JP2001298745A - Device and method for retrieving motion vector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce an arithmetic operation amount and a processing time required for motion vector retrieval while maintaining the retrieval accuracy of a motion vector. SOLUTION: A detection object image division section 1 divides a coding object frame into clusters, a prediction reference image division section 12 generating a reduction object cluster extracts the reduced object cluster from a reference prediction frame resulting from reducing an area at the same position of the reduction object cluster. An error calculation section 3 obtains a sum SADCL of absolute values of differences of pixels overlapped between the reduction object cluster and the area of the reference prediction frame at the same position as that of the reduction object cluster. When the sum SADCL is smaller than processing threshold value, since it depicts that the motion of the cluster is small, it is discriminated that no retrieval is required for the motion vector and the retrieval processing for the motion vector is skipped. The motion vector is calculated from other clusters.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion vector search apparatus and method, and more particularly, to a motion vector search apparatus and method based on a macroblock macro clustering method.

[0002]

2. Description of the Related Art In recent years, ISDN (Integrated Services)
Digital Network) or GSTN (General Switched)
Video communication services such as video telephony and video conferencing are promising as services that make effective use of the Telephone Network, and research on high-efficiency coding for efficient transmission of such moving images has been actively conducted. ing. In this study, we generally use the statistical properties of images to
The amount of information is reduced by removing the redundancy included in the image. As such an encoding method, a hybrid encoding method combining motion compensation prediction and discrete cosine transform is well known.

For example, an ITU (International Teleco
mmunications Union) -T Recommendation H. At 263,
The following motion compensated prediction is used. A current frame to be coded (hereinafter referred to as a “frame to be coded”) is divided into coding processing units called macroblocks, and frames that are temporally earlier or later than the frame to be coded (hereinafter “prediction reference frame”). ) Is detected from within the predetermined search area of the macroblock of the encoding target block.

[0004] The amount of information is reduced by encoding and transmitting a positional change of a macroblock between a current frame and a predicted reference frame as a motion vector.

[0005] As the most general motion vector search method, there is known a full search method for evaluating the correlation between all pixels included in a search area and a macroblock. Hereinafter, the full search method will be described with reference to the drawings.

Referring to FIG. 8, a frame to be encoded 10
0 is divided into macroblocks of a predetermined size (for example, 16 × 16 pixels). One macro block 101 is selected as a motion vector detection target block. Referring to FIG. 9, search area 103 is set for a motion vector search in prediction reference frame 102 one frame before encoding target frame 100. The size of the search area 103 is larger than that of the macroblock 101. For example, a range of 16 pixels each in the negative direction of the X axis and the Y axis of the macroblock 101 and a range of 15 pixels each in the positive direction around the macroblock 101 , And 47 × 47 pixels.

In search area 103, macro block 10
While shifting 1 by one pixel in the horizontal or vertical direction, the sum of absolute values of pixel value differences between overlapping pixels is obtained. The partial area 104 having the minimum total value is a part having the highest correlation with the macroblock 101. Therefore, a vector 105 from the macroblock 101 to the partial area 104 is obtained, and the vector 105 is set as a motion vector of the macroblock 101.

In the motion vector search processing by the full search method, differences in pixel values are obtained for all pixels in the macro block 101 at all positions in the search area 103. Therefore, there is a problem that the amount of calculation is enormous. Several techniques have been proposed so far to reduce the amount of computation, but when the amount of computation is reduced, the accuracy of motion vector search is reduced. Thus, it can be said that the calculation amount and the search accuracy have a trade-off relationship. In addition, there is a problem that if the search accuracy of the motion vector is low, the compression efficiency deteriorates.

In such a background, as a method of obtaining a highly accurate motion vector with a reduced amount of calculation, "Gen
Fujita et al., "A VLSI Architecture for Motion Est
imation Core Dedicated to H.263 Video Coding ", IEI
CE TRANS. ELECTRON., VOL.E81-C, NO.5, pp.702-707, May
1998 "has been proposed.

Referring to FIG. 10, a motion vector search apparatus for searching for a motion vector using a macro clustering method receives a frame to be coded, divides the frame to be coded, and extracts a region to be detected of the motion vector. A detection target image dividing unit 1 that receives a prediction reference frame, sets a search area in the prediction reference frame, and cuts out an area of the same size as the detection target area from the search area; An error calculating unit 3 connected to the target image dividing unit 1 and the prediction reference image dividing unit 2 for calculating an error between the cut-out areas (for example, a sum of absolute values of differences between pixels); An error storage unit 5 that stores a motion vector of the detection target area, and is connected to the error calculation unit 3 and the error storage unit 5, evaluates the error, and calculates a minimum error value and an error value. There and an error evaluation unit 4 for determining the position in the search area at which the minimum.

The motion vector detection processing will be described.
Referring to FIG. 11, detection target image dividing section 1 converts encoding target frame 200 to a predetermined size (for example, 16 × 16
The pixel 201 is divided into macroblocks (pixels), and a cluster 201 is formed by m × n macroblocks of the macroblocks. Here, the case where m = n = 2 will be described. That is, the size of the cluster is 32 × 32 pixels. Referring to FIG. 4, detection target image division unit 1 thins out the pixels included in cluster 201 such that the ratio in the X direction and the Y direction is m: n. Here, by setting the upper left pixel in the 2 × 2 pixel area as a pixel in that area, 32
An area of 16 × 16 pixels (hereinafter referred to as “cluster to be detected”) is created by thinning out the cluster 201 of × 32 pixels.

The predicted reference image dividing unit 2 creates a reduced predicted reference frame 204 by thinning out the predicted reference frame by the same thinning method as the detection target image dividing unit 1. The prediction reference image dividing unit 2 sets a search area 205 in the reduced prediction reference frame 204. The search area 205 is larger than the reduction target cluster 203. For example, in the X direction and the Y direction of the reduction target cluster 203, 2 ⁱ −1 (i =
1, 2,. . . ) Has a size obtained by adding pixels. Here, the case where i = 3 will be described. Therefore, the size of the search area 205 is 30 × 30 pixels.

The error evaluator 4 uses the above-described full search method to shift the cluster 203 to be reduced in the search area 205 in the horizontal or vertical direction by a predetermined pixel (generally, one pixel), and adjusts the distance between overlapping pixels. , The absolute value of the difference between the pixel values is sequentially obtained, and the sum of the absolute values of the differences in the reduction target cluster 203 is calculated.

Referring to FIG. 6, error evaluating section 4 sequentially compares the calculated sum of absolute values of the differences with the minimum value of the sum of absolute values of the differences stored in error storing section 5, The partial area cluster 206 that minimizes the sum of the absolute values of the differences is detected. The error evaluator 4 stores the minimum value of the sum of the absolute values of the differences in the error storage 5. The error evaluator 4 obtains a motion vector 207 from the center of the reduction target cluster 203 to the center of the partial area cluster 206, and stores the motion vector 207 in the error storage 5.

Referring to FIG. 12, predictive reference image dividing section 2
Is a vector 209 obtained by multiplying the motion vector 207 by m in the X direction and by n in the Y direction as an initial vector of the macroblock 202 included in the cluster 201. With reference to FIG. 7, the prediction reference image dividing unit 2
, A search area 211 centered on a partial area macroblock 210 obtained by moving the macroblock 202 according to the vector 209 is set.

The search area 211 has (2 ^j -1) pixels (j = j) in the positive and negative directions in the X and Y directions of the partial area macroblock 210, respectively.
1, 2,. . . ) Has the added size.

By finding the position corresponding to the macroblock 202 in the search area 211 using the full search method,
Finally, the motion vector of the macro block 202 can be obtained.

As described above, a rough motion vector search is performed in the cluster hierarchy, and then a detailed motion vector search is performed in the block hierarchy.
Perform a stage search. Therefore, compared to the case where a motion vector is implemented only in the block hierarchy, a search region of the same size can be searched with high accuracy and with a small amount of calculation.

[0019]

In the macroblock macro clustering method, a motion vector search is performed also on a macroblock which does not originally need to obtain a motion vector, that is, a macroblock whose change is small compared with frames preceding and succeeding in time. ing. For this reason, there is a problem that the amount of calculation required for the search processing increases and processing time is wasted.

There is also a problem that the power consumption of the motion vector search device increases with an increase in the amount of calculation and the processing time.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to reduce the amount of calculation and processing time required for a motion vector search while maintaining the accuracy of motion vector search. It is an object of the present invention to provide a motion vector search device that can perform the motion vector search.

Another object of the present invention is to provide a motion vector search apparatus with low power consumption.

[0023]

A motion vector search apparatus according to an aspect of the present invention uses a motion compensated prediction to set a position corresponding to a block included in a first frame in a predetermined relationship with the first frame. A motion vector of the block is searched for from the second frame. The motion vector search device is connected to a cluster generation unit for generating a cluster including one or more blocks in a first frame including a plurality of blocks, and is connected to the cluster and the second frame. A cluster correlation judging means for judging the correlation between the cluster and the area at the same position in the
Cluster motion vector detecting means for searching the second frame for a position corresponding to the cluster only for a cluster determined to have no correlation with the corresponding region in the frame of ,
The second frame is connected to the cluster correlation determining means and the cluster motion vector detecting means, and based on the output of the cluster correlation determining means and the motion vector of the cluster, a position corresponding to a block included in the cluster in the second frame is obtained. Motion vector detecting means for detecting a motion vector.

When it is determined that there is a correlation between the clusters at the same position in the first and second frames, it can be determined that the cluster has no motion. For this reason, for a cluster determined to have no motion, it is possible to omit the subsequent process of searching for a motion vector of the cluster.
Therefore, it is possible to reduce the amount of calculation and the processing time required for the motion vector search without lowering the accuracy of the motion vector. Accordingly, the power consumption of the motion vector search device can be reduced.

Preferably, the motion vector detecting means is connected to the cluster correlation determining means and the cluster motion vector detecting means, and based on the output of the cluster correlation determining means and the motion vector of the cluster, the motion of the block included in the cluster is determined. Initial vector setting means for setting an initial vector representing an initial value of the vector, and a correlation between the block and a region in the second frame after the block is moved according to the initial vector, And a block correlation judging unit for judging the position of the second frame, the position corresponding to the block is determined only for the block judged to have no correlation with the corresponding area in the second frame. A block for searching from the second frame and detecting a motion vector of the block. And a click motion vector detecting means.

If it is determined that the blocks have correlation at the same position in the first and second frames, it can be determined that the blocks have no motion. Therefore, for a block determined to have no motion, it is possible to omit the process of searching for a motion vector of the subsequent block.
Therefore, it is possible to reduce the amount of calculation and the processing time required for the motion vector search without lowering the accuracy of the motion vector. Accordingly, the power consumption of the motion vector search device can be reduced.

More preferably, the block motion vector detecting means is connected to the cluster correlation judging means and the block correlation judging means, and detects a block determined to have no correlation with the corresponding area in the second frame. Only the search area setting means for setting the search area in the second frame based on the output of the cluster correlation judgment means for the cluster including the block, and the search area setting means connected to the search area setting means, and Means for searching for a position corresponding to the block by using the method, and detecting a motion vector of the block.

[0028] More preferably, the search area setting means includes:
The search area when it is determined that there is no correlation between the cluster and the corresponding area of the cluster is set to be larger than the search area when it is determined that there is correlation.

If it is determined that the clusters have a correlation but the blocks constituting the cluster have no correlation, a global motion vector has not been obtained in the cluster hierarchy. Therefore, by setting the search area to be large, the motion vector can be obtained with high accuracy.

[0030] More preferably, the cluster correlation judging means comprises a cluster and a first correlation judging means for judging a correlation between the cluster and an area located at the same position as the cluster in the second frame. For each of the blocks, the block is connected to second correlation determining means for determining the correlation between the block and an area at the same position as the block in the second frame, and to the first and second correlation determining means. And means for judging the correlation between the cluster and the corresponding area in the second frame based on the output of the first correlation judging means and the output of the second correlation judging means.

When determining whether or not there is a correlation between clusters, it is also determined whether or not the blocks constituting the cluster have a correlation. Therefore, the reliability of the correlation between the clusters is improved.

In a motion vector search method according to another aspect of the present invention, a position corresponding to a block included in a first frame is determined from a second frame having a predetermined relationship with the first frame by using motion compensated prediction. Then, the motion vector of the block is searched. In the motion vector search method, in a first frame including a plurality of blocks, a cluster including one or more blocks is generated, and a correlation between the cluster and an area at the same position as the cluster in the second frame is determined. Is determined, and only for clusters determined to have no correlation with the corresponding area in the second frame, a position corresponding to the cluster is searched from the second frame to detect a motion vector of the cluster. For a block included in the cluster determined to have no correlation between the step and the corresponding region in the second frame, a corresponding position in the second frame is obtained,
Detecting a motion vector of the block.

When it is determined that there is a correlation between the clusters at the same position in the first and second frames, it can be determined that the cluster has no motion. For this reason, for a cluster determined to have no motion, it is possible to omit the subsequent process of searching for a motion vector of the cluster.
Therefore, it is possible to reduce the amount of calculation and the processing time required for the motion vector search without lowering the accuracy of the motion vector.

Preferably, the step of detecting the motion vector of the block includes the step of initializing the motion vector of the block included in the cluster based on the correlation between the cluster and the corresponding region of the cluster in the second frame and the motion vector of the cluster. Setting an initial vector representing a value; determining a correlation between the block and a region in the second frame after moving the block according to the initial vector; And searching only the position corresponding to the block from the second frame and extracting the motion vector of the block only for the block determined to have no correlation.

If it is determined that the blocks have correlation at the same position in the first and second frames, it can be determined that the blocks have no motion. Therefore, for a block determined to have no motion, it is possible to omit the process of searching for a motion vector of the subsequent block.
Therefore, it is possible to reduce the amount of calculation and the processing time required for the motion vector search without lowering the accuracy of the motion vector.

[0036] More preferably, the step of extracting a motion vector of a block is performed only for a block including a block that is determined to have no correlation with the corresponding region in the second frame. Setting a search area in the second frame based on the correlation between the search area and the corresponding area in the second frame; searching for a position corresponding to the block in the set search area; Detecting the vector.

More preferably, the step of setting the search area in the second frame includes the step of determining that the search area when there is no correlation between the cluster and the corresponding area of the cluster has the correlation. Is set to be larger than the search area in the case where the search is performed.

When it is determined that the clusters have correlation but the blocks that constitute the cluster have no correlation, a global motion vector has not been obtained in the cluster hierarchy. Therefore, by setting the search area to be large, the motion vector can be obtained with high accuracy.

More preferably, the step of judging the correlation between the cluster and the area located at the same position includes the step of:
Determining the correlation between the cluster and the region at the same position in the frame, and determining, for each of the blocks constituting the cluster, the correlation between the block and the region at the same position as the block in the second frame. Based on the correlation with the cluster and the correlation with the blocks making up the cluster,
And determining the correlation with the corresponding area in the frame of (a).

When determining whether or not there is a correlation between clusters, it is also determined whether or not the blocks constituting the cluster have a correlation. Therefore, the reliability of the correlation between the clusters is improved.

[0041]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Referring to FIG. 1, a motion vector search device according to an embodiment of the present invention receives a frame to be coded and divides the frame into a detection target region for motion vector search. 1 and a prediction reference frame that is one frame earlier than the encoding target frame in time, set a motion vector search area, and cut out an area of the same size as the detection target area from the search area. Prediction reference image division unit 12 and detection target image division unit 1
And an error calculation unit 3 connected to the prediction reference image division unit 12 and calculating an error between the cut-out regions (for example, a sum of absolute values of differences between pixels); a minimum value of the error and a motion of the detection target region. An error storage unit 5 for storing a vector,
An error evaluation unit 4 is connected to the error calculation unit 3 and the error storage unit 5 and evaluates the error, and obtains a minimum value of the error and a position in the search area when the error is maximum.

The motion vector search device is further connected to the error calculation unit 3 and, when searching in the cluster hierarchy, determines whether or not to skip the search. Are connected to a macroblock layer skip determination unit 8, a cluster layer skip determination unit 6, and a macroblock layer skip determination unit 8 that determine whether to skip the search when searching in the macroblock layer. A search area determining section 7 for determining a search area for a motion vector based on the outputs of the cluster layer skip determining section 6 and the macroblock layer skip determining section 8, and an initial setting for storing an initial setting value for defining the search area Value register 9 and initial setting value register 9
Set value register 10 for storing an enlarged set value for setting a search area larger than the initial set value stored in
Is connected to the search area determination unit 7, the initial setting value register 9, and the enlargement setting value register 10, and according to the output of the search area determination unit 7, changes the value of one of the initial setting value register 9 and the enlargement setting value register 10. A selector 11 for selecting and supplying the selected reference image to the prediction reference image dividing unit 12.

The motion vector search processing will be described with reference to FIGS. The detection target image dividing unit 1
The encoding target frame is divided into clusters, and a reduction target cluster 203 shown in FIG. 4 is created (S1). Since the method of creating reduction target cluster 203 is the same as the conventional method, detailed description thereof will not be repeated here.

The prediction reference image dividing unit 12 extracts a region at the same position as the cluster 203 to be reduced from the reduced reference prediction frame. The error calculation unit 3 calculates the reduction target cluster 2
The sum of the absolute values of the differences between the pixels overlapping between the target pixel 03 and the region of the reference prediction frame at the same position as the reduction target cluster 203 (hereinafter referred to as “error evaluation value”) SAD _CL is obtained (S2).

Referring to FIG. 5, reduction target cluster 203
Is composed of four reduced macroblocks, and the error information of the macroblocks is represented by SAD _CL1 and SA, respectively.
_Let D _CL2 , SAD _CL3 and SAD _CL4 . At this time,
The relationship shown in the following equation (1) is established.

SAD _CL = SAD _CL1 + SAD _CL2 + SAD _CL3 + SAD _CL4 (1) The cluster hierarchy skip determination unit 6 checks whether or not all the determination conditions of the following equations (2) to (6) are satisfied (S
3).

SAD _CL <ThSAD _CL (2) SAD _CL1 <ThSAD _CLq (3) SAD _CL2 <ThSAD _CLq (4) SAD _CL3 <ThSAD _CLq (5) SAD _CL4 <ThSAD _CLq (6) , ThSAD _CL and ThSAD _CLq are ThSA
Satisfy D _CL> ThSAD _CLq the relationship is a predetermined threshold value.

When all of the determination conditions of the equations (2) to (6) are satisfied, it is determined that it is not necessary to search for a motion vector for that cluster because the cluster has little motion (YES in S3). ), And proceeds to S9.

Referring to FIG. 6, if at least one of the judgment conditions of equations (2) to (6) does not satisfy the condition, a motion vector is calculated for that cluster (N in S3).
O), the prediction reference image division unit 12 sets a search area of the reduction target cluster 203 in the reduced prediction reference frame 204 (S4). The search area 205 is an area obtained by adding eight pixels to each of the upper, lower, left, and right sides of the reduction target cluster 203. The value of 8 pixels is stored in the initial setting value register 9 in advance.

The predictive reference image dividing unit 12 searches the search area 20
5, the search point is updated (S5), and a partial area cluster 206 is set in the search area 205. The error calculation unit 3
An error evaluation value between the reduction target cluster 203 and the partial area cluster 206 is calculated (S6). At the same time, the error evaluator 4 compares the calculated error evaluation value with the previously calculated minimum value of the error evaluation value in the search area 205 stored in the error storage 5 and stores the error storage value. The unit 5 stores the minimum value of the error evaluation value and the position of the partial area cluster 206 where the error evaluation value is the minimum.

It is checked whether or not an error evaluation value has been calculated for all search points in the search area 205 (S7). If an unprocessed search point exists (YE in S7)
S), the processing after S5 is repeated.

If error evaluation values have been obtained for all search points in the search area 205 (NO in S7), FIG.
, The error evaluation unit 4 determines that the cluster to be reduced 203
Region cluster 2 with the smallest error evaluation value from the center of
A motion vector 207 heading to the center of 06 is obtained and stored in the error storage unit 5 (S8).

It is determined whether the processing has been completed for all the reduction target clusters 203 to be detected (S9). If there is an unprocessed reduction target cluster 203 (YES in S9), after updating the reduction target cluster 203 (S10), the processing from S2 onward is repeated.

If the processing has been completed for all the clusters 203 to be reduced (NO in S9), an initial vector is set for each of a plurality of macroblocks included in the cluster corresponding to the cluster 203 to be reduced. (S
11). If it is determined in step S3 that the cluster has little motion, (0, 0) is given as the initial vector of the macroblock. In other cases, the motion vector 207 for the cluster 203 to be reduced obtained in the processing of S8, which is m times in the X direction and n times in the Y direction, is set as the initial vector of the macroblock. Here, m and n are thinning intervals used when creating the reduction target cluster 203.

Referring to FIG. 7, error calculating section 3 includes macro block 202 cut out from the encoding target frame,
An error evaluation value SAD _MB between the macroblock 202 and the partial area macroblock 210 existing at the position where the macroblock 202 is moved by the initial vector in the prediction reference frame 208 is calculated (S12).

The macroblock hierarchy skip determination unit 8
Upon receiving the error evaluation value SAD _MB , it is checked whether or not the following equation (7) holds (S13).

SAD _MB <ThSAD _MB (7) Here, ThSAD _MB is a predetermined threshold value.

When the expression (7) is satisfied, (YE in S13)
S), the initial vector set in S11 is set as the motion vector of the macroblock (S14), and the process proceeds to S20.

If the equation (7) is not satisfied (NO in S13), a search area 211 of the macroblock 202 is set in the prediction reference frame 208 to calculate a motion vector for the macroblock (S15). The size of the search area differs depending on whether or not the motion vector search in the cluster hierarchy has been skipped. That is, when the search for the motion vector is skipped in the cluster hierarchy, it is considered that the search for the motion vector is not sufficient. Therefore, a larger search area is set as compared with the case where a search for a motion vector is performed in the cluster hierarchy. The search area determination unit 7 controls the output of the selector 11 as described below.

(1) When Search for a Motion Vector is Skipped in the Cluster Layer A global motion vector is obtained in the cluster layer. Therefore, in the macroblock hierarchy, a search area having a normal size may be set to search for a motion vector. Therefore, the search range determining unit 7 controls the selector 11 so that the value held in the initial setting value register 9 is output from the selector 11.

(2) When the search for a motion vector is not skipped in the cluster layer A global motion vector is not found in the cluster layer. Therefore, it is necessary to set a search area larger than a search area of a normal size and search for a motion vector. Therefore, the search range determination unit 7 controls the selector 11 so that the value held in the enlargement setting value register 10 is output from the selector 11.

As an example, when the search for the motion vector is skipped in the cluster hierarchy, a region obtained by adding six pixels to each of the upper, lower, left, and right sides of the partial region macroblock 210 is set as the search region 211. If the search for the motion vector is not skipped in the cluster hierarchy, an area obtained by adding four pixels to the upper, lower, left, and right sides of the partial area macroblock 210 is set as the search area 211.

After the search area 211 is set, the search points are updated in the search area 211, and a partial area macro block 210 is set from the search area 211. The error calculator 3 calculates an error evaluation value between the macroblock 202 and the partial area macroblock 210 (S17). At the same time, the error evaluation unit 4 calculates the calculated error evaluation value,
The minimum value of the error evaluation value within the search area 211 calculated so far stored in the error storage unit 5 is compared, and the minimum value of the error evaluation value and the minimum value of the error evaluation value are stored in the error storage unit 5. And the position of the partial area macroblock 210 as follows.

It is checked whether or not the error evaluation value has been calculated for all the search points in the search area 211 (S18). If an unprocessed search point exists (Y in S18)
ES), and the processes in and after S16 are repeated.

If error evaluation values have been obtained for all search points in search area 211 (YES in S18),
The error evaluator 4 obtains a motion vector from the center of the macroblock 202 toward the center of the partial area macroblock 210 having the minimum error evaluation value based on the data stored in the error storage 5, and stores the motion vector in the error storage 5. Remember (S1
9).

After the processing of S14 or S19, it is determined whether or not the motion vector search processing has been completed for all the macroblocks 202 to be detected (S20). If there is an unprocessed macroblock 202 (YES in S20), after updating the macroblock 202 (S21), the processes in and after S11 are repeated.

If the processing has been completed for all the macroblocks 202 (NO in S20), a series of processing ends.

As described above, according to the present embodiment, when searching for a motion vector in the cluster layer and the macroblock layer, the motion vector search processing is performed only when the error evaluation value at the initial position is large. did. As a result, without reducing the accuracy of the motion vector,
It is possible to reduce the amount of calculation and processing time required for a motion vector search. Accordingly, the power consumption of the motion vector search device can be reduced.

The embodiments disclosed this time are to be considered in all respects as illustrative and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

[0070]

According to the present invention, the amount of calculation and the processing time required for searching for a motion vector can be reduced without lowering the accuracy of the motion vector. Accordingly, the power consumption of the motion vector search device can be reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a motion vector search device according to an embodiment of the present invention.

FIG. 2 is a flowchart of a motion vector search process.

FIG. 3 is a flowchart of a motion vector search process.

FIG. 4 is a diagram for explaining a cluster to be reduced;

FIG. 5 is a diagram for explaining an error evaluation value of a cluster.

FIG. 6 is a diagram for explaining a motion vector search in a cluster hierarchy.

FIG. 7 is a diagram for explaining a motion vector search in a macroblock hierarchy.

FIG. 8 is a diagram for explaining macroblocks constituting a frame to be encoded.

FIG. 9 is a diagram for explaining a motion vector search by a full search method.

FIG. 10 is a block diagram showing a configuration of a conventional motion vector search device.

FIG. 11 is a diagram showing a configuration of a cluster.

FIG. 12 is a diagram for explaining a relationship between a motion vector of a cluster and an initial vector of a macroblock in a macroblock macroclustering method.

[Explanation of symbols]

DESCRIPTION OF REFERENCE NUMERALS 1 detection target image division unit, 2, 12 prediction reference image division unit, 3 error calculation unit, 4 error evaluation unit, 5 error storage unit, 6 cluster layer skip determination unit, 7 search area determination unit, 8 macro block layer skip determination Part, 9 Initial setting value register, 10 Enlarged setting value register, 11
Selector, 100, 200 encoding target frame, 10
1,202 macroblock, 102,208 prediction reference frame, 103,205,211 search area, 10
4 partial areas, 105, 207, 209 vectors, 2
01 cluster, 203 cluster to be reduced, 204
Reduced prediction reference frame, 206 partial area cluster, 2
10 Partial area macroblock.

Continued on front page F term (reference) 5C059 KK19 MA00 MA05 MA23 NN03 NN09 NN24 NN28 NN30 NN34 PP04 SS07 UA02 5J064 AA03 BA01 BB03 BC04 BC25 BC29 BD02

Claims (10)

[Claims] 1. A position corresponding to a block included in a first frame is obtained from a second frame having a predetermined relationship with the first frame using motion compensation prediction, and a motion vector of the block is searched. A motion vector search device, comprising: a cluster generating unit for generating a cluster including one or more blocks in the first frame including a plurality of blocks; A cluster correlation determining unit for determining a correlation between the cluster and the region at the same position in the second frame; and a corresponding region in the second frame connected to the cluster correlation determining unit. Only for the clusters determined to be uncorrelated, the position corresponding to the cluster is A cluster motion vector detecting means for searching and detecting a motion vector of the cluster; connected to the cluster correlation determining means and the cluster motion vector detecting means; an output of the cluster correlation determining means and a motion of the cluster; Based on the vector,
A motion vector search device comprising: a position corresponding to a block included in the cluster in the second frame; and a motion vector detecting means for detecting a motion vector of the block. 2. The motion vector detecting means is connected to the cluster correlation judging means and the cluster motion vector detecting means, and based on an output of the cluster correlation judging means and a motion vector of the cluster, Initial vector setting means for setting an initial vector representing an initial value of a motion vector of the included block; connected to the initial vector setting means, and the block and the block after moving the block according to the initial vector. A block correlation determining unit for determining a correlation with a region in the second frame; connected to the block correlation determining unit, and determining that there is no correlation with a corresponding region in the second frame Only for the block that has been set, the position corresponding to the block is set to the second frame. The motion vector search device according to claim 1, further comprising: a block motion vector detecting unit configured to perform a search and detect a motion vector of the block. 3. The block motion vector detecting means is connected to the cluster correlation determining means and the block correlation determining means, and it is determined that there is no correlation with a corresponding region in the second frame. Only for the block, a search area setting means for setting a search area in the second frame based on an output of the cluster correlation determining means for a cluster including the block, and a search area setting means connected to the search area setting means. Means for searching for a position corresponding to the block in the set search area, and detecting a motion vector of the block.
The motion vector search device according to claim 2. 4. The search area setting means determines a search area when it is determined that there is no correlation between the cluster and a corresponding area of the cluster from a search area when it is determined that there is correlation. The motion vector search device according to claim 3, wherein is set to be large. 5. The cluster correlation determining unit includes: a first correlation determining unit configured to determine a correlation between the cluster and an area at the same position as the cluster in the second frame; For each of the blocks constituting the second frame, a second correlation determining means for determining the correlation between the block and an area at the same position as the block in the second frame; and the first and second correlation determination means Connected to a correlation determining means, and based on an output of the first correlation determining means and an output of the second correlation determining means, determine a correlation between the cluster and a corresponding region in the second frame. The motion vector search device according to any one of claims 1 to 4, further comprising means for determining. 6. A position corresponding to a block included in a first frame is obtained from a second frame having a predetermined relationship with the first frame using motion compensation prediction, and a motion vector of the block is searched. A motion vector search method, wherein in the first frame including a plurality of blocks, a cluster including one or more blocks is generated; and the cluster is the same as the cluster in the second frame. Determining the correlation between the position and the region; and determining the position corresponding to the cluster in the second frame only for the cluster determined to have no correlation with the corresponding region in the second frame. Explore more,
Detecting a motion vector of the cluster; and determining a corresponding position in the second frame for a block included in the cluster determined to have no correlation with a corresponding region in the second frame. Detecting a motion vector of the block. 7. The step of detecting a motion vector of the block is included in the cluster based on a correlation between the cluster and a corresponding region of the cluster in the second frame and a motion vector of the cluster. Setting an initial vector representing an initial value of a motion vector of the block; determining a correlation between the block and an area in the second frame after the block is moved according to the initial vector; Only for the block determined to have no correlation with the corresponding region in the second frame, a position corresponding to the block is searched from the second frame,
Extracting the motion vector of the block. 8. The step of extracting a motion vector of the block, the step of extracting a motion vector of the cluster including the block only for the block determined to have no correlation with the corresponding region in the second frame. Setting a search area in the second frame based on the correlation between the cluster and the corresponding area in the second frame; and determining a position corresponding to the block in the set search area. Searching and detecting a motion vector of the block. 9. The step of setting a search area in the second frame includes: determining a search area when it is determined that there is no correlation between the cluster and a corresponding area of the cluster; 9. The motion vector search method according to claim 8, wherein the search area is set to be larger than a search area when it is determined that there is a motion vector. 10. The step of determining the correlation between the cluster and a region located at the same position, the step of determining the correlation between the cluster and the region located at the same position as the cluster in the second frame. Determining, for each of the blocks constituting the cluster, a correlation between the block and an area at the same position as the block in the second frame; and a correlation for the cluster and the block constituting the cluster. 10. The motion vector search method according to claim 6, further comprising: determining a correlation between the cluster and a corresponding region in the second frame based on the correlation with respect to.