JP2005348318A - Image predictive coding method, image predictive decoding method, image coding device, image decoding device, image coding program,image decoding programs, and computer readable record medium recording those programs - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To correct a prediction signal for an invalid prediction mode to raise a selection probability of the invalid prediction mode having a very small selection probability, in order to improve the coding efficiency. <P>SOLUTION: A prediction signal generation portion 15 generates a prediction signal from an encoded image signal in a memory 14. An invalid prediction mode determination portion 16 detects an invalid prediction mode from the generated prediction signal for each prediction mode. A prediction signal correction portion 17 corrects the prediction signal in the detected invalid prediction mode. A prediction mode determination portion 18 computes a coding cost for each prediction mode using the corrected prediction signal and selects a prediction mode with a minimum cost. Then, the prediction signal of the selected prediction mode is sent to a subtractor 10, and a prediction mode information is sent to a coding portion 19. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a video encoding method having a plurality of prediction modes, and in particular, an image predictive encoding method, an image predictive decode method, and a method for encoding / decoding by correcting a prediction signal of a prediction mode having a very low selection probability, The present invention relates to an image encoding device, an image decoding device, an image encoding program, an image decoding program, and a computer-readable recording medium on which these programs are recorded.

  ITU-TH. In H.264 (hereinafter referred to as H.264), when performing intra-frame coding, a prediction signal is generated from information of an already-encoded neighboring block of a block to be encoded, and a prediction residual is encoded (non-coding) Patent Document 1).

  Intraframe coding is a method for predicting an image signal of an encoding target block from an encoded image signal in the same frame. H. In H.264, intra-frame prediction is performed using 4 × 4 pixels (4 × 4 blocks) or 16 × 16 pixels (16 × 16 blocks) as a unit for performing prediction (hereinafter referred to as a prediction unit). Four to nine types of prediction directions can be selected for each prediction unit.

  FIG. It is a figure explaining the prediction mode of 16 * 16 size of H.264. When predicting an image signal of an encoding target block surrounded by a thick line from an already encoded image signal indicated by a hatched line in FIG. 8A, four types from FIG. 8B to FIG. There is a prediction mode.

  FIG. 8B shows a prediction mode for vertical prediction in which the image signal of the encoding target block is predicted from the already encoded image signal of the column existing above the encoding target block. FIG. 8C shows a prediction mode for horizontal prediction in which the image signal of the encoding target block is predicted from the already encoded image signal of the column existing on the side surface portion of the encoding target block. FIG. 8D shows a prediction mode for DC prediction in which the image signal of the block to be encoded is predicted from the average value of the already encoded image signals. FIG. 8E shows a prediction mode of Plane prediction in which a prediction value is generated according to a position from an already encoded image signal.

  In FIG. 2 shows a prediction mode of H.264 4 × 4 size. The encoding target 4 × 4 blocks a to p in FIG. 9A are predicted from the already encoded image signals A to L and O. FIG. 9B shows the prediction direction excluding the DC prediction mode. According to the prediction direction corresponding to the prediction mode, the values of a to p of the 4 × 4 block to be encoded are predicted from the already encoded image signal.

  FIG. 10 is a diagram illustrating a configuration of a conventional image encoding device. As shown in FIG. 10, the image encoding device 3 includes a subtracter 31, a prediction residual signal encoding unit 32, a prediction residual signal decoding unit 33, an adder 34, a memory 35, a prediction signal generation unit 36, a prediction mode. A determination unit 37 and an encoding unit 38 are provided.

  In the image encoding device 3, the prediction signal generation unit 36 generates a prediction signal from the encoded image signal in the memory 35. In the prediction mode determination part 37, the encoding cost of each prediction mode is calculated, and the prediction mode of the minimum cost is selected. As will be described in detail later, the coding cost is composed of the prediction residual power of the prediction signal and the input signal, and the overhead code amount of each prediction mode.

  The prediction mode determination unit 37 sends a prediction signal of the selected prediction mode to the subtracter 31 and sends prediction mode information to the encoding unit 38. The subtractor 31 obtains a prediction residual signal that is the difference between the input signal and the prediction signal, and the prediction residual signal encoding unit 32 performs processing such as orthogonal transformation and quantization of the prediction residual signal.

  The prediction residual signal decoding unit 33 performs inverse quantization and inverse orthogonal transform on the output signal from the prediction residual signal encoding unit 32 to decode the prediction residual signal. The adder 34 adds the decoded prediction residual signal and the prediction signal to generate a decoded image. The generated decoded image is stored in the memory 35 in order to generate a prediction signal.

  FIG. 11 is a diagram illustrating a configuration of a conventional image decoding apparatus. The image decoding device 4 includes an encoded data decoding unit 41, an adder 42, a memory 43, and a prediction signal generation unit 44.

  In the image decoding device 4, the encoded data decoding unit 41 decodes the encoded data into a prediction residual signal and prediction mode information. The prediction signal generation unit 44 generates a prediction signal from the decoded image signal in the memory 43 based on the prediction mode information from the encoded data decoding unit 41. The adder 42 adds the decoded prediction residual signal and the prediction signal to generate a decoded image.

  The prediction mode determination process on the encoding side will be described more specifically. FIG. 12 shows a flow of prediction mode determination processing on the encoding side. H. In the H.264 reference software, when determining the prediction unit and the prediction direction, the intraframe coding mode is determined in consideration of the prediction residual signal power and the code amount of the overhead part in each mode (see Non-Patent Document 2). .

That is, a prediction signal is generated (step S41), the code amount of the overhead part is calculated (step S42), and the prediction residual power is calculated (step S43). Then, the encoding cost J _i of the prediction mode number _i is calculated from the following equation (1) (step S44).

J _i = D _i + λR _i (1)
Here, D _i is the prediction residual power, and R _i is the predicted generated code amount. λ is a value that varies depending on the quantization step size. For each prediction mode, the encoding cost J _i is calculated, the prediction mode number of the minimum cost is obtained, and the prediction mode is determined (step S45). The obtained prediction mode number is transmitted to the decoding side.

  Next, processing on the decoding side will be described. FIG. 13 shows a flow of processing on the decoding side. On the decoding side, the prediction mode number included in the encoded bitstream is decoded (step S51), a prediction signal is generated in the prediction mode indicated by the prediction mode number (step S52), and a decoding process is performed (step S53).

Thus, encoding efficiency can be improved by encoding in the prediction mode of the minimum cost.
ITU-T Rec. H.264, “Advanced video coding for generic audiovisual services”, 2003. H.264 reference software, [online], Internet, <URL: http: //bs.hhi.de/suehring/tml/download/>

  In the conventional method, in order to improve the coding efficiency, the prediction mode is selected in consideration of the overhead code amount. However, the prediction mode to be selected may be limited by weighting with the overhead code amount. In the following, prediction mode determination when only prediction residual power is used and when overhead code amount is considered will be described.

First, the case where the prediction mode is determined from the prediction residual power will be described. Now, an N × N size block is considered as an N × N-dimensional signal space. For example, a 4 × 4 size block has a 16-dimensional signal space vector. Input signal encoding target block vector ↑ s, when a prediction signal of each of the prediction modes and prediction signal vector ↑ p _i, prediction residual power can be obtained from the distance between the two vectors.

d _i = | ↑ s− ↑ p _i | (2)
It is to detect a prediction signal vector ↑ p _i the distance in the above equation is minimized to determine a prediction mode by paying attention to the predictive residual power. In other words, the prediction signal vector ↑ _pi is a representative vector of the divided area, and the prediction mode determination is equivalent to detecting the area to which the input signal vector ↑ s belongs.

  FIG. 14 shows an example of area division in the case of two dimensions. A point indicated by a black circle in the figure represents a representative vector. When focusing only on the predicted residual power, the distance from the representative point of the adjacent region is equal at the boundary of each region.

Next, a case where the prediction mode is determined in consideration of the overhead code amount will be described. In the conventional method, since the encoding cost is obtained from the overhead code amount in addition to the prediction residual power, the boundary of each region is a position where the encoding cost J _i is equal.

FIG. 15 shows an example of area division based on the coding cost J _i . In this example, for simplicity, the area is divided into two areas A and B, and the representative points are also two points A and B. The encoding cost of each prediction mode is as follows:

J _a = D _a + λR _a (3)
J _b = D _b + λR _b (4)
The boundary of the region is that J _a and J _b are equal,
D _a + G _ab = D _b (5)
It becomes. Here, G _ab represents the cost difference λ (R _a −R _b ) depending on the amount of overhead codes in the prediction modes A and B.

From equation (5), when the amount of overhead code in prediction mode A is larger than that in prediction mode B (G _ab > 0), the position of the boundary is G _{ab as} compared with region division using only prediction residual power. It can be seen that the region A is shifted by the amount.

Further, when the overhead code amount cost difference G _ab is larger than the half distance D _ab / 2 between the representative points A and B (G _ab > D _ab / 2), the representative point A is expressed as follows. Are included in the region B.

That is, in FIG. 15, if the boundary line X is a boundary between the region A and the region B when focusing only on the predicted residual power, if 0 <G _ab <D _ab / 2, the region A and the boundary between the region B, the boundary line Y in the position shifted to the G _ab amount corresponding region a side of the boundary line X. When G _ab further increases and G _ab > D _ab / 2, the boundary between the region A and the region B becomes the boundary line Z, and the representative point A is included in the region B.

Thus, the possibility that the prediction mode A is selected decreases as the overhead code amount cost difference G _ab increases. Even if the prediction mode A is selected in a situation where the representative point A is included in the region B, the position is a place away from the representative point A, and the prediction residual power increases.

  Also, when the correlation between adjacent pixels is very high, such as in a flat region, there is a high possibility that even if a prediction signal is generated in a plurality of prediction modes, the prediction signal is almost the same. In such a case, the distance between representative vectors is 0, and a prediction mode with a large overhead cost is not selected at all.

  As described above, when the difference in overhead code amount is large, or when the correlation between the prediction signals is high and the distance between the representative points is short, a prediction mode with a low probability of being selected or not selected at all (hereinafter referred to as an invalid prediction mode). appear. In the conventional method, a code is assigned to the invalid prediction mode, so that there is a problem that the coding efficiency is lowered.

  An object of the present invention is to solve the above-described problems of the prior art and increase the selection probability of an invalid prediction mode having a very low selection probability, thereby improving the coding efficiency.

  In the conventional method, the coding cost is obtained in consideration of the overhead code amount and the prediction mode is selected, so that there is a problem that an invalid prediction mode occurs. In the present invention, the prediction signal of the invalid prediction mode is corrected to improve the selection probability of the invalid prediction mode, thereby improving the coding efficiency.

  FIG. 1 shows an example of the configuration of an image encoding device according to the present invention. The image encoding device 1 includes a subtracter 10, a prediction residual signal encoding unit 11, a prediction residual signal decoding unit 12, an adder 13, a memory 14, a prediction signal generation unit 15, an invalid prediction mode determination unit 16, a prediction signal. A correction unit 17, a prediction mode determination unit 18, and an encoding unit 19 are provided.

  In the image encoding device 1 shown in FIG. 1, the prediction signal generation unit 15 generates a prediction signal from the encoded image signal in the memory 14. The invalid prediction mode determination unit 16 detects the invalid prediction mode from the generated prediction signal of each prediction mode. The prediction signal correction unit 17 corrects the prediction signal in the invalid prediction mode from the detected invalid prediction mode and the prediction signal in each prediction mode. The corrected all prediction signal is transmitted to the prediction mode determination unit 18. A method for correcting the prediction signal in the invalid prediction mode will be described in detail later.

  The prediction mode determination unit 18 calculates the encoding cost of each prediction mode using the corrected prediction signal, and selects the prediction mode with the minimum cost. The coding cost includes the prediction residual power of the prediction signal and the input signal, and the overhead code amount of each prediction mode.

  The prediction mode determination unit 18 sends the prediction signal of the selected prediction mode to the subtracter 10 and the prediction mode information to the encoding unit 19. The subtracter 10 obtains a prediction residual signal that is a difference between the input signal and the prediction signal. The prediction residual signal encoding unit 11 performs processing such as orthogonal transformation and maximization of the prediction residual signal. The prediction residual signal decoding unit 12 performs inverse quantization and inverse orthogonal transform on the output signal of the prediction residual signal encoding unit 11 to decode the prediction residual signal. The adder 13 adds the decoded prediction residual signal and the prediction signal to generate a decoded image. The generated decoded image is stored in the memory 14 in order to generate a prediction signal.

  The encoding unit 19 performs variable length encoding on the output of the prediction residual signal encoding unit 11 and the prediction mode information, and outputs the result as encoded data.

  FIG. 2 is a diagram showing an example of the configuration of the image decoding apparatus according to the present invention. The image decoding device 2 includes an encoded data decoding unit 20, an adder 21, a memory 22, a prediction signal generation unit 23, an invalid prediction mode determination unit 24, and a prediction signal correction unit 25.

  In the image decoding apparatus 2 shown in FIG. 2, the encoded data decoding unit 20 decodes the encoded data into a prediction residual signal and prediction mode information. The prediction signal generation unit 23 generates a prediction signal for each prediction mode from the decoded image signal in the memory 22. The invalid prediction mode determination unit 24 detects the invalid prediction mode from the generated prediction signal of each prediction mode. The prediction signal correction unit 25 corrects the detected prediction signal in the invalid prediction mode. Further, the prediction signal correction unit 25 outputs a corrected prediction signal based on the prediction mode information. The adder 21 adds the decoded prediction residual signal and the prediction signal to generate a decoded image.

FIG. 3 shows a processing flow on the encoding side of the present invention. First, a prediction signal is generated for each prediction mode (step S1), and an overhead cost that is a cost of an overhead code amount is calculated (step S2). A prediction signal of a prediction mode number i prediction signal vector (representative vector) ↑ p _i, the overhead costs and g _i. And a representative vector ↑ p _i and overhead costs g _i for each of these prediction modes, check for invalid prediction mode (step S3).

  When the invalid prediction mode exists, the prediction signal as the representative vector of the invalid prediction mode is corrected (step S4). When the invalid prediction mode does not exist, the process is the same as the conventional process, and the process proceeds to step S5.

After all the corrections are completed, after calculating the prediction residual power of each prediction mode for the encoding target block (step S5), the encoding cost J _i is calculated (step S6), and the minimum encoding cost and A prediction mode number is obtained to determine a prediction mode (step S7), and an encoding process is performed. The obtained prediction mode number is transmitted to the decoding side.

  Next, processing on the decoding side will be described. In the conventional method, the decoding side generates only the prediction signal having the prediction mode number indicated in the encoded bit stream. This is because the prediction signal can be uniquely generated from the prediction mode number because the prediction signal correction is not performed.

  In the present invention, in the same way as on the decoding side, the decoding side generates prediction signals in all available prediction modes, and corrects the prediction signals as necessary. FIG. 4 shows a processing flow on the decoding side. The decoding side also performs prediction signal correction processing in the same manner as the processing flow on the encoding side described above.

That is, after generating the prediction signal (step S11), the overhead cost is calculated (step S12), the prediction signal of the prediction mode number i is the representative vector ↑ p _i , the overhead cost is g _i , and ↑ p _i and g _i From this, the presence / absence of the invalid prediction mode is checked (step S13). If the invalid prediction mode exists, the representative vector (prediction signal) of the invalid prediction mode is corrected (step S14). If the invalid prediction mode does not exist, the process proceeds to step S15. Next, the prediction mode number is decoded (step S15), and decoding processing is performed based on the decoded prediction mode number (step S16).

  Next, determination of the invalid prediction mode and correction of the prediction signal will be described. FIG. 5 shows a flow of invalid prediction mode determination and prediction signal correction. Let M be the number of prediction modes. First, a list is created by rearranging each prediction mode from 0 to (M−1) in ascending order of overhead cost (step S31). Next, after initial value 0 is set to i (step S32), j = i + 1 is set (step S33), and the following steps S34 to S39 are executed.

First, the cost of the prediction mode number i with the lowest overhead cost in the list is compared with the cost of the remaining prediction mode number j (> i), and the invalid prediction mode is determined (step S34). The criterion for determining the invalid prediction mode in step S34 is that the representative vector ↑ p _j of the remaining prediction mode number j is included in the prediction mode number i area of the minimum overhead cost in the list. That is, when the following equation is established, the prediction mode number j is determined as the invalid prediction mode.

g _j −g _i > | ↑ p _j − ↑ p _i | / 2 (6)
When the prediction mode number j is determined to be the invalid prediction mode, the prediction signal of the prediction mode number j determined to be the invalid prediction mode is corrected (step S35).

  In step S34, the cost of the prediction mode number i of the minimum overhead cost in the list and the cost of the remaining prediction mode number j (> i) are compared until j ≧ M (step S36). Repeat while adding 1 to j (step S38).

  When j ≧ M, and when the comparison between the prediction mode with the minimum overhead cost at a certain point in time and the remaining prediction modes ends (step S36), until i ≧ M−1 (step S37), the prediction mode number 1 is added to i (step S39), and the processing after step S33 is continued. That is, the prediction mode with the minimum overhead cost is sequentially removed from the list, and the invalid prediction mode determination and the prediction signal correction process for the remaining prediction modes are repeated. When the correction of the prediction signals in all invalid prediction modes is completed (step S37), the correction processing of the prediction signals in the invalid prediction mode is finished.

  There are several methods for correcting the prediction signal in the invalid prediction mode, that is, the representative vector correction method as described below.

[Correction by correction vector]
↑ p ' _i = ↑ p _i + ↑ c _k (7)
As shown in the above equation (7), the correction vector ↑ c _k is added to the representative vector ↑ p _i to obtain a new representative vector ↑ p ′ _i . For example, a predetermined fixed vector is used as the correction vector ↑ c _k . Alternatively, as shown in the following equation (8), by multiplying the correction coefficient c to the difference vector between the representative vector ↑ p ₀ of a prediction mode of a region including the representative vector ↑ p _i invalid prediction mode and the invalid prediction mode , A new representative vector ↑ p ′ _i may be used.

↑ p ' _i = c · (↑ p _i − ↑ p ₀ ) (8)
[Correction by correction coefficient]
↑ p ' _i = c ・ ↑ _pi (9)
As shown in the above equation (9), a new representative vector is obtained by multiplying the representative vector in the invalid prediction mode by the correction coefficient c. As the correction coefficient c, a constant given in advance, the distance between both representative vectors of the invalid prediction mode and the prediction mode of the region including the invalid prediction mode, or the like is used.

[Replace representative vector]
↑ p ' _i = ↑ ck (10)
As shown in the above equation (10), the representative vector in the invalid prediction mode is replaced with another vector. As this replacement vector, a predetermined fixed vector or a midpoint between regions where the distance between representative vectors is maximum is used.

  It is also possible to combine these methods. In addition to intra-frame coding, the present invention is useful not only for coding methods having a plurality of prediction modes, but also for coding methods for determining a prediction mode by a cost other than prediction residual power such as overhead cost.

  According to the present invention, it is possible to increase the selection probability by detecting a prediction mode that is not or hardly selected even though a code is assigned and correcting the prediction signal. it can. Thereby, encoding efficiency can be improved.

Embodiments of the present invention will be described below. In the present embodiment, an intra-frame coding scheme that can use a plurality of prediction modes is assumed. The prediction mode selection considers the overhead code amount of each prediction mode. Note that the invalid prediction mode is determined only for the prediction mode with the minimum overhead cost. The condition of disabled prediction mode, and if the representative vector ↑ p _i is included in the area of the prediction mode of the minimum overhead cost.

First, the encoding side will be described. FIG. 6 shows a processing flow on the encoding side. First, a prediction signal for each prediction mode is generated (step S61). A prediction signal for each prediction mode generated the representative vector ↑ p _i. Calculating the overhead costs g _i for each prediction mode (step S62), obtains the minimum value. The representative vector of the prediction mode of the minimum overhead cost is ↑ p ₀ and the overhead cost is g ₀ .

Then, from the representative vector ↑ p _i and overhead cost g _i of each of the prediction mode number i (> 0), examine the presence or absence of the invalid prediction mode (step S63). If the invalid prediction mode exists, the representative vector ↑ _pi (prediction signal) of the invalid prediction mode is corrected (step S64).

When the invalid prediction mode does not exist, the process is the same as the conventional one, and the process proceeds to step S65. After all the correction processes are completed, the prediction residual power is calculated for the encoding target block (step S65), and the encoding cost J _i for each prediction mode is calculated (step S66). The prediction mode with the lowest cost is selected to determine the prediction mode (step S67), and the encoding process is performed.

  Next, processing on the decoding side will be described. Also on the decoding side, a prediction signal for each prediction mode is generated and the prediction signal is corrected in the same manner as the encoding side. FIG. 7 shows a processing flow on the decoding side. First, the overhead cost of each prediction mode is obtained (step S71). Next, the prediction mode number is decoded (step S72), and it is determined whether the decoded prediction mode is the prediction mode with the minimum overhead cost (step S73).

  If the decoded prediction mode is the prediction mode with the minimum overhead cost, the decoding process is continued without performing the correction process (step S77). In other prediction modes, after generating a prediction signal for each prediction mode (step S74), the presence / absence of the invalid prediction mode is checked (step S75).

  If the invalid prediction mode does not exist, the process proceeds to step S77. If the invalid prediction mode exists, the representative vector (prediction signal) of the invalid prediction mode is corrected (step S76) and the decoding process is performed (step S77). ).

Next, an example of an invalid prediction mode determination and prediction signal correction method performed at the time of encoding and decoding will be described. First, the size of the representative vector ↑ p ₀ of a prediction mode of the minimum overhead cost differential vector of the representative vector ↑ p _i of each of the prediction modes _{d i = | ↑ p i -} ↑ p 0 | a seek.

At this time, the prediction mode in which d _i / 2 <g _i −g ₀ is the invalid prediction mode. Of the prediction modes determined to be the invalid prediction mode, the prediction signal is corrected from the one with a small overhead code amount. The correction method is as follows. Note that the corrected representative vector is ↑ p ′ _i .

[When d _i = 0]
↑ p ' _i = ↑ c _k + ↑ p ₀ (11)
Here, ↑ c _k is a predetermined correction vector. A plurality of correction vectors ↑ c _k are prepared, and when there are a plurality of invalid prediction modes with d _i = 0, different correction vectors ↑ c _k are used.

[When d _i ≠ 0]
↑ p ′ _i = (1+ (g _i −g ₀ ) / d _i ) (↑ p _i − ↑ p ₀ ) (12)
Here, g _i is the overhead cost of the invalid prediction mode. As a result, the invalid prediction mode can be determined and the prediction signal can be corrected.

  In this embodiment, the invalid prediction mode determination is performed only for the prediction mode with the minimum overhead cost. However, it is more efficient to sequentially perform the invalid prediction mode determination with a prediction mode with a low overhead cost.

Further, in the present embodiment, as a criterion for disable prediction mode, although the case where the representative vector ↑ p _i is included in the area of the prediction mode of the minimum overhead cost, regardless of the representative vector, in the region of each of the prediction modes You may determine by an occupation area.

  The above image predictive encoding and image predictive decoding processes can be realized not only by hardware and firmware but also by a computer and a software program, and the program can be stored on a computer-readable recording medium. It can be recorded and provided through a network.

It is a figure which shows an example of a structure of an image coding apparatus. It is a figure which shows an example of a structure of an image decoding apparatus. It is a figure which shows the processing flow by the side of an encoding. It is a figure which shows the processing flow by the side of decoding. It is a figure which shows the flow of invalid prediction mode determination and prediction signal correction | amendment. It is a figure which shows the processing flow by the side of the encoding of embodiment of this invention. It is a figure which shows the processing flow by the side of decoding of embodiment of this invention. H. It is a figure explaining the prediction mode of 16 * 16 size of H.264. H. It is a figure which shows the prediction mode of H.264 4x4 size. It is a figure which shows the structure of the conventional image coding apparatus. It is a figure which shows the structure of the conventional image decoding apparatus. It is a figure which shows the processing flow by the side of an encoding. It is a figure which shows the processing flow by the decoding side. It is a figure which shows the example of the area | region division in the case of two dimensions. Is a diagram showing an example of area division by the encoding cost J _i.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1,3 Image coding apparatus 2,4 Image decoding apparatus 10,31 Subtractor 11,32 Prediction residual signal encoding part 12,33 Prediction residual signal decoding part 13,21,34,42 Adder 14,22, 35, 43 Memory 15, 23, 36, 44 Prediction signal generation unit 16, 24 Invalid prediction mode determination unit 17, 25 Prediction signal correction unit 18, 37 Prediction mode determination unit 19, 38 Encoding unit 20, 41 Encoded data decoding Part

Claims (16)

An image predictive coding method for coding a picture by determining a prediction mode of a target block from a plurality of prediction modes,
Generating a prediction signal for each prediction mode;
Calculating overhead cost from overhead code amount for each prediction mode;
Detecting an invalid prediction mode having a low selection probability according to a predetermined criterion from the calculated relationship between the overhead cost of each prediction mode and the prediction signal;
Correcting the detected invalid prediction mode prediction signal;
Calculating a coding cost for each prediction mode from the prediction residual power and overhead code amount of each prediction mode;
And a step of determining a prediction mode that minimizes the encoding cost as a prediction mode of the encoding target block. The image predictive encoding method according to claim 1,
In the step of detecting the invalid prediction mode,
Detecting a prediction mode with a minimum overhead cost;
And a step of detecting an invalid prediction mode by comparing a prediction mode having a minimum overhead cost with another prediction mode. An image predictive coding method for coding a picture by determining a prediction mode of a target block from a plurality of prediction modes,
Generating a prediction signal for each prediction mode;
Calculating overhead cost from overhead code amount for each prediction mode;
Rearranging prediction modes in ascending order of overhead cost to generate a list;
Detecting an invalid prediction mode by comparing a prediction mode with a minimum overhead cost in the list with the remaining prediction modes;
Correcting the detected invalid prediction mode prediction signal;
Removing the prediction mode with the lowest overhead cost from the list after correction;
The number of prediction modes registered in the list includes the step of detecting the invalid prediction mode, the step of correcting the prediction signal of the prediction mode detected as the invalid prediction mode, and the step of deleting the prediction mode from the list. Step to repeat until
Calculating a coding cost for each prediction mode from the prediction residual power and overhead code amount of each prediction mode;
And a step of determining a prediction mode that minimizes the encoding cost as a prediction mode of the encoding target block. The image predictive encoding method according to any one of claims 1 to 3,
In the step of detecting the invalid prediction mode,
Calculating a distance between prediction signal vectors, which is a distance between a prediction signal vector of a reference prediction mode and a prediction signal vector of a test target prediction mode;
Calculating an overhead cost difference that is a difference between the overhead cost of the reference prediction mode and the overhead cost of the inspection target prediction mode;
The half of the calculated inter-prediction signal vector distance is compared with the overhead cost difference, and if the half of the inter-prediction signal vector distance is smaller than the overhead cost difference, the inspection target prediction mode is detected as an invalid prediction mode. An image predictive encoding method characterized by the above. The image predictive encoding method according to any one of claims 1 to 4,
In the step of correcting the prediction signal in the invalid prediction mode,
An image predictive coding method, comprising: correcting a prediction signal in an invalid prediction mode using a predetermined correction vector or correction coefficient. The image predictive encoding method according to any one of claims 1 to 4,
In the step of correcting the prediction signal in the invalid prediction mode,
An image predictive coding method, wherein a prediction signal in an invalid prediction mode is replaced with a predetermined prediction vector. The image predictive encoding method according to any one of claims 1 to 4,
In the step of correcting the prediction signal in the invalid prediction mode,
Calculating a difference vector between a prediction signal vector of the invalid prediction mode and a prediction signal vector of a prediction mode in a region including the invalid prediction mode;
Calculating a correction vector from the calculated difference vector,
An image predictive coding method, wherein the prediction signal in the invalid prediction mode is corrected by the calculated correction vector. The image predictive encoding method according to claim 7,
In the step of calculating a correction vector from the calculated difference vector,
Calculate the correction factor from the overhead cost difference,
An image predictive coding method, comprising: calculating a correction vector from a difference vector and a correction coefficient. An image predictive decoding method for decoding an image encoded using the image predictive encoding method according to claim 1, comprising:
Generating a prediction signal for each prediction mode;
Calculating overhead cost from overhead code amount for each prediction mode;
Detecting an invalid prediction mode having a low selection probability according to a predetermined criterion from the calculated relationship between the overhead cost of each prediction mode and the prediction signal;
Correcting the detected prediction signal of the invalid prediction mode,
An image predictive decoding method comprising performing decoding after correcting a prediction signal in an invalid prediction mode. The image predictive decoding method according to claim 9,
In the step of detecting the invalid prediction mode,
Detecting the invalid prediction mode using the same method used to detect the invalid prediction mode when encoding the image,
In the step of correcting the prediction signal in the invalid prediction mode,
An image predictive decoding method, wherein a prediction signal in an invalid prediction mode is corrected using the same method as that used for correcting a prediction signal in an invalid prediction mode when an image is encoded. An image encoding apparatus for encoding an image by determining a prediction mode of an encoding target block from a plurality of prediction modes,
Prediction signal generating means for generating a prediction signal for each prediction mode;
Invalid prediction mode detection means for calculating an overhead cost from the overhead code amount for each prediction mode and detecting an invalid prediction mode having a low selection probability according to a predetermined criterion from the relationship between the calculated overhead cost of each prediction mode and the prediction signal When,
Prediction signal correction means for correcting the detected prediction signal in the invalid prediction mode;
A prediction mode determining means for calculating a coding cost for each prediction mode from a prediction residual power and an overhead code amount of each prediction mode, and determining a prediction mode that minimizes the coding cost as a prediction mode of an encoding target block; An image encoding device comprising: An image decoding device for decoding an image encoded using the image encoding device according to claim 11,
Prediction signal generating means for generating a prediction signal for each prediction mode;
An invalid prediction mode detecting means for calculating an overhead cost from an overhead code amount for each prediction mode, and detecting an invalid prediction mode having a low selection probability from a relationship between the calculated overhead cost of each prediction mode and a prediction signal;
Prediction signal correction means for correcting the detected prediction signal of the invalid prediction mode,
An image decoding apparatus, wherein a decoding process is performed after correcting a prediction signal in an invalid prediction mode.   An image encoding program for causing a computer to execute the image predictive encoding method according to any one of claims 1 to 8.   An image decoding program for causing a computer to execute the image predictive decoding method according to claim 9.   9. A computer-readable recording medium on which an image encoding program for causing a computer to execute the image predictive encoding method according to claim 1 is recorded.   The computer-readable recording medium which recorded the image decoding program for making a computer perform the image predictive decoding method of Claim 9 or Claim 10.

Images