JPH10304371A - Moving vector detecting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To operate switching to the shape of an adaptive and appropriate retrieval area through a simple arithmetic operation by using the efficiency of an inter-frame prediction calculated by selecting a moving vector with minimum prediction error. SOLUTION: In a retrieval area decision circuit 12, a retrieval area shape is determined from prediction efficiency in a coded frame calculated by a prediction efficiency calculating circuit 11, and each retrieval area is assigned to (n) pieces of moving vector detecting circuits 2-1,..., 2-n. In a prediction error judging circuit 13, a moving vector with minimum prediction error is selected from among the moving vectors detected by the plural moving vector detecting circuits 2-1,..., 2-n, and the moving vector is outputted to a movement compensation prediction circuit 4, and the information of a prediction error for the moving vector is outputted to a prediction efficiency calculating circuit 11 and an encoding mode decision circuit 3. Thus, an exact moving vector can be detected even in a small moving vector retrieval area, and a circuit scale can be reduced.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion vector detecting apparatus applied to a moving picture signal coding apparatus in recording, communication, transmission and broadcasting of an image. The present invention relates to a motion vector detecting device for motion compensation prediction encoding that encodes a difference between motion vector information indicating whether a coded block has moved from a region and a prediction block indicated by the vector information.

[0002]

2. Description of the Related Art Since a moving image has a very large data amount, a process of compressing the image on the transmitting or recording side and expanding it on the receiving or reproducing side is often performed. Such compression /
Decompression is also called encoding / decoding. In recent years, MPEG-1 (ISO / IEC11172) has been used as a moving image encoding method,
Inter-frame coding methods using motion-compensated inter-frame prediction such as MPEG-2 (ISO / IEC13818) are being used in the fields of storage, communication, and broadcasting. In these systems, each frame of the moving image sequence is divided into coding blocks, and a motion compensation frame for obtaining a prediction block using a motion vector detected from a past or future, or past and future frame for each coding block. Interim predictions have been made.

[0003] The coding mode of a coding block in MPEG includes an inter-frame prediction coding mode for obtaining a prediction block from a past or future frame or a past and future frame, and a coding mode for coding the coding block itself without using prediction. There is an intra-frame coding mode to be converted.

FIG. 14 shows a block diagram of a conventional motion vector detecting device. Here, 1 is a frame memory for temporarily storing an image, 2 is a motion vector detection circuit that calculates a motion vector for motion compensation inter-frame prediction, and 3 is a coding mode determined from information calculated by the motion vector detection circuit. 4 is a motion compensation prediction circuit that generates a prediction block, 5 is a selector, 6 is a subtractor that calculates a difference block between the coding block and the prediction block, and 7 is coding that encodes the difference block. , 8 a decoder for decoding the encoded data, 9 an adder for adding the decoded difference block and the prediction block to generate a decoded block, and 10 a frame memory for storing a decoded reference frame. .

The operation of FIG. 14 will be described below. The input image is temporarily written to the frame memory 1, and the pixel data of the encoded block read from the frame memory 1 is input to the motion vector detection circuit 2.

The motion vector detecting circuit 2 reads the data of the reference frame from the frame memory 10 and performs a matching error calculation with the encoded block to detect a motion vector. At this time, the motion vector detection circuit 2 calculates the prediction error in the motion compensation inter-frame prediction and the activity of the coded block, and outputs the calculated error to the coding mode determination circuit 3.

[0007] The coding mode determination circuit 3 uses the information such as the prediction error and the activity output from the motion vector detection circuit 2 to determine the coding mode of the coding block. Here, the prediction error is generally compared with the magnitude of the activity. If the prediction error is smaller, the inter-frame prediction coding is selected, and if the activity is smaller, the intra-frame coding is selected.

[0008] The motion compensation prediction circuit 4 generates a prediction block using the pixel data of the reference frame read from the frame memory 10 in accordance with the coding mode determined by the coding mode determination circuit 3.

[0009] The selector 5 switches the output according to the coding mode determined by the coding mode determination circuit 3. Here, “0” in the intra-frame encoding mode.
Is selected, otherwise the output (prediction block) of the motion compensation prediction circuit 4 is selected.

A difference block between the coded block and the predicted block is calculated by a subtractor 6. The difference block is encoded by the encoder 7, and encoded data is output.

Here, the encoded data is decoded by the decoder 8 and the decoded difference block is added to the prediction block by the adder 9 so as to be used as a reference frame for prediction of the subsequent frames. The output (decoded block) of the adder 9 is stored in the frame memory 10.

FIG. 15 shows an example of a motion vector in motion compensation prediction. As a method for detecting a motion vector, a block matching method is generally used. In the block matching method, prediction block candidates in the search area are
The amount of error between the prediction block candidate and the coded block is calculated. Then, the candidate with the smallest error amount is set as the prediction block, and the relative shift amount of the prediction block position from the coded block position is set as the motion vector.

By the way, in the block matching method, when an image having a large motion such that the motion vector exceeds the search range is input, an appropriate prediction block cannot be detected, so that the coding efficiency is reduced. .
However, simply expanding the search area increases the number of times of matching between the candidate of the prediction block and the coded block, thereby increasing the amount of calculation, increasing the processing time, or performing a parallel processing, so that the motion vector detection circuit is required. There was a drawback that the scale of the large size.

As one means for solving the drawback, Japanese Patent Application Laid-Open No. Hei 5-328333 discloses that a shape of a motion vector search area (search area shape) is made variable and a past motion vector at a position of a coding block is stored. There is disclosed a method for improving the motion vector detection performance by determining the shape of a search area of a target coded block according to the size / direction of the motion vector.

[0015]

Problems to be Solved by the Invention
In No. 3, since the current search area shape is determined based on the past motion vector at the coding block position, if the direction of the motion of the motion vector suddenly changes, it cannot cope with the change, and Vectors are erroneously detected and image quality deteriorates. Further, once the past motion vector is mistaken, the search area shape for the current coding block is not appropriately set, so that the motion vector in the current coding block is also erroneously detected. There is a possibility that a vicious cycle in which the search region shape for the transformed block is not properly set may be repeated, and in this case, an accurate motion vector cannot be detected forever. The present invention has been made in view of the above problems, and has as its object to provide a motion vector detecting device capable of adaptively switching to an appropriate search area shape by a simple calculation.

[0016]

In the motion vector detecting apparatus according to the present invention, each frame of a moving image sequence is divided into coding blocks, and each coding block is inter-frame predicted from a reference frame. In a motion vector detecting apparatus for video coding that encodes using a motion vector, a prediction efficiency calculating circuit that calculates prediction efficiency of inter-frame prediction, and a motion vector search range is determined for each of the plurality of motion vector detecting circuits. The search area determining circuit, one or more motion vector detecting circuits for detecting each of the motion vectors, and a plurality of the motion vector detecting circuits, the most out of the plurality of detected motion vectors The prediction error determination circuit that selects a motion vector that reduces the prediction error, and is calculated by the prediction efficiency calculation circuit. Utilizing the prediction efficiency of the prediction between frames, and also changes a search area shape of the reference frame adaptively.

In the motion vector detecting apparatus according to a second aspect of the present invention, when the prediction efficiency calculation circuit calculates the prediction efficiency, a prediction error or a prediction error in inter-frame prediction of a coded block, It is characterized by using the activity of the conversion block.

In the motion vector detecting device according to a third aspect of the present invention, when the prediction efficiency is calculated by the prediction efficiency calculation circuit, the prediction error is used, and the prediction error is more than a predetermined threshold value. It is characterized in that a small coded block is determined as a high-efficiency prediction block, and the number of high-efficiency prediction blocks in a frame is used.

In the motion vector detecting device according to a fourth aspect of the present invention, when the prediction efficiency calculation circuit calculates the prediction efficiency, the prediction error and the activity are used, and the prediction error is applied to the activity. A coded block whose prediction error is smaller than a value multiplied by a predetermined coefficient or whose prediction error is smaller than a predetermined threshold is determined as a high-efficiency prediction block, and the number of high-efficiency prediction blocks in a frame is defined as prediction efficiency. It is characterized by the following.

In the motion vector detecting apparatus according to a fifth aspect of the present invention, when the prediction efficiency is calculated by the prediction efficiency calculation circuit, the prediction error is used, and the reciprocal of the prediction error is used to calculate the cumulative value of the reciprocal of the prediction error in the entire frame. Is used as the prediction efficiency.

In the motion vector detecting apparatus according to a sixth aspect of the present invention, when the prediction efficiency is calculated by the prediction efficiency calculation circuit, the activity is represented by the prediction error using the prediction error and the activity. It is characterized in that a cumulative value in the frame of the divided value is set as the prediction efficiency.

In the motion vector detecting apparatus according to a seventh aspect of the present invention, as the prediction error used in the prediction efficiency calculation circuit, a sum of absolute differences between a coded block and a predicted block, or It is characterized in that a sum of squared differences between prediction blocks is used.

In the motion vector detecting device according to the eighth aspect of the present invention, the activity used in the prediction efficiency calculating circuit may be a sum of absolute values of AC components in a coding block or a square of the AC component in the coding block. It is characterized by using a sum.

In the motion vector detecting device according to the ninth aspect of the present invention, when the prediction efficiency calculated by the prediction efficiency calculation circuit falls below a predetermined threshold value after the end of the encoding, the search area determination circuit determines , Wherein the shape of the search area in subsequent encoding is changed.

In the motion vector detecting apparatus according to the tenth aspect of the present invention, the prediction efficiency is preliminarily calculated by the prediction efficiency calculation circuit before encoding, so that the calculated prediction efficiency is equal to a predetermined threshold value. If it is less than the above, the calculation of the prediction efficiency by changing the search area may be continued until the prediction efficiency exceeds a predetermined threshold value or until the number of changes in the search area shape reaches the predetermined number. When the calculated prediction efficiency is lower than a predetermined threshold, a motion vector with the highest prediction efficiency is detected by selecting a search area shape with the highest prediction efficiency.

In the motion vector detecting apparatus according to the eleventh aspect of the present invention, when the search area shape is determined by the prediction efficiency calculated by the prediction efficiency calculation circuit, the search area determination circuit is predetermined. A plurality of search area shapes are sequentially switched.

In the motion vector detecting device according to the twelfth aspect of the present invention, by providing a recording device capable of recording the search region shape used for encoding, the search region shape selected in the past can be used. Recording can be performed, and a search area shape with high prediction efficiency can be obtained in a shorter time based on the recording.

The motion vector detecting apparatus according to the thirteenth aspect of the present invention is characterized in that the motion vector detecting processing according to the first to twelfth aspects can be performed not only at the time of frame coding but also at the time of field coding.

[0029]

FIG. 1 is a block diagram showing an embodiment of a motion vector detecting device according to the present invention. Here, 11 is a prediction efficiency calculation circuit, 12 is a search area determination circuit, and 13 is a prediction error determination circuit. The blocks having the same functions as those in FIG. 14 are given the same numbers. However, unlike FIG. 14, FIG. 1 has n motion vector detecting circuits 2, and
... 2-n is attached.

In the prediction efficiency calculation circuit 11, the output from the prediction error determination circuit 13 and the coding mode determination circuit 3
Is used to calculate the prediction efficiency in the encoded frame.

The search area determination circuit 12 determines the shape of the search area from the prediction efficiency in the encoded frame calculated by the prediction efficiency calculation circuit 11 and sends the result to n motion vector detection circuits 2-1,. On the other hand, each search area is assigned. In the prediction error determination circuit 13, a motion vector with the smallest prediction error is selected from among the motion vectors detected from the plurality of motion vector detection circuits 2-1,..., 2-n, and the motion vector is subjected to motion compensation. The information is output to the prediction circuit 4, and information such as a prediction error corresponding to the motion vector is output to the prediction efficiency calculation circuit 11 and the coding mode determination circuit 3.

FIG. 2 shows an example of a search area shape when n = 2. The search area shape in FIG. 2 is constituted by two search areas. Here, search areas 1 and 2 are assigned to the two motion vector detection circuits, respectively.
By performing a matching operation in the search areas 1 and 2, the respective motion vectors 1 and 2 are detected. The prediction error determination circuit 13 selects one of the motion vectors 1 and 2 having a smaller prediction error, and outputs the selected motion vector as a motion vector for the encoded block.

FIG. 3 shows a block diagram of an embodiment of the prediction efficiency calculation circuit 11 according to the third aspect of the present invention. The comparator 14 compares the prediction error in the coding block with a predetermined threshold, and determines a coding block in which the prediction error is smaller than the predetermined threshold as a high-efficiency prediction block. The counter 15 counts the number of high-efficiency prediction blocks in the frame, and calculates the value of the counter as prediction efficiency.

FIG. 4 is a block diagram showing an embodiment of the prediction efficiency calculation circuit 11 according to the fourth aspect of the present invention. 1
6 is a constant multiplier for multiplying the activity by a constant α; 17 is a comparator for comparing the output of the constant multiplier 16 with the prediction error and the threshold to determine whether or not the block is a highly efficient prediction block; Is a counter for counting the number of highly efficient prediction blocks.

FIG. 5 shows an embodiment of the characteristics of the comparator 17. Generally, it can be said that the prediction error is large when the motion prediction efficiency is low. On the other hand, when a complicated pattern, that is, the activity is large, the prediction error becomes large. This means that even if the motion prediction successfully tracks the motion of the block, the prediction error will increase if the activity is always high,
The efficiency of motion prediction cannot be determined only by the magnitude of the prediction error. Since the determination of the search area should be determined with reference to the efficiency of the motion prediction, it is considered effective to cancel the influence of the activity. Equation (1) reduces the effect of the activity by dividing the prediction error by the activity. When the efficiency of the motion prediction is high, the equation (1) becomes small, and when the efficiency of the motion prediction is low, the equation (1) Is to be larger.

[0036]

(Equation 1)

FIG. 5 shows that equation (1) determines a coded block not exceeding the constant α as a block having a high prediction efficiency (hereinafter referred to as a high-efficiency prediction block). However, when the prediction error is sufficiently small, the coding efficiency does not decrease even if the equation (1) does not exceed α. Therefore, a predetermined threshold value is set and the prediction efficiency is high even when the prediction error is smaller than the predetermined threshold value. It is determined as an efficiency prediction block. Therefore, the activity and the prediction error of the encoded block are plotted in FIG. 6, and if the activity enters the shaded area, it is determined that the block is a highly efficient predicted block.

The counter 18 is reset to 0 for each frame, counts the number of high-efficiency prediction blocks, and uses the count value as the prediction efficiency in the encoded frame. The determination of the high-efficiency prediction block may be made for all the coded blocks in the frame, or may be made for only some of the coded blocks.

In the present embodiment, since the use of the activity reduces the prediction efficiency depending on the complexity of the picture of the coding block itself, a more accurate prediction efficiency is calculated as compared with FIG. It is possible to do.

FIG. 6 is a block diagram showing an embodiment of the prediction efficiency calculation circuit 11 according to the fifth aspect of the present invention. 19
Is a reciprocal calculator for calculating the reciprocal of the prediction error, 20 is an adder, 21 is a register for holding the accumulated value obtained by the adder 20, and 20 and 21 constitute a cumulative adder.

In the present embodiment, the cumulative value of the reciprocal of the prediction error in the coded block in the entire frame is defined as the prediction efficiency. Note that the cumulative addition of the reciprocal of the prediction error may be performed on all the coded blocks in the frame, or may be performed on only some of the coded blocks.

FIG. 7 is a block diagram of an embodiment of a prediction efficiency calculation circuit according to claim 6 of the present invention. Reference numeral 22 denotes a divider for dividing the activity by the prediction error to calculate the prediction efficiency for each coding block, reference numeral 23 denotes an adder, and reference numeral 24 denotes a register for holding an addition result.

The adder 23 and the register 24 constitute a cumulative adder. The register 24 is reset for each frame, and accumulates and adds one frame of prediction efficiency for each coded block output from the divider 22. The cumulative addition of the prediction efficiency for each block may be performed on all the coded blocks in the frame, or may be performed on only some of the coded blocks.

Here, the prediction efficiency for each coded block calculated by the divider 22 is expressed by equation (2).

[0045]

(Equation 2)

Here, "+1" in the denominator of equation (2)
Is added to prevent the denominator of equation (2) from becoming 0 when the prediction error is 0, and a value other than 1 may be added.

Also in this embodiment, it is possible to calculate the prediction efficiency more accurately than in FIG. 6 by using the activity as in FIG.

In the above embodiment, the prediction error and the activity are used for calculating the prediction efficiency. Since these parameters are used for determining the encoding mode, there is no need to newly calculate them. Therefore, there is almost no increase in the scale of hardware due to the calculation of the prediction efficiency.

In the motion vector detecting device according to the seventh aspect of the present invention, Expression (3) or Expression (4) is used as a prediction error in a coded block.

[0050]

(Equation 3)

[0051]

(Equation 4)

Where x is the pixel value of the coding block, r
Is the pixel value of the prediction block, B is the number of pixels in the block, i
Is the index for the pixel.

In the motion vector detecting device according to the eighth aspect of the present invention, Expression (5) or Expression (6) is used as an activity in a coded block.

[0054]

(Equation 5)

[0055]

(Equation 6)

[0056]

(Equation 7)

In the equation (7), d is the average value (DC component) of the pixel values in the coding block. Therefore, the activity in Expression (5) is the sum of the absolute values of the AC components in the coding block, and the activity in Expression (6) is the sum of the squares of the AC components in the coding block.

FIG. 8 is a flow chart showing an embodiment according to the ninth aspect of the present invention. In FIG. 1, after the coding of the coded frame is completed, the prediction efficiency in the coded frame is input from the prediction efficiency calculation circuit 11 to the search area determination circuit 12. The search area determination circuit 12 compares the prediction efficiency with a predetermined threshold value (8-1). If the prediction efficiency is equal to or higher than the predetermined threshold value, the subsequent coding is performed with the search area shape as it is. However, when the value falls below the threshold value, the shape of the search area of the motion vector is changed (8-
2) is performed, and a new search area is allocated to the motion vector detection circuit (8-3), and encoding of the next encoded frame is started.

FIG. 9 shows an embodiment of the search area shape determination processing according to claim 10 of the present invention. In the present embodiment, even when the prediction efficiencies in all the search area shapes are below the threshold value, it is possible to select the search area shape having the highest prediction efficiency.

The flow of the process in FIG. 9 will be described below.
Here, the index for the search area shape is p. First, as an initial setting, a variable i = 0, an optimum value MA
X is set to 0 (9-1). Before coding the encoded frame, the prediction efficiency calculation circuit 11 of FIG. 1 calculates the prediction efficiency in the search area shape p (9-2), and the calculated prediction efficiency is calculated by the search area determination circuit 12. , And a predetermined threshold value (9-3). When the calculated prediction efficiency is equal to or larger than the threshold, the search area shape p used for the prediction efficiency calculation is used as it is.

In the comparison between the calculated prediction efficiency and the predetermined threshold value (9-3), if the calculated prediction efficiency is smaller than the predetermined threshold value, the calculated prediction efficiency and the optimum prediction efficiency are compared. The value MAX is compared (9-4), and when the calculated prediction efficiency exceeds the optimum value MAX of the prediction efficiency, the calculated prediction efficiency is set as the optimum value MAX of the prediction efficiency (9-5). Is saved as the search area shape p _MAX with the highest prediction efficiency (9-
6). The number of prediction efficiency calculations is N at maximum, and the variable i is used to count the number of prediction efficiency calculations, and is incremented by one when the prediction efficiency falls below a predetermined threshold (9-9). If the number of calculations of the prediction efficiency has not reached N (9-7 _NO ), the search area shape p is changed (9-1).
0), the search area shape is assigned to the motion vector detection circuit (9-11), and the prediction efficiency is calculated again (9-11).
2) Compare with the threshold. When the prediction efficiency does not exceed a predetermined threshold value during the N calculations (9-7 _YES )
Uses the search area shape p _MAX having the highest prediction efficiency in the calculation up to that time as the search area shape p.

FIG. 10 shows an embodiment of the search area shape according to claim 11 of the present invention. The search area shape A is determined with respect to the position of each encoded block in the reference frame in order to improve the vertical motion vector detection performance.
The search area 1 and the search area 2 corresponding to the coding block are set in the vertical direction. The search area 1 and the search area 2 are overlapped in the row of the coding block position so that no gap is formed in the search range.

Similarly, in order to improve the performance of detecting a motion vector in the horizontal direction, the search area shape B is used for the search area 1 and the search area 2 corresponding to the coding block with respect to the position of each coding block in the reference frame. Are set in the horizontal direction. The search area 1 and the search area 2 are overlapped in the column of the coded block position so that no gap is formed in the search range.

FIG. 11 (a) shows an embodiment of the search area shape switching when the example of FIG. 10 corresponding to claim 11 of the present invention is applied to claim 9. Each frame to be encoded is a frame j-1 when a frame to be encoded is j.
The encoding is performed by the inter-frame prediction from. FIG.
In FIG. 1A, it is assumed that the screen moves largely in the horizontal direction in frames 2 to 5.

In the coding of frame 1, when the search area shape is set to A in FIG. 10, when the prediction efficiency calculated by prediction efficiency calculation circuit 11 in FIG. 1 exceeds a predetermined threshold value, It is not necessary to switch the search area shape, and encoding of frame 2 is performed with the search area shape shown in FIG.

However, in the encoding of frame 2, if the prediction efficiency calculated by the prediction efficiency calculation circuit 11 of FIG. 1 falls below a predetermined threshold value, it can be determined that the coding efficiency of the motion vector has decreased. The search area shape is switched to B in FIG. 10 by the search area determination circuit 12 in FIG.

In the coding up to frame 6, if the prediction efficiency exceeds a predetermined threshold value, it is not necessary to switch the search area shape, and the search area shape is changed to B in FIG.
Will remain.

However, when the prediction efficiency is less than the predetermined threshold value in the encoding of frame 6, it can be determined that the encoding efficiency of the motion vector has decreased, and the shape of the search area is switched to A in FIG.

As described above, by switching between the search area shape A and the search area shape B in FIG. 10 every time the prediction efficiency falls below a predetermined threshold value in the coded frame, the present invention provides a conventional example (Japanese Unexamined Patent Application Publication No. 5-328333
Unlike the control of the search area shape in (1), even if the search area shape is different from the actual image due to a rapidly moving image, the search area shape is switched in the next frame, so that the tracking ability is good and the motion vector detection accuracy is high. Will be higher.
Further, even when an error occurs in the past motion vector, the motion vector to be obtained is not affected, so that it is possible to prevent the image quality from being deteriorated even when the motion of the subject changes abruptly.

FIG. 11B shows an embodiment of the search area shape switching when the example of FIG. 10 corresponding to claim 11 of the present invention is applied to claim 10. In FIG. 11B, as in FIG. 11A, it is assumed that the screen moves greatly in the horizontal direction in frames 2 to 5. Claim 9
In the embodiment of the present invention, the prediction efficiency is calculated after the frame coding, whereas in the present embodiment, the prediction efficiency is calculated in advance before the frame coding, so that the prediction efficiency is calculated in the frames 2 and 6 in FIG. Is below a predetermined threshold before encoding. As a result, in FIG. 11A, the search area is switched between the frames 3 and 7, whereas in FIG. 11B of the present embodiment, the search areas are switched between the frames 2 and 6, that is, 1 from FIG. 11A. It is possible to switch the search area before the frame, and as a result, it is possible to perform a highly accurate motion vector detection that is more responsive to the screen motion than in the ninth aspect.

FIG. 12 shows an embodiment of a search area shape changing process according to claim 12 of the present invention. In the present embodiment, the use frequency of the search area shape is recorded, and when the search area shape is changed, the search area shape is selected in order from the search frequency having the highest use frequency, thereby predicting the search area shape in the order of high possibility of selection. Since the efficiency can be calculated, the optimal search area can be determined efficiently.

The flow of the process in FIG. 12 will be described below. In FIG. 12, the same processes as those in FIG. 9 are denoted by the same reference numerals. Also, let p be the index for the search area shape and hi be the frequency of occurrence of the search area shape p.
It is represented by st [p].

Assume that the search area shape p is currently selected. When the prediction efficiency in the search area shape p exceeds a predetermined threshold value (9-3 _NO ), the search area shape p
Is used. On the other hand, when the prediction efficiency in the search area shape p falls below the predetermined threshold value (9-3 _YES ), p ′
= 0, 1, 2, ... in order (12-5), the prediction efficiency in the search area shape p 'is calculated (12-3), and the search in which the prediction efficiency first exceeds a predetermined threshold value is performed. The region shape p 'is used (12-4 _NO ). However, as in FIG. 9, the number of calculations of the prediction efficiency is N at the maximum. If the prediction efficiency does not exceed a predetermined threshold value during the N calculations, the search having the highest prediction efficiency in the calculation up to that time is performed. The region shape p _MAX is adopted (9-8).

For the employed search area shape p, the occurrence frequency hist [p] is incremented (12-7).
The search area shapes are rearranged in descending order of hist [p],
The correspondence between p and the search area shape is updated (12-8).
FIG. 13 shows the processes of 12-7 and 12-8.

FIG. 13 shows an embodiment of the search area shape storage means for performing the above-mentioned processes 12-7 and 12-8. (A)
Denotes search area shape storage means, and the search area shape p (0,
1, 2,...), The occurrence frequency hist [p], and the actual search area shape (A, B, C,...) As one group.

(B) shows an example of the processing in 12-7. When the search area shape p = 1 is selected, hi corresponding to p = 1 is selected.
The value of st [p] is incremented to a value of 9.

(C) shows the processing of 12-8,
The shapes are sorted in descending order of the value of [p], and the correspondence between p and the search area shape is exchanged.

As described above, in the next frame encoding process, the prediction efficiency calculation process is performed from the shape B having the highest frequency.

In the above-mentioned claims 1 to 12, all the processes are described as frame encoding processes, but the same processes can be performed in the field encoding process.

[0080]

According to the present invention, since the shape of the motion vector search area is adaptively changed according to the coded image, an accurate motion vector can be detected even in a small motion vector search area. Can be reduced in circuit scale.

Further, in the present invention, Japanese Patent Application Laid-Open No. 5-328
Unlike the control of the search area shape in No. 333, the search area shape used in the motion vector detecting device can be appropriately set without being influenced by the error even if the past motion vector is incorrect, and accurate A simple motion vector can be detected.

Further, according to the ninth aspect of the present invention, unlike the control of the search area shape in JP-A-5-328333, the switching of the search area shape is adaptively controlled in accordance with the prediction efficiency, so that the object and the camera are controlled. It is possible to select an appropriate search area shape in real time, for example, when the movement is severe, and it is possible to detect an optimal motion vector.

In the third, fourth, fifth, sixth, seventh and eighth aspects of the present invention, since the prediction efficiency can be calculated using the parameters used for the coding mode determination, the circuit scale in the prediction efficiency calculation is reduced. be able to.

According to the tenth aspect of the present invention, a search area shape can be selected and switched before coding of a target frame, so that an appropriate and more appropriate action can be taken in a case where the movement of a subject or a camera is severe. A search area shape can be used, and an optimal motion vector can be detected.

According to the eleventh aspect of the present invention, the shape of the search area can be determined efficiently by making the determination in order from the more likely search area shape.

According to claim 12 of the present invention, claim 1
By storing the appearance frequency of the search area shape used in the 0 motion vector detection device, it becomes possible to obtain an appropriate motion vector search area shape in a short time.

According to the thirteenth aspect of the present invention, similar processing can be performed not only in frame coding but also in field coding.

[Brief description of the drawings]

FIG. 1 is a block diagram of one embodiment of the present invention.

FIG. 2 is an explanatory diagram of a motion vector search area shape according to the present invention.

FIG. 3 is a block diagram of an embodiment according to claim 2 of the present invention.

FIG. 4 is a block diagram of an embodiment according to claim 4 of the present invention.

FIG. 5 is an embodiment of the characteristics of the comparator in FIG. 6;

FIG. 6 is a block diagram of an embodiment according to claim 5 of the present invention.

FIG. 7 is a block diagram of an embodiment according to claim 6 of the present invention.

FIG. 8 is a flowchart of an embodiment according to claim 9 of the present invention.

FIG. 9 is a flowchart of an embodiment according to claim 10 of the present invention.

FIG. 10 is an example of a search area shape according to claim 11 of the present invention.

FIG. 11 is an explanatory diagram of one embodiment according to claim 11 of the present invention.

FIG. 12 is a flowchart of one embodiment according to claim 12 of the present invention.

FIG. 13 is an explanatory diagram of one embodiment according to claim 12 of the present invention.

FIG. 14 is a block diagram of a conventional video encoding device.

FIG. 15 is an explanatory diagram of motion vector detection.

[Explanation of symbols]

 Reference Signs List 1 frame memory 2-1 to 2-n motion vector detection circuit 3 coding mode determination circuit 4 motion compensation prediction circuit 5 selector 7 encoder 8 decoder 10 frame memory

Continued on the front page (72) Inventor Hiroshi Kusao 22-22 Nagaikecho, Abeno-ku, Osaka-shi, Osaka

Claims (13)

[Claims] 1. A motion vector detecting apparatus for video coding, wherein each coded frame of a video sequence is divided into coding blocks, and each coding block is coded using inter-frame prediction from a reference frame. A predetermined motion vector search area (hereinafter, referred to as a search area)
And one or more motion vector detecting circuits for detecting a motion vector by performing a predetermined operation between the pixel data of the prediction block and the pixel data of the coded block, and calculate the prediction efficiency in inter-frame prediction. A prediction efficiency calculation circuit; a search area determination circuit for determining a search area in one or more motion vector detection circuits; and a plurality of outputs of the plurality of motion vector detection circuits when a plurality of the motion vector detection circuits are provided. And a prediction error determination circuit that selects one from the following. The search area determination circuit uses a prediction efficiency calculated by the prediction efficiency calculation circuit to calculate a shape of a motion vector search area (hereinafter, a search area shape and a search area shape). Is determined, and the shape of the search area is decomposed into one or a plurality of search areas, and one or a plurality of motion vector detections are performed. The motion vector detection circuit detects a motion vector for each of the search areas allocated to the output circuit, and determines a motion vector having the smallest prediction error from among the plurality of detected motion vectors. A motion vector detecting device, wherein a motion vector is detected by selection by a circuit. 2. The prediction efficiency calculation circuit calculates a prediction efficiency using a prediction error in inter-frame prediction of a coding block, or a prediction error and an activity of the coding block. Motion vector detection device. 3. The prediction efficiency is used to calculate a prediction efficiency. A coding block in which the prediction error is smaller than a predetermined threshold value is determined as a high-efficiency prediction block, and the number of high-efficiency prediction blocks in a frame is determined. 3. The motion vector detection device according to claim 2, wherein the prediction efficiency is set as the prediction efficiency. 4. The method according to claim 1, wherein the prediction error and the activity are used to calculate a prediction efficiency, and the prediction error is smaller than a value obtained by multiplying the activity by a predetermined coefficient, or the prediction error is smaller than a predetermined threshold value. The motion vector detecting device according to claim 2, wherein a small coded block is determined as a high-efficiency prediction block, and the number of high-efficiency prediction blocks in the frame is set as the prediction efficiency. 5. The motion vector detecting device according to claim 2, wherein the prediction error is used for calculating the prediction efficiency, and a cumulative value of the reciprocal of the prediction error in the entire frame is used as the prediction efficiency. 6. A value obtained by cumulatively adding a value obtained by dividing the activity by the prediction error in each coding block in all coding blocks in a frame, using the prediction error and the activity in calculating prediction efficiency. 3. The motion vector detecting device according to claim 2, wherein is a prediction efficiency. 7. The method according to claim 2, wherein a sum of absolute differences between a coded block and a predicted block or a sum of squares of differences between a coded block and a predicted block is used as the prediction error. 4. The motion vector detecting device according to 4, 5, or 6. 8. The motion vector detecting apparatus according to claim 2, wherein the activity is a sum of absolute values of AC components in a coding block or a sum of squares of AC components in the coding block. . 9. When the prediction efficiency of the coded frame calculated by the prediction efficiency calculation circuit falls below a predetermined threshold after the coding of the coded frame is completed, the search area determination circuit outputs the code. 9. The motion vector detecting device according to claim 1, wherein the search area shape in encoding after the coded frame is changed. 10. A coding method according to claim 1, wherein said coding step further comprises: calculating a prediction efficiency in said prediction efficiency calculation circuit before encoding said coded frame; and when said prediction efficiency falls below a predetermined threshold value, said search area determination circuit includes: Change the shape and re-estimate the prediction efficiency, and continue this until the prediction efficiency exceeds a predetermined threshold or until the number of changes in the search region shape reaches a predetermined number, and the search region shape with the highest prediction efficiency The motion vector detecting device according to claim 1, 2, 3, 4, 5, 6, 7, or 8, wherein is a search area shape for the encoded frame. 11. The motion vector detecting device according to claim 9, wherein the search area determination circuit switches a plurality of search area shapes in a predetermined order. 12. A search area shape having a high prediction efficiency can be selected in a shorter time by recording a determination result in the search area determination circuit and sequentially selecting a search area shape having a higher selection frequency. 2. The method according to claim 1, wherein
The motion vector detection device according to 0 or 11. 13. The prediction efficiency calculation circuit according to claim 1, wherein a similar process is performed in a field encoding process instead of the frame encoding process.
The motion vector detection device according to 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12.