JP2001145114A - Device and method for detecting moving vector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a moving vector detecting device which actualizes the detection of an accurate moving vector actualizing high image quality, without increasing the circuit scale. SOLUTION: This moving vector detecting device 100 is equipped with a parameter setting means 30, which has parameters regarding a search range obtained by dividing at least part of the whole search range in a search frame by N which is a natural number larger than 2, a decording target block for target frame, a moving vector detecting means 20 which calculates evaluated value showing the correlativity between an object block to be encoded in an object frame and a candidate block for one of the N-divided search ranges and calculates a moving vector on the basis of the evaluated value, a result storage means 40, which stores the moving vector calculated by the moving vector detecting means 20 and the evaluated value, and a control means 10 which determines a moving vector for the object block to be encoded by making the moving vector detecting means 20 calculates a moving vector of a different search range and an evaluated value as many times as M (1<=M<=N), until a predetermined condition is satisfied, and receiving the moving vector and evaluated value from the result storage means 40.

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 device used in motion compensated predictive coding which is a technique for compressing and coding digital video data.

[0002]

2. Description of the Related Art As a method of realizing image coding of a moving image, information of a certain portion of the current image has been moved from the position of the immediately preceding image (movement vector), and the time of the data is calculated. There is a method of reducing the objective and quantitative redundancy. One method for extracting the motion vector is a block matching method.

FIG. 7 is a diagram showing the principle of the block matching method. The image of the target frame 702 to be encoded is compared with the image of the search frame 701 for searching for a motion vector, and the evaluation value most similar to the encoding target block 703 of the target frame 702 (that is, the evaluation value having the highest correlation) is compared. (Best match block 706) is extracted from within the search range 704 of the search frame 701,
A motion vector 705 is detected. This block matching method is widely used for motion compensation prediction in image compression coding.

FIG. 8 is an enlarged view of a search frame 701 shown in FIG.
Shown in In the block matching method, a plurality of candidate blocks 707 within the search range 704 of the search frame 701 (scan the candidate block 707 shown in FIG. And a candidate block having the highest correlation is selected as the best match block 706, and the best match block 706 is selected.
The difference between the position of the current position and the position of the current block 703 is detected as a motion vector 705.

[0005] As a conventional motion vector detecting device using the block matching method, there is a device for reducing the amount of hardware.

For example, in Japanese Patent Application Laid-Open No. Hei 6-141304, M × N arithmetic units each including a pixel storage register, a multiplexer and a difference absolute value calculator are arranged in an M × N matrix, and a reference block size is set. In the case of M × N pixels, a candidate vector having the highest degree of correlation for the number of candidate blocks M × N is selected. However, in this conventional example, since only the correlation degree of M × N candidate blocks can be calculated, it is difficult to expand the search area and increase the number of candidate blocks even when trying to select a candidate block having a higher correlation degree. It is.

For example, Japanese Patent Application Laid-Open No. 9-224249 discloses a motion vector detecting device 9 as shown in FIG.
00 is a first motion vector detection circuit 915 that performs motion vector detection with a wide detection range and coarse accuracy for a certain encoding target block, and a second motion vector detection circuit that performs motion vector detection with a narrow detection range and fine accuracy. And two types of motion vector detection circuits are disclosed. The detection device 900 is configured such that the detection range of the first motion vector detection circuit 915 having a wide detection range includes the detection range of the second motion vector detection circuit 916 having a narrow detection range. Thereby, the fast motion vector is detected by the first motion vector detection circuit 915 having a wide detection range, while the slow motion vector is detected by two different detection circuits (that is, the first motion vector having the wide detection range). And a second motion vector detection circuit 916) having a narrow detection range.
The correlation value detected by each detector is compared with a comparator 9
By performing the comparison using No. 17, the detection accuracy of a slow-moving motion vector is improved.

However, in this conventional example, it is necessary to provide two types of motion vector detection circuits 915 and 916, and the circuit scale becomes large and complicated.

[0009]

In the conventional motion vector detecting device, the search range is limited as described above. On the other hand, in order to widen the search range, a solution is provided by providing two types of motion vector detection circuits of fine precision and coarse precision.
00.

However, the motion vector detecting device 90
0, two types of motion vector detection circuits 915 and 916
Is required, which causes a problem that the circuit scale is increased. In particular, the motion vector detection using the block matching method detects the optimum place (motion vector) by moving the entire search range for the accumulation of the difference for each pixel in block units, thereby increasing the circuit scale. Problem.

It is inevitable that detection in a larger search range is desired in order to respond to the demand for high definition and high image quality corresponding to a large screen, and in this case, the circuit scale is further increased. .

SUMMARY OF THE INVENTION It is an object of the present invention to provide a motion vector detecting device which realizes accurate detection of a motion vector realizing high image quality without increasing the circuit scale.

[0013]

In order to solve this problem, the present invention focuses on performing a search by dividing a search range into a plurality of regions when a wide search range is assumed, and using the same coding target block. , A plurality of motion vector detections are performed for each search range. That is, by providing a means for setting a plurality of parameters in a search range and a means for storing a plurality of results in one motion vector detecting means, an accurate motion vector corresponding to high image quality can be obtained without increasing the circuit scale. A detectable motion vector detecting device is provided.

The motion vector detecting apparatus according to the present invention comprises: a parameter setting means having a parameter relating to a search range obtained by dividing at least a part of the entire search range in a search frame into N natural numbers of 2 or more; Calculating an evaluation value indicating a degree of correlation between a block to be converted and a candidate block in one of the N search ranges, and calculating a motion vector based on the evaluation value; Means for storing the motion vector and the evaluation value calculated by the motion vector detecting means; and a result storing means for storing the motion vectors and the motion vectors in different search ranges until the motion vector detecting means satisfies a predetermined condition. An operation is performed to calculate the evaluation value M times (1 ≦ M ≦ N), and the motion vector and the evaluation value are received from the result storage means. And a control means for determining a motion vector for the block. With this configuration, accurate motion vector detection in a wide search range can be realized with a small circuit scale.

[0015] The control means may input a parameter relating to the N divided search range to the parameter input means.

The M may be equal to the N.

After storing the N motion vectors and the evaluation values, the result storage means may output the evaluation values and the motion vectors to the control means.

[0018] The parameter relating to the search range may include a search position indicating a position of the search range.

[0019] The parameter relating to the search range may include a search size indicating the size of the search range.

[0020] The parameter relating to the search range may include a subsample indicating a size and a direction for thinning out the search accuracy.

[0021] The parameter relating to the search range may include a priority indicating an arbitrary value to be added or multiplied to the evaluation value.

The priority of the search range may be higher in the search range shifted in the horizontal direction with respect to the center of the entire search range than in the search range shifted in the vertical direction.

[0023] The parameter related to the search range may include a search order indicating an order in which the motion vector and the evaluation value are calculated.

The search order of the search range may be higher in a search range shifted in a horizontal direction with respect to a center of the entire search range than in a search range shifted in a vertical direction.

The parameters relating to the search range may include at least two of a search position, a search size, a subsample, a priority, and a search order.

[0026] The predetermined condition may be that the evaluation value is smaller than a predetermined value of the control means.

[0027] The predetermined condition may be limited by a predetermined time.

The N may be 2.

The N may be 3.

At least a part of the entire search range may be divided only in the vertical direction.

At least a part of the entire search range may be divided only in the horizontal direction.

At least a part of the entire search range may be divided such that the search range has an elliptical shape.

[0033] The elliptical shape may be longer in the horizontal direction than in the vertical direction.

[0034] At least a part of the entire search range may be divided so that the search range has a rectangular shape.

3. The motion vector detecting device according to claim 1, wherein the N search ranges are divided such that the candidate blocks of the N search ranges do not completely overlap each other.

[0036] At least two of said N search areas may have different sizes.

At least two of the N search areas may have different sizes, and the large search area may be formed so as to surround the small search area.

[0038] The size of the search range including the center of the entire search range may be the smallest among the sizes of the N search ranges.

The size of the search range near the outer periphery of the entire search range may be the largest among the sizes of the N search ranges.

[0040] At least a part of the entire search range may be rectangular, and the size of the search range may be such that the size of the search range at the corner of the rectangle is the widest.

At least two of said N search areas may have different priorities.

[0042] Among the N search ranges, the search range including the center of the entire search range may have the highest priority.

[0043] Among the N search ranges, the search range near the outer periphery of the entire search range may have the lowest priority.

[0044] At least a part of the entire search range may be rectangular, and the priority of the search range of the corner of the rectangle may be the lowest among the priorities of the N search ranges.

[0045] At least two of said N search regions may have different subsamples.

The search range located vertically shifted with respect to the center of the entire search range may be sub-sampled in the vertical direction.

[0047] A search range located horizontally shifted with respect to the center of the entire search range may be subsampled in the horizontal direction.

[0048] Of the N search range sub-samples, the search range sub-sample near the outer periphery of the entire search range may be the largest.

[0049] At least a part of the entire search range may be rectangular, and among the N sub-samples of the search range, the sub-sample of the search range at the corner of the rectangle may be the largest.

In the motion vector detecting method according to the present invention, an input step includes inputting parameters relating to a search range obtained by dividing at least a part of the entire search range in a search frame into N natural numbers equal to or more than two, to a motion vector detecting means. An evaluation value indicating a correlation value between a coding target block of a target frame and a candidate block in the search range based on parameters of a certain search range among the N divided search ranges. Calculating a motion vector based on the evaluation value; and calculating whether the control unit satisfies a predetermined condition based on the calculated evaluation value and the motion vector. And a step.

When the predetermined condition is satisfied, the method may further include a step of ending the calculation of the motion vector.

When the predetermined condition is not satisfied, the method may further include a step of repeating the inputting step and the calculating step.

[0053]

Embodiments of the present invention will be described below with reference to the drawings.

(Embodiment 1) FIG. 1A shows a configuration of a motion vector detecting apparatus 100 according to Embodiment 1 of the present invention.

The motion vector detecting device 100 includes a control means 10, a motion vector detecting means 20, a parameter setting means 30, and a result storing means 40. Control means 1
0 inputs the setting parameter of the search range to the parameter setting means 30 and activates the motion vector detection means 20. The parameter setting means 30 sends the entire search range (which may be a part of the search frame or the search frame itself) from the control means 10 to N (N
Is a setting parameter (search position in this embodiment) relating to a certain search range divided into two or more natural numbers. The motion vector detecting means 20 is started by the control means 10, and receives a target frame, a search frame, and a parameter relating to one search range from the parameter setting means 30 from outside. Then, the motion vector detection means 20 calculates a motion vector and an evaluation value between the target frame and the search frame based on the input parameters in one search range, and stores the motion vector and the evaluation value in the result storage means. Output to 40. The result storage means 40 stores the motion vector calculated by the motion vector detection means 20 and the evaluation value. The control unit 10 receives the stored motion vector and the evaluation value from the result storage unit 40, and receives an end signal from the motion vector detection unit 20. The evaluation value indicates the degree of correlation between the encoding target block of the target frame and the candidate blocks within the search range of the entire search range.

In the above description, for the purpose of explanation, the control means 1
Although 0 is input to the parameter setting means 30 regarding the parameters relating to the N divided search ranges, the parameter setting means 30 may store the parameters relating to the N search ranges in advance. Further, the result storage unit 40 may output the motion vector to the control unit 10 after the N motion vectors are stored.

The operation of the motion vector detecting device 100 will be described with reference to the operation flowchart of FIG. 2A. Step 1: The control means 10 sets N parameters (search positions) relating to the search range in the parameter setting means 30. Step 2: The control means 10 is the motion vector detection means 20
Start. Step 3: The motion vector detection means 20 calculates a motion vector and an evaluation value for one search position and outputs them to the result storage means 40, and ends the motion vector detection processing corresponding to one search position. Step 4: Steps 2 to 3 are repeated N times. Step 5: The control means 10 fetches N sets of motion vectors and evaluation values from the result storage means 40. Step 6: The control means 10 determines an optimal motion vector for the current block based on the N sets of motion vectors and the evaluation values.

As a specific example, a search method in the case where the entire search range 300 is divided into four search ranges (search 1, search 2, search 3, and search 4) (N = 4) will be described with reference to FIG. 3A. I do. In FIG. 3A, the upper left, upper right,
The lower left and lower right regions are divided as search 1, search 2, search 3, and search 4, respectively. In FIG. 3A, the search ranges are formed so as not to overlap each other. This does not mean that the candidate block at the boundary of the search range is not searched. The candidate block at the boundary of the search range is also searched as a candidate block of any of the search ranges located at the boundary. That is, the candidate blocks included in different search ranges are formed so as not to completely overlap.

In FIG. 3A, the search range of the entire search range 300 is divided so as to have a rectangular shape.

For example, when motion vector detection processing is performed for all search ranges (ie, calculation and storage of motion vectors and evaluation values are performed four times), the parameter setting means 30 searches for four search ranges. Set the position,
The result storage means 40 is configured to be able to store four sets of motion vectors and evaluation values. The control means 10 writes each search position, for example, the upper left coordinate of the search position, into the parameter setting means 30.

3A is set to the upper left of the entire search range 300. “Search 2” is set to the upper right of the entire search range 300. After setting the four coordinates of the search position in this way, the control unit 10 sets the motion vector detection unit 20
To start.

When the first activation is performed, the motion vector detecting means 20 calculates a motion vector using “search 1” in FIG. 3A as a search range, and evaluates the motion vector as a first storage result in the result storage means 40. This is stored together with the value, and an end signal is returned to the control means 10 to end the first motion vector detection process.

Then, a motion vector is calculated in the same manner for "search 2", and the motion vector and the evaluation value are stored in the result storage means 40 as the second storage result. Hereinafter, the same applies to “search 3” and “search 4”.

More specifically, the correlation (that is, the evaluation value) between the current block 301 and the candidate block 302 for search 1 is obtained, and then the candidate block 302 is scanned (candidate blocks are searched vertically or horizontally in search 1). After shifting, the evaluation value is obtained in the same manner. By repeating such an operation, the best matching block in search 1 that shows the best correlation with the coding target block 301 is obtained, and as a result, the motion vector and evaluation value in search 1 are calculated. By repeating the same operation in searches 2 to 4, a motion vector and an evaluation value are calculated for each search range.

When the execution of the motion vector detecting means 20 corresponding to the four search ranges has been completed four times, the result storing means 4
The control means 10 takes in each calculation result stored in 0, and
Based on each evaluation value, one of the four motion vectors is selected as the motion vector of the current block to be encoded.
That is, an appropriate motion vector is determined for the current block. Generally, since the evaluation value is calculated by accumulating (or squaring) the difference between the corresponding pixels, the one with the smallest evaluation value is selected.

As described above, in the present embodiment, the entire search range is divided into four, and one motion vector detecting means
By obtaining motion vectors in one search range, it is possible to detect a motion vector in a wide search range with a small circuit scale, and thus it is possible to select an accurate motion vector.

In this embodiment, the number of divisions (N) of the entire search range is set to 4, but N may be any natural number of 2 or more.

Further, in the embodiment of the present invention, the entire search range is divided into N (N = 4 in the above description) and the search range is searched without omission. However, the partial search is not performed by the division method. A range may exist. In this case, only a part of the entire search range needs to be searched, so that a short search can be performed.

For example, as shown in the operation flowchart of FIG. 2B, after Steps 1 to 3 of the flowchart described with reference to FIG. 2A, Step 4 ′: the calculated motion vector and the evaluation value for the parameters in a certain search range are obtained. Output from the storage means 40 to the control means 10. Step 5 ': The control means 10 determines whether or not the evaluation value for a parameter in a certain search range satisfies a predetermined condition. If the predetermined condition is not satisfied, steps 2 to 5 are repeated. Step 6 ': The control means 10 determines an optimal motion vector for the encoding target block based on the calculated motion vector and the evaluation value.

As a specific example of the predetermined condition, for example, when the evaluation value is smaller than a predetermined value, a motion vector corresponding to the evaluation value is determined as a motion vector.
In this case, it is not necessary for the motion vector detection means 20 to search all N search ranges, and it is sufficient to obtain M (1 ≦ M ≦ N) motion vectors until a predetermined condition is satisfied. If the evaluation value is larger than the predetermined value, the control unit 10 instructs the motion vector detection unit 20 to obtain a motion vector and an evaluation value for another search range parameter.

As another example of the predetermined condition, an optimum motion vector may be obtained from M motion vectors calculated within a predetermined time limited by a timer or the like.

In the present embodiment, the control means counts the number of repetitions. However, the repetition number counting means is provided in the motion vector detecting means, and an end signal is returned to the control means 10 after the end of N times. There is no problem in operation. In this case, after the result storage means 40 determines the optimal motion vector based on the N evaluation values,
An optimal motion vector may be output to the control means 10.

In this embodiment, N sets of parameters are set at first, but one set of parameter setting means and one set of result storage means are provided.
It is also feasible to perform execution and fetch output results N times.

FIG. 1B shows a specific configuration. The motion vector detection block 50 includes a motion vector detection unit 20, a parameter setting unit 30, and a result storage unit 40, as indicated by a broken line in FIG. 1A. Control means 1
0 indicates that each of the different motion vector detection blocks 50
Control is performed to obtain the motion vector and the evaluation value for the different search ranges as described above. Thus, a motion vector for a certain encoding target block can be quickly obtained in the entire search range.

Further, using the configuration shown in FIG. 1B, the control means 10 may detect a motion vector corresponding to a different encoding target block by using a plurality of motion vector detection means. By simultaneously obtaining the motion vectors of the different encoding target blocks of the target frame, it is possible to quickly detect a motion vector between the entire target frame and the entire search frame.

Although the parameter setting means 30 and the result storage means 40 are provided in the motion vector detection block 50 in FIG. 1B, the control means is configured to include the parameter setting means and the result storage means. Is also good.

Further, the entire search range 300 shown in FIG.
Is completely divided into N (4) search ranges, but the invention is not limited to this. As shown in FIG. 3B, the entire search range 350 need not be completely divided into N (4) search ranges, and the search range is not limited to being in contact with another search range (for example, FIG. 3B). Search 1
And search 4). That is, it is sufficient that at least a part of the entire search range 350 is divided into N pieces. In this case, if a position where a motion vector is likely to be detected is set as a search position in the search range, an appropriate motion vector can be obtained in a short time.

Further, in FIG. 3A, the coding target block 301 of the target frame corresponds to the position of the center of the entire search range 300. When determining an appropriate motion vector after calculating the motion vectors of all search ranges with the above configuration,
The motion vector is determined on average within the entire search range. However, the position of the center of the entire search range and the block to be coded need not correspond. For example, when a large upper left motion vector is estimated, an appropriate motion vector can be obtained in a short time by forming Search 1, Search 2, and Search 3 as shown in FIG. 3B.

(Embodiment 2) FIG. 4 shows a motion vector detecting device 400 according to Embodiment 2 of the present invention.

In FIG. 4, a motion vector detecting device 40
0 is the control means 10, the motion vector detection means 20,
It comprises a parameter setting means 130 and a result storage means 140. The parameter setting means 130 inputs the parameters related to the search range (in this embodiment, the search position, the search size, the subsample, the priority, and the search order) to the motion vector detection means 20, and the result storage means 140 The motion vector and the evaluation value calculated by the means 20 are stored. Otherwise, the configuration is the same as that of the first embodiment. The parameters related to the search range include at least one of a search position, a search size, a subsample, a priority, and a search order;
Alternatively, at least two may be included.

Regarding the operation, the search size, subsample, priority, and search order are set as parameters in step 1 of the first embodiment, in addition to the search position. Otherwise, this is the same as that described in the first embodiment with reference to the flowchart in FIG. 2A.

More specifically, a description will be given of a search method when the entire search range 500 is divided into nine (N = 9) as shown in FIG. 5A.

To calculate the motion vector nine times, nine parameters are set in the parameter setting means 130,
The result storage means 140 is configured to be able to store nine sets of motion vectors and evaluation values. Then, the control unit 10 writes each search position, search size, subsample, priority, and search order in the parameter setting unit 130.

In FIG. 5A, the entire search range 500 is divided into nine search ranges having different sizes (search A, search B, search C, search D, search E, search F, search G, search H, and search I). Is done. As the parameters related to the search range, a search position, a search size, a subsample, a priority, and a search order are set for each search range. The priority means an arbitrary value to be multiplied or added to the evaluation value calculated based on the search range in order to determine a motion vector. The search order means the order of searching the search range in the entire search range. Generally, the search order is set in order from a search range in which an appropriate motion vector is likely to be obtained.

In the present specification, “different parameters of the search range (search position, search size, subsample, priority or search order)” means that the parameters of the N search ranges are all different. Does not mean that there are at least two search ranges having different parameters among the N search ranges.

In the case of "search A" in FIG. 5A, the search position is set to "center" of the entire search range 500, the search size is set to "small", and the subsample is set to "none". In the entire search range 500, the search size increases as the search position moves away from the center, and the setting is such that the sub-sampling tends to be performed.

The reason for “sub-sampling when the search position is away from the center” utilizes the human visual characteristic that the resolution for a fast object is low.

That is, when the motion vector is large, it can be inferred that an object having a high moving speed is present. Therefore, at the search position far from the center, the search pixels are subsampled (thinned out) to detect the motion vector with coarse accuracy. But sufficient performance can be demonstrated. Therefore, as the search position is farther from the center, more subsamples may be performed.

The reason that the search size is increased when the search position is away from the center is that the amount of calculation is reduced in a region far from the center by performing the sub-sampling as described above. This is because it is possible to increase the size and search over a wider range. Therefore, it is desirable to increase the search size as the search position moves away from the center.

As an example of the setting parameters, FIG. 5A shows a case where the entire search range 500 is 112 pixels (± 48 pixels) in the horizontal direction and 80 pixels (± 32 pixels) in the vertical direction. The search position here is represented by upper left coordinates of each search range.

Further, regarding the sub-samples, search B,
Regarding C, since the horizontal movement is large, it is set to perform horizontal sub-sampling. On the other hand, for the searches D and E, the vertical motion is large, so that the vertical sub-sample is set. Searches F, G, H, and I were set to perform sub-sampling in both the horizontal and vertical directions.

The setting of the above parameters can be changed for each encoding target block. Specifically, when the motion vector of the previous encoding target block indicates a large value, the size of the search range far from the center of the search position is increased to further coarsen the subsamples, It is also possible to set a position far away.

After setting the nine setting parameters in this way, the control means 10 activates the motion vector detection means 20. When the first activation is performed, the motion vector detection means 20 calculates a motion vector using “search A” in FIG. 5A as a search range, and compares the motion vector calculated as the first result in the result storage means 140 with an evaluation value. And the end signal is returned to the control means 10 to end the first motion vector detection process.

Next, a motion vector is similarly calculated for “search B”, and the motion vector and the evaluation value are stored in the result storage means 140 as a second storage result. Hereinafter, the same applies to “search C” to “search I”.

When the execution of the motion vector detecting means 20 corresponding to the nine search ranges is completed nine times, the result storing means 1
The control means 10 takes in each calculation result stored in 40 and selects one from nine motion vectors as a motion vector of the current block to be coded based on each evaluation value. Generally, since the evaluation value is calculated by accumulating (or squaring) the difference between the corresponding pixels, the one with the smallest evaluation value is selected.

At this time, as described with reference to FIG. 2B in the first embodiment, if the predetermined condition is satisfied, the motion vector need not be calculated N times, but M times (1 ≦ M ≦
N) is sufficient. As described above, the predetermined condition includes, for example, a case where the evaluation value is lower than a certain predetermined value, or a case where the search time is limited to a predetermined time by a timer or the like.

However, as shown in FIG. 5A, it is possible to easily make a selection with a priority depending on a search position with this configuration. The reason for giving priority to the search position is that if the motion vector is large, the code amount of the motion vector itself becomes large, which may adversely affect the compression ratio.

As an implementation method, numbers from 1 (higher priority) to 4 (lower priority) are described in the priority item of FIG. 5A. There is a method of selecting one as the motion vector of the current block to be encoded based on the correction value obtained by adding the above number to the value.

Further, the entire search range 500 shown in FIG.
In the example, the priorities of the searches B and C in the horizontal direction of the search A at the center are set higher than the priorities of the searches D and E in the vertical direction. In general, since the image display device is longer in the horizontal direction than in the vertical direction, there is a high possibility that the motion vector has a large horizontal component. Therefore, it is preferable that the priority of the search range in the horizontal direction is higher than the priority of the search range in the vertical direction.

Further, in order to shorten the time for calculating the motion vector, it is not necessary to obtain the motion vectors of the search ranges (search F, search G, search H, and search I in FIG. 5A) of the corners of the entire search range. May be.

Similarly, according to the above configuration, it is possible to select a search position with a search order. For example, as described with reference to FIG. 2B, when the calculation of the motion vector is ended when the evaluation value of the search range is smaller than a predetermined value, the search range in which an appropriate motion vector is likely to be calculated is determined. By searching in order, the search time can be reduced. Also, even if the operation of the motion vector detection device 400 stops for some reason during the search, if the search is performed in the search order, there is a high possibility that an appropriate motion vector is obtained. In FIG. 5A, the priority and the search order are described by the same numeral, but it goes without saying that the present invention is not limited to this.

As described above, in the present embodiment, a small number of circuits are obtained by obtaining a motion vector by setting a search position, a search size, a subsample, a priority, and a search order that are suitable for a plurality of search ranges. With various settings despite the scale, it is possible to select an accurate motion vector in a wide search range.

Further, by adding or multiplying an arbitrary value to a plurality of evaluation values, it is possible to provide a selection priority based on a search position.

In the present embodiment, the number of divisions (N) of the search range is set to 9, but N may be a natural number of 2 or more.

For example, FIG. 5B is a diagram in which the number of divisions of the entire search range 520 is set to 3, and the entire search range 520 is horizontally divided into a search J, a search K, and a search L. In this case, it is appropriate that the width x1 of the search J is smaller than the width x2 of the search K and the width x3 of the search L. That is, the search size of the search range including the search center of the entire search range is smaller than the search size of the other search ranges, and it is suitable to perform detection with fine accuracy in consideration of the viewing angle characteristics and the circuit scale. This is because the resolution of a human visual characteristic is high with respect to a still image or an object that moves slowly (that is, an object having a small motion vector).

Similarly, FIG. 5C is a diagram in which the number of divisions of the entire search range 540 is set to 3, and the entire search range 540 is vertically divided into a search M, a search N, and a search O. In this case, for the same reason, the width x4 of the search M is equal to the width x5 of the search N and the width of the search O.
It is suitable that the width is smaller than the width x6.

Further, in the present embodiment, the search range at the center is densely performed. However, the result of the surrounding coding target block, the information of the motion vector which is temporally previous, and the external information such as the motion of the camera are used. Alternatively, the part to be searched densely may be a part other than the center.

Further, in the present embodiment, the center search range is set to be the smallest, but the size for dense search may be changed according to image characteristic information that can be estimated in advance.

Further, in this embodiment, one search range is set for performing a dense search. However, when it is possible to preliminarily estimate that there are a plurality of locations having a high existence probability, a plurality of densely searched portions are set. You may.

In the above description, the entire search range is divided in units of rectangles, but the present invention is not limited to this. Ideally, it is desirable to divide the search range with an elliptical shape. FIG. 5D shows an example in which the search range 560 is divided into elliptical shapes.

Considering the horizontal to vertical ratio of the screen, NTS
Standards such as C are 4: 3, and high-definition types are 16: 9. Therefore, the movement of the object in the video and the movement of the entire video tend to be longer in the horizontal direction than in the vertical direction. Therefore, ideally, FIG.
It is desirable that the search range be an elliptical search as shown in FIG. At that time, it is suitable to detect a narrow range of the search center with fine precision, and to perform detection with coarser accuracy as the distance from the search center increases, in consideration of the viewing angle characteristics and the circuit scale.

However, since an elliptical search range is difficult to manufacture, it is practical to provide a rectangular search range that is close to an ellipse. Specifically, for example, when the entire search range is divided into nine as described above, four corners are not searched,
Alternatively, reducing the search size of the four corners is considered as a practical means.

Also, instead of the elliptical shape shown in FIG. 5D, as shown in FIG. 5E, a search P having a small search size and a search Q having a large search size occupy the entire search range 580. Is formed so as to surround the. When searching for the search P that is close to the search center, the search size is reduced, the search is performed with finer search accuracy without performing sub-sampling, and the search Q that is farther from the search center is searched. In this case, the search size may be widened, sub-samples may be performed, and the search accuracy may be reduced for detection.

(Embodiment 3) FIG. 6 is a configuration diagram showing Embodiment 3 of the present invention. 6, the motion vector detecting device 1 of the present invention and a general-purpose computer 600 are connected via a main bus, and the general-purpose computer 600 includes a CPU, a main storage, a large-capacity storage medium, a monitor, a terminal, and the like.

The operation of the present embodiment will be described. The moving image data including the search frame and the target frame is stored in the large-capacity storage medium, and the necessary search frame and the target frame are read out from the large-capacity storage medium via the main bus according to the instruction of the motion vector detecting device 1. Do that.

In this case, since the general-purpose computer 600 uses the main bus for exchanging data with each device, the frequency of use is uncertain. On the other hand, the transfer amount of the search frame and the target frame requested by the motion vector detecting device 1 is almost constant and requires a transfer amount of about several giga bps.

Therefore, depending on the use status of the general-purpose computer 600, there are cases where the required search range can be transferred within the allowable time and cases where the required search range cannot be transferred. There is no problem when the transfer is completed in time. However, when the transfer is not completed in time, the motion vector detection processing is delayed, and all the N sets of parameters set in the first and second embodiments cannot be executed.

Even in such a case, by dividing the entire search range into a plurality of parts and sequentially setting the parameters in the parameter setting means so as to start from a region having a high possibility that a motion vector exists, even if the processing time is insufficient, Even if the search operation is stopped halfway, it is possible to detect an effective motion vector having a good evaluation value.

In the above-described embodiment, the execution order is set in the parameter setting means. However, a dedicated execution order storage unit for storing a predetermined execution order may be provided, and the execution may be performed according to the order.

[0120]

As described above, according to the present invention, after dividing the entire search range into an arbitrary number of search positions, and calculating a motion vector at the search position a plurality of times, a motion vector can be selected. It is possible to select an accurate motion vector in a wide search range while having a large scale.

Also, according to the present invention, since a motion vector can be selected after setting the arbitrary search position, search size, subsample, priority, and search order a plurality of times, the characteristics of the image can be reduced even with a small circuit scale. Various settings in accordance with this enable selection of an accurate motion vector in a wide search range.

Further, according to the present invention, it is possible to provide a priority of selection based on a search range by adding or multiplying an arbitrary value to a plurality of evaluation values.

Further, since the present invention has a configuration in which a search can be performed from a high search order range, motion vector detection cannot be performed for all search positions set in the parameter setting means, and even if the operation is stopped halfway, Increase the possibility that a valid motion vector is detected.

[Brief description of the drawings]

FIG. 1A is a schematic diagram of a motion vector detection device according to the present invention.

FIG. 1B is a schematic diagram using a plurality of motion vector detection blocks.

FIG. 2A is a flowchart showing the operation of the present invention.

FIG. 2B is a flowchart illustrating another operation of the present invention.

FIG. 3A is a diagram showing division of the entire search range according to the present invention.

FIG. 3B is a diagram showing another division of the entire search range according to the present invention.

FIG. 4 is a schematic diagram of another motion vector detecting device of the present invention.

FIG. 5A is a diagram showing another division of the entire search range according to the present invention.

FIG. 5B is a diagram showing a horizontal division of the entire search range according to the present invention.

FIG. 5C is a diagram showing the vertical division of the entire search range according to the present invention.

FIG. 5D is a diagram showing the division of an elliptical shape of the entire search range according to the present invention.

FIG. 5E is a diagram showing division of a shape in which a small size search range is surrounded by a large size search range according to the present invention.

FIG. 6 is a schematic diagram showing a motion vector detection circuit and a general-purpose computer according to the present invention.

FIG. 7 is a diagram illustrating block matching;

FIG. 8 is a diagram illustrating an enlarged search frame.

FIG. 9 is a diagram illustrating a conventional motion vector detection circuit.

[Explanation of symbols]

Reference Signs List 1 motion vector detecting device 10 control means 20 motion vector detecting means 30, 130 parameter setting means 40, 140 result storing means 50 motion vector detecting block 100, 200, 400 motion detecting device 300, 500, 520, 540, 560, 580 Search range 301,703 Encoding target block 302,707 Candidate block 303 Best match block 600 General-purpose computer 701 Search frame 702 Target frame 704 Search range 705 Motion vector 706 Best match block 900 Motion detection device

Claims (39)

[Claims] 1. A parameter setting means having a parameter relating to a search range obtained by dividing at least a part of an entire search range in a search frame into N natural numbers equal to or more than 2, a coding target block of a target frame, and the N Motion vector detecting means for calculating an evaluation value indicating a degree of correlation with a candidate block in one of the search ranges and calculating a motion vector based on the evaluation value; A result storage unit that stores the motion vector and the evaluation value calculated by the unit; and a motion vector and the evaluation value in different search ranges until the motion vector detection unit satisfies a predetermined condition. M ≦ N) times, receiving the motion vector and the evaluation value from the result storage means,
A control unit that determines a motion vector for the current block. 2. The motion vector detection device according to claim 1, wherein the control unit inputs a parameter relating to the N divided search ranges to the parameter input unit. 3. The motion vector detecting device according to claim 1, wherein said M is equal to said N. 4. The motion vector detecting device according to claim 3, wherein the result storage unit outputs the evaluation value and the motion vector to the control unit after storing the N motion vectors and the evaluation values. 5. The motion vector detection device according to claim 1, wherein the parameter related to the search range includes a search position indicating a position of the search range. 6. The motion vector detection device according to claim 1, wherein the parameter related to the search range includes a search size indicating a size of the search range. 7. The motion vector detection device according to claim 1, wherein the parameter related to the search range includes a subsample indicating a magnitude and a direction for thinning out search accuracy. 8. The motion vector detection device according to claim 1, wherein the parameter relating to the search range includes a priority indicating an arbitrary value to be added or multiplied to the evaluation value. 9. The search range according to claim 8, wherein the priority of the search range shifted in the horizontal direction with respect to the center of the entire search range is higher than the priority of the search range shifted in the vertical direction. The motion vector detecting device as described in the above. 10. A parameter related to the search range,
The motion vector detection device according to claim 1, further comprising a search order indicating an order in which the motion vector and the evaluation value are calculated. 11. The search order of the search range, wherein the search order of the search range shifted horizontally with respect to the center of the entire search range is higher than the search order of the search range shifted vertically. The motion vector detecting device as described in the above. 12. A parameter relating to the search range,
The motion vector detection device according to claim 1, further comprising at least two of a search position, a search size, a subsample, a priority, and a search order. 13. The motion vector detecting device according to claim 1, wherein the predetermined condition is that the evaluation value is smaller than a predetermined value of the control means. 14. The motion vector detecting device according to claim 1, wherein the predetermined condition is limited by a predetermined time. 15. The motion vector detecting device according to claim 1, wherein said N is 2. 16. The motion vector detecting device according to claim 1, wherein said N is three. 17. The motion vector detection device according to claim 1, wherein at least a part of the entire search range is divided only in a vertical direction. 18. The motion vector detection device according to claim 1, wherein at least a part of the entire search range is divided only in a horizontal direction. 19. At least a part of the entire search range includes:
The motion vector detecting device according to claim 1, wherein the search range is divided so as to have an elliptical shape. 20. The motion vector detecting device according to claim 19, wherein the elliptical shape is longer in a horizontal direction than in a vertical direction. 21. At least a part of the entire search range is:
The motion vector detection device according to claim 1, wherein the search range is divided so as to have a rectangular shape. 22. The motion vector detecting apparatus according to claim 1, wherein the N search ranges are divided such that the candidate blocks of the N search ranges do not completely overlap each other. 23. The motion vector detecting device according to claim 6, wherein at least two of the N search areas have different sizes. 24. The motion of claim 6, wherein at least two of the N search regions have different sizes, and wherein a large size search range is formed to surround a small size search range. Vector detection device. 25. In the size of the N search ranges,
24. The motion vector detection device according to claim 23, wherein the size of the search range including the center of the entire search range is the smallest. 26. In the size of the N search ranges,
24. The motion vector detection device according to claim 23, wherein the size of the search range near the outer periphery of the entire search range is the widest. 27. The motion vector detection device according to claim 23, wherein at least a part of the entire search range is a rectangle, and the size of the search range is the largest in the size of the search range of the corner of the rectangle. 28. The motion vector detection device according to claim 8, wherein at least two of the N search areas have different priorities. 29. Among the priorities of the N search ranges,
29. The motion vector detection device according to claim 28, wherein the search range including the center of the entire search range has the highest priority. 30. Among the priorities of the N search ranges,
29. The motion vector detection device according to claim 28, wherein the search range near the outer periphery of the entire search range has the lowest priority. 31. The search method according to claim 28, wherein at least a part of the entire search range is a rectangle, and among the N search ranges, a search range of a corner of the rectangle has the lowest priority. Motion vector detection device. 32. The motion vector detection device according to claim 7, wherein at least two of the N search areas have different subsamples. 33. The motion vector detecting device according to claim 32, wherein a search range vertically shifted from a center of the entire search range is sub-sampled in a vertical direction. 34. The motion vector detection device according to claim 32, wherein a search range positioned horizontally shifted from the center of the entire search range is sub-sampled in the horizontal direction. 35. The motion vector detection device according to claim 32, wherein among the N sub-samples in the search range, the sub-sample in the search range near the outer periphery of the entire search range is the largest. 36. At least a part of the entire search range is rectangular, and among the sub-samples of the N search ranges,
33. The motion vector detection device according to claim 32, wherein a subsample in a search range of a corner of the rectangle is the largest. 37. An input step of inputting parameters relating to a search range obtained by dividing at least a part of the entire search range in the search frame into N natural numbers equal to or more than two to the motion vector detecting means; The motion vector detecting means calculates a correlation between an encoding target block of a target frame and a candidate block within the search range based on a parameter of a certain search range of the search range. A calculating step of calculating a motion vector, and a determining step of determining whether the control means satisfies a predetermined condition based on the calculated evaluation value and the motion vector, and a determining step. Motion vector detection method. 38. The motion vector detection method according to claim 37, further comprising ending the calculation of the motion vector when the predetermined condition is satisfied. 39. The motion vector detecting method according to claim 37, further comprising a step of repeating the inputting step and the calculating step when the predetermined condition is not satisfied.