It is possible to prevent the selection of a predictive mode that may cause the occurrence of linear noise in the decoded image.

  As described above, according to the image encoding device and the image encoding program of the present invention, a prediction function is switched to a predetermined cost function based on a prediction mode of an adjacent block that is referred to when performing intra prediction. Selection of a prediction mode in which linear noise may occur can be prevented, and generation of linear noise in the decoded image can be suppressed.

  Hereinafter, the best mode for carrying out the present invention will be described. In addition, this invention is not limited to the range demonstrated by the following Example, In the range which does not deviate from the summary, it can change and implement suitably.

  Examples of the present invention will be described below. FIG. 1 shows an H.264 image coding program according to the present invention. 1 is a block diagram of an encoding device that realizes H.264 / AVC intra coding. In FIG. 1, 101 is an intra prediction unit, 102 is an orthogonal transform unit, 103 is a quantization unit, 104 is a lossless encoding unit, 105 is a buffer memory, 106 is a rate control unit, 107 is an inverse quantization unit, and 108 is an inverse unit. An orthogonal transform unit, 109 is a frame memory, 110 is a differentiator, 111 is an adder, 401 is an adjacent prediction mode reference unit, and 402 is a cost function switching unit. The main blocks of the present invention are a cost function switching unit 401, an adjacent prediction mode reference unit 402, and an intra prediction unit 101. The same reference numerals are used for the same components as those of the conventional coding apparatus shown in FIG. Is attached.

  First, H.C. Each processing function of the image coding apparatus that implements H.264 / AVC intra coding will be described in order. As described above, the intra prediction unit 101 inputs a decoded image stored in the frame memory 109 and performs intra prediction. A process that uses a high correlation between adjacent pixels is performed, and an H. One mode is selected from a plurality of prediction modes defined by H.264 / AVC, and an image called a predicted image is generated. The generated prediction image is output to the differentiator 110 and the adder 111 together with the prediction mode. In the orthogonal transform unit 102, the difference image between the input image and the predicted image output from the differentiator 110 is input, and the difference image is subjected to orthogonal transformation such as discrete cosine transformation or kernel-label transformation. The orthogonally transformed difference image is output to the quantization unit 103 as a transform coefficient. The quantization unit 103 receives the transform coefficient after the orthogonal transform output from the orthogonal transform unit 102 and the quantization parameter output from the rate control unit 106, and performs quantization on the transform coefficient. The quantized orthogonal transform coefficient is output to the lossless encoding unit 104 and the inverse quantization unit 107 as an encoded transform coefficient. The lossless encoding unit 104 receives the encoded transform coefficient output from the quantization unit 103, performs lossless encoding based on variable length coding or arithmetic coding on the transform coefficient, and generates a codeword. . The generated code word is output to the buffer memory 105. In the buffer memory 105, the code word output from the lossless encoding unit 104 is input, and the code word is stored. The stored codeword is referred to by the rate control unit 106 and then output as an encoded image. The rate control unit 106 determines the compression rate with reference to the buffer occupation amount of the buffer memory 105. The determined compression rate is converted into a quantization parameter and output to the quantization unit 103. The inverse quantization unit 107 receives the encoded transform coefficient output from the quantization unit 103, and performs inverse quantization on the transform coefficient. The inversely quantized encoded transform coefficient is output to the inverse orthogonal transform unit 108 as a decoded transform coefficient. The inverse orthogonal transform unit 108 receives the decoded transform coefficient output from the inverse quantum unit 107 and performs inverse orthogonal transform on the transform coefficient. The inverse transform transform decoded transform coefficient is output to the adder 111 as a decoded difference image. In the frame memory 109, the prediction image output from the intra prediction unit 101 and the addition image of the decoded difference image output from the inverse orthogonal transform unit 108 are input, and the addition image is stored as a decoding image. The stored decoded image is referred to when the intra prediction unit 101 performs intra prediction.

  The adjacent prediction mode reference unit 401 inputs the prediction mode of the decoded image stored in the intra prediction unit 101, and refers to the prediction mode of the adjacent block that has already been selected in the block adjacent to the encoding target block. The encoding target block is a block that is a reference target for determining a prediction mode, and refers to the smaller prediction mode among the mode numbers of adjacent blocks located above or to the left of the encoding target block. The referenced adjacent prediction mode is output to the cost function switching unit 402.

The cost function switching unit 402 receives the adjacent prediction mode output from the adjacent prediction mode reference unit 401, and switches to a predetermined cost function according to the referenced adjacent prediction mode. Specifically, when the adjacent prediction mode is the horizontal / vertical prediction mode, the cost function of Expression 4 with only the distortion amount D is switched. In other cases, the cost function is changed to the cost function of Equation 1 in consideration of the header code amount. The cost function switching is output to the intra prediction unit 101 as a cost function switching command.
<When adjacent prediction mode is horizontal / vertical prediction mode>

＜隣接予測モードが水平・垂直予測モードでない場合＞

<When adjacent prediction mode is not horizontal / vertical prediction mode>

図２は、コスト関数の計算処理を示すフロー図である。以下の処理フローは、４×４ブロック単位に処理が施される。はじめに、ステップ５１において、隣接予測モード参照部４０１は、符号化対象ブロックの上または左に位置する隣接ブロックの予測モードを参照する。具体的には、上または左の隣接ブロックのモード番号の内、小さい方の予測モードを参照する。参照された隣接予測モードは、コスト関数切替え部４０２へ出力される。

FIG. 2 is a flowchart showing a cost function calculation process. In the following processing flow, processing is performed in units of 4 × 4 blocks. First, in step 51, the adjacent prediction mode reference unit 401 refers to the prediction mode of the adjacent block located above or to the left of the encoding target block. Specifically, the smaller prediction mode is referred to among the mode numbers of the upper and left adjacent blocks. The referenced adjacent prediction mode is output to the cost function switching unit 402.

  Next, in step 52, the cost function switching unit 402 receives the adjacent prediction mode output from the adjacent prediction mode reference unit 401, and the pixel value of the pixel in which the referenced adjacent prediction mode is adjacent to the encoding target block. Is extrapolated in a predetermined direction to determine whether or not it is a mode for generating a predicted image. In the present embodiment, it is determined whether the horizontal or vertical prediction mode is statistically likely to occur.

  When the adjacent prediction mode is the horizontal / vertical prediction mode, in step 53, the cost function switching unit 402 switches to the cost function of Equation 4 defined only by the distortion amount D. In the case of a low bit rate, when the adjacent prediction mode is the horizontal / vertical prediction mode, the intra prediction unit 101 preferentially selects the horizontal / vertical prediction mode in which the header code amount is small, and as a result, linear This is because noise is generated. For example, when the same vertical prediction mode as the prediction mode MY of the adjacent block Y shown in FIG. 14 shown in the description of the prior art is selected, the vertical prediction mode is determined and Expression 4 is used. In this case, if the cost function is defined only by the distortion amount D as shown in Expression 4, it is determined that the similarity between the input image O and the predicted image P is low, and as a result, the vertical prediction that has been selected in the past is determined. A prediction image is not generated using the mode. As a result, linear noise has conventionally occurred in the prediction image generated using the vertical prediction mode, but in this embodiment, when a prediction mode that may cause linear noise in the prediction image is selected. Uses a cost function that does not consider the amount of header codes, can prevent the prediction mode from being selected, and can suppress the occurrence of linear noise in the decoded image.

  On the other hand, when the adjacent prediction mode information is other than the horizontal / vertical prediction mode, in step 54, the cost function switching unit 402 switches to the cost function of the above formula 1 in consideration of the header code amount. As a result, when a prediction mode that does not cause the occurrence of linear noise in the predicted image is selected, a cost function that considers the header code amount is used to suppress the generation of linear noise and achieve high coding efficiency. .

  Finally, in step 54, the intra prediction unit 101 calculates a cost based on the cost function after switching. Then, the intra prediction unit 101 performs cost calculation using all nine modes defined by 4 × 4 blocks, selects a prediction mode having a minimum value among them, and generates a prediction image using the prediction mode. .

  As described above, according to the present embodiment, a prediction mode that is switched to a predetermined cost function based on a prediction mode of an adjacent block that is referred to when performing intra prediction, and that linear noise may occur in the prediction image. Selection can be prevented, and the occurrence of linear noise in the decoded image can be suppressed.

  The image coding apparatus of the present invention can realize a still compressed image with high image quality, and thus has excellent random accessibility and editability. Therefore, application to an image storage device or an image editing device having an image search function is useful.

The block diagram which shows the whole structure of the image coding apparatus of this invention The flowchart which shows the calculation process of the cost function of this invention The block diagram which shows the whole structure of the conventional image coding apparatus. The figure which shows the prediction mode number and its direction of 4x4 block The figure which shows the prediction mode number and its direction of 16x16 block Flow chart showing processing performed by intra prediction unit Flow chart showing 4 × 4 block intra prediction processing The figure which shows the encoding object block and the block adjacent to it The figure which shows the prediction image produced | generated when the vertical prediction mode is selected The figure which shows the prediction image produced | generated when horizontal prediction mode is selected H. The figure which shows the syntax of the header prescribed | regulated by H.264 / AVC Diagram showing input image (no linear noise is generated) Diagram showing decoded image (linear noise occurs) The figure which shows the example at the time of selecting the prediction mode with small header code amount

Explanation of symbols

101 Intra Prediction Unit 401 Adjacent Prediction Mode Reference Unit 402 Cost Function Switching Unit

Claims (6)

An image encoding device that divides an input image into blocks and encodes a difference from a prediction image generated according to a prediction mode selected by performing intra prediction,
An adjacent prediction mode reference unit that refers to a prediction mode of a block adjacent to the encoding target block;
A cost function switching unit that outputs a command to switch to a predetermined cost function according to the referenced adjacent prediction mode;
An intra prediction unit that switches to a predetermined cost function according to a cost function switching command output from the cost function switching unit and generates a prediction image using a prediction mode selected based on a result of a calculation process using the cost function; ,
An image encoding device comprising: The cost function switching unit, when the referenced prediction mode is a mode for generating a predicted image by extrapolating pixel values of pixels adjacent to the encoding target block in a predetermined direction, an input image and a predicted image Outputs a command to switch to a cost function consisting only of the distortion amount indicating the similarity of
2. The image encoding apparatus according to claim 1, wherein in other cases, an instruction for switching to a cost function including a header code amount required for describing a prediction mode in a header and the similarity is output.   3. The image encoding apparatus according to claim 2, wherein the predetermined direction is a horizontal or vertical direction.   The image coding apparatus according to claim 1, wherein the adjacent prediction mode reference unit refers to a prediction mode of a block adjacent to the encoding target block and positioned on the upper or left side.   The image coding apparatus according to claim 4, wherein the adjacent prediction mode reference unit refers to a prediction mode having a smaller mode number among the blocks located above or to the left. By running the computer,
An adjacent prediction mode reference unit that refers to a prediction mode of a block adjacent to the encoding target block;
A cost function switching unit that outputs a command to switch to a predetermined cost function according to the referenced adjacent prediction mode;
And an intra prediction unit that switches to a predetermined cost function according to a cost function switching command output from the cost function switching unit and generates a predicted image using a prediction mode selected based on a result of calculation processing using the cost function An image encoding program that functions as:

Images