JP3237815B2 - Motion vector search method and apparatus - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention divides the image data of the current frame into small blocks with respect to a digital image signal, and stores the divided past motion detection target small blocks stored in a frame memory. Or search for a plurality of candidate blocks in the future reference frame, determine the similarity with each candidate block based on a predetermined evaluation criterion,
The present invention relates to a motion vector search method in which a vector to a candidate block most similar to the evaluation value is used as a motion vector.

[0002]

2. Description of the Related Art In highly efficient coding of moving images, motion compensation prediction is often used to remove inter-frame correlation. Motion compensation prediction utilizes similarity between the current frame and another frame that is temporally close.
In order to perform efficient motion compensation prediction, the direction and magnitude (motion vector) of motion indicating where a certain region of the encoding target frame exists in another frame that is temporally close to the frame must be accurately determined. It is important to ask.

In selecting a motion vector, the following method is usually used. The image data of the motion vector detection target small block has a size of N × N (N is an integer of 1 or more) and is set to B (i, j). B (i, j) is spatially represented by B (i, j) on the reference image in the frame memory.
j) is similar to a small block B ′ (i + u, j + v) shifted horizontally and vertically by (u, v).
Calculate (u, v). As the evaluation value, a sum of absolute differences and a sum of squares of differences are often used. (U, v) a defined search area _{(-R u <u <R u} , -R v <v <R v) is changed in, which gives the minimum value of the evaluation value D (u, v) ( u,
v) is a final motion vector (U, V), and a small block in the reference image defined by the motion vector (U, V) is a reference small block.

FIG. 5 is a block diagram showing the configuration of a typical conventional motion vector search apparatus. In this example, the reference image is an image one frame before in time, and the sum of absolute differences is used as an evaluation function.

[0005] Input image 1 input from input terminal 101
02 is divided by the small block dividing unit 103 into small blocks 104 which are units of motion search. On the other hand, the input image 102 is input to the frame memory 105, and an image 106 one frame before is obtained. Candidate vector setting unit 10
7 generates a vector in the search range determined by the search range setting unit 108 and generates a vector memory 113.
To send to. In the shift unit 109, the frame memory 10
5 is spatially shifted by the candidate vector set by the candidate vector setting unit 107 by that amount. In the subtracter 110, the difference between the shifted small block and the small block to be encoded 104 is calculated, and the operation result is input to the absolute value sum calculation unit 111. The absolute value sum calculator 111 calculates the sum of the absolute values of the differences in the small block, sends the calculated absolute value sum to the minimum value memory 112, and compares it with the minimum value stored in the memory at that time. Is performed. When the sum of absolute values is smaller than the minimum value, the minimum value is updated, an update instruction command is sent to the vector memory 113, and the stored vector is input from the candidate vector setting unit 107 at that time. Is replaced by the candidate vector.
When the generation of all the vectors in the search range determined by the candidate vector setting unit 107 is completed, the vector stored in the vector memory 113 is output to the output terminal 115 as the motion vector 114 of the small block to be encoded. At the same time, the minimum value memory 112 and the vector memory 113 are reset.

Here, the search range setting unit 108 normally
In advance -R_u <u<R_u , -R _v <v<R_v Search
The search range is set. Try to follow big movements
And R_u , R_v Must be set large, but the search range
The amount of computation to detect the motion vector when the surrounding area increases
Becomes enormous. When the amount of computation is a problem
Conventionally, R_u , R_v To a reasonable amount of computation
I was searching for something smaller.

[0007]

In the conventional method described above, when the amount of calculation must be reduced, the search range is narrowed for all the small blocks. There is no problem for small blocks, but for small blocks whose motion is larger than the search range, the most similar block is searched only within the search range, so an incorrect motion vector is obtained. There was a problem that it would.

An object of the present invention is to provide a motion vector search method and apparatus capable of detecting a correct motion vector by reducing the amount of calculation required for motion detection processing.

[0009]

According to the motion vector search method of the present invention, the small blocks of the current frame are classified into two types, a first small block and a second small block. Calculates a motion vector by searching for a candidate block in a predetermined first search range around a point in a past or future reference block that is spatially at the same position as that block, and calculates a second motion vector. motion to the small blocks, in the motion detection target small block at the same position, which in terms of the past or in the future reference blocks, are determined in the first small block in the vicinity of the motion detection target small block It is characterized in that a motion vector is calculated by searching for a candidate block in a second search range smaller than the first search range around a point shifted by the vector.

According to this method, the search range of the second small block is determined by using the motion vector obtained from the nearby first small block. Generally, in a moving image, there is a high probability that adjacent small blocks are moving in the same direction. Therefore, the search range of the second small block is
Even if it is limited to a small area centered on the motion vector obtained by the nearby first small block, it can be said that the probability of obtaining an incorrect motion vector is extremely low. From the viewpoint of the amount of calculation, it is possible to greatly reduce the amount of calculation for the second small block.

According to the embodiment of the present invention, a search center vector and a search range for a motion detection target small block in a second small block can be obtained from a plurality of first small blocks in the vicinity thereof. Set according to the distribution state of the motion vector.

At the boundary of a moving object, adjacent motion vectors do not always have the same direction and magnitude. In such a case, an erroneous motion vector may be obtained by referring to the motion vector of only one nearby first small block. The above embodiment is for avoiding this, and when setting the search range of the second small block, the motion vectors of a plurality of neighboring first small blocks are referred to. Therefore, if all of the plurality of motion vectors have the same direction and magnitude, the second small block is set to search a very small range around the same motion vector. If a plurality of motion vectors in the vicinity vary, it is dangerous to limit the search range, so that the search range may be set large.

A motion vector search apparatus according to the present invention includes a block classifying unit for classifying a small block of a current frame into two types, a first small block and a second small block, and a first classifying unit output from the block classifying unit. A motion vector storage unit that stores a motion vector obtained as a result of the motion vector search processing for the small block, and a motion detection target small block in the second small block output from the block classification unit,
In the motion vector detection target small block at the same position, the movement of the point in the reference vectors of the past or future, which are stored in the motion vector storage unit, the first small block
A search range setting unit that sets a second search range smaller than the first search range, centered on a point shifted by a motion vector with respect to a small block located near the small block to be subjected to vector detection processing.

[0014]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a diagram showing an example of small block classification in a motion vector search method according to an embodiment of the present invention, and FIG. 2 is an explanatory diagram of the motion vector search method.

First, the small blocks in the encoding target frame are classified into a first small block and a second small block as shown in FIG.

Next, a first small block (A in FIG. 2)
For and C), in the block and the same spatial position, predetermined around a point in the reference frame in the past or _{future, -R u <u <R u} , -R v <v
The search is performed in a first range < _Rv . On the other hand, the second
For small blocks (B and D in FIG. 2),
From the point at the same spatial position as the small block to be detected, the first small block in the vicinity (A,
A search is performed in a second range centered on a point shifted by the motion vector (U ₁ , V ₁ ) obtained in C).
_{That, U 1 -R' u <u <} U 1 + R'u, V 1 -R' v <
V <probed with V ₁ + _R'v Scope is executed. here,
The range searched in the second small block is set smaller than the range searched in the first small block. In other words, it is _{R'u <R u, R'v} <R v.

FIG. 3 is an explanatory diagram of a motion vector search method according to another embodiment of the present invention.

In the present embodiment, the small blocks are classified so as to alternate with the first small blocks and the second small blocks in the horizontal direction.

To set the search range of the second small block,
The motion vectors obtained in the first small blocks on both the left and right sides are used. That is, in FIG. 3, the small blocks A and C are used for setting the search range of the small block B.
Each motion vector determined by the small blocks A and _{C (U A, V A)} , and (U _C, V _C). At this time,
The search ranges u and v of the second small block B are set by the following equations.

[0021]

(Equation 1)In other words, the center is the motion vector of the left and right small blocks
Mean vector ((U _A + U_C ) / 2, (V_A + V_C )
/ 2), and ± | U_A -U _C ｜ /
2, ± | V in the vertical direction_A -V_C Search in the range of | / 2
You. U_A And U_C Are similar, the horizontal search range
The enclosure is very small and V_A And V_C But similar
If the value is set, the vertical search range becomes very small. sand
That is, as shown in FIG.
Very small range if the vectors are similar
Only search, and conversely, as shown in FIG.
When the motion vectors of the left and right blocks are significantly different
Is set to search a relatively large range.

As a result, if the motion vectors of the left and right small blocks A and C are almost the same, the motion vector of the small block B is likely to be in the same direction. For example, the amount of calculation can be reduced without detecting an erroneous motion. Conversely, if the motion vectors of the left and right small blocks A and C are significantly different,
Although the motion of the small block B cannot be predicted, no problem arises in that case because the setting is made to search a large range.

FIG. 4 is a block diagram showing the configuration of the motion vector search device according to one embodiment of the present invention. The same reference numerals in FIG. 5 indicate the same components.

The motion vector search apparatus according to the present embodiment classifies the small block 104 into two types, a first small block 117 and a second small block 118.
6 and a switch 119 for switching and outputting the first small block 117 or the second small block 118 to the subtractor 110.
And a motion vector storage unit 120 for storing the vector stored in the vector memory 113, which is obtained as a result of the motion vector detection processing for the first small block 117.
, The switch 122, and the motion vector storage unit 120 during the motion vector detection process for the second small block 118.
Is provided with a search range setting unit 108 for setting a search range around the motion vector 121 stored in.

Next, the operation of this embodiment will be described.

Input image 1 input from input terminal 101
02 is divided by the small block dividing unit 103 into small blocks 104 which are units of motion search. The divided small blocks 104 are classified into two types of a first small block 117 and a second small block 118 by a block classification unit 116. On the other hand, the input image 102 is stored in the frame memory 105.
To obtain the image 106 one frame before.

Firstly, at the time of motion vector detection processing for the first small block, switch 119 is switched to the upper side, the switch 12 2 is opened. Candidate vector setting unit 1
07 generates a vector in the search range predetermined by the search range setting unit 108 and sends it to the vector memory 113. The shift unit 109 spatially shifts the reference image 106 stored in the frame memory 105 by the candidate vector set by the candidate vector setting unit 107. Subtractor 118
In, the difference between the shifted small block and the small block to be encoded 117 is calculated, and the calculation result is input to the absolute value sum calculation unit 111. The absolute value sum calculator 111 calculates the sum of the absolute values of the differences in the small block, sends the calculated absolute value sum to the minimum value memory 112, and compares it with the minimum value stored in the memory at that time. Is performed.
When the sum of absolute values is smaller than the minimum value, the minimum value is updated, and an update instruction command is sent to the vector memory 113 to replace the stored vector with the candidate vector input at that time. Is done. When the generation of all the vectors in the search range determined by the candidate vector setting unit 107 ends, the vector stored in the vector memory 113 is output to the output terminal 115 as the motion vector 114 of the small block to be encoded. At the same time, after being sent to the motion vector storage unit 120, the minimum value memory 112 and the vector memory 113 are reset.

At the time of the motion vector detection processing for the second small block, the switch 119 is switched to the lower side, and the switch 122 is closed. In the search range setting unit 108, the first
With reference to the motion vector 121 for the small block located near the small block to be processed among the small blocks,
The motion vector is set to the second small
A search range is set around a point indicated by applying to the block . The candidate vector setting unit 107 generates a vector in the search range set by the search range setting unit 108. The following processing is the same as that of the first small block 117.

[0029]

As described above, the present invention has the following effects. (1) According to the first and third aspects of the present invention, it is possible to control the search range of the motion vector to be set small when the estimation from the neighboring motion vectors is easy, and to be set large when the estimation is difficult. . As a result, it is possible to detect a correct motion vector even if the amount of calculation required for the motion detection process is reduced, and it is possible to provide high reproduction image quality with a small amount of calculation when applied to high-efficiency encoding of an image or the like. Becomes (2) According to the second aspect of the invention, it is possible to avoid finding an erroneous motion vector.

[Brief description of the drawings]

FIG. 1 is a diagram showing an example of small block classification in a motion vector search method according to an embodiment of the present invention.

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

FIG. 3 is an explanatory diagram of a motion vector search method according to another embodiment of the present invention.

FIG. 4 is a block diagram of a motion vector search device according to an embodiment of the present invention.

FIG. 5 is a block diagram of a conventional example of a motion vector search device.

[Explanation of symbols]

 Reference Signs List 101 input terminal 102 input image 103 small block division unit 104 motion detection target block 105 frame memory 106 image one frame before 107 candidate vector setting unit 108 search range setting unit 109 shift unit 110 subtractor 111 absolute value sum calculation unit 112 minimum value Memory 113 Vector memory 114 Obtained motion vector 115 Output terminal 116 Small block classification unit 117 First small block 118 Second small block 119 Switch 120 Motion vector storage unit 121 Motion vector of neighboring small block 122 Switch

Continuation of front page (58) Field surveyed (Int.Cl. ⁷ , DB name) H04N ^7/ 24-7/68

Claims (3)

(57) [Claims] An image data of a current frame is divided into small blocks with respect to a digital image signal, and a divided motion detection target small block is divided into past or future reference frames stored in a frame memory. A plurality of candidate blocks are searched for, a similarity with each candidate block is obtained based on a predetermined evaluation criterion, and a motion vector search method using a vector to the most similar candidate block by the evaluation value as a motion vector The small blocks of the current frame are classified into two types, a first small block and a second small block, and the first small block is located at the same position as the small block in the past or the future. , A motion vector is calculated by searching for a candidate block in a predetermined first search range around a point in the reference block of For a click, dynamic
In the detection target small block at the same position can, mainly in terms of the past or in the future reference blocks, point shifted motion vector component obtained in the first small block in the vicinity of the motion detection target small block A motion vector is calculated by searching for a candidate block in a second search range that is smaller than the first search range. 2. A search center vector and a search range for a motion detection target small block in a second small block are determined by a motion vector distribution state obtained by a plurality of first small blocks in the vicinity thereof. set, the motion vector search method of claim 1. 3. For a digital image signal, image data of a current frame is divided into small blocks, and a divided motion detection target small block is divided into past or future reference frames stored in a frame memory. A plurality of candidate blocks are searched for, a degree of similarity with each candidate block is obtained based on a predetermined evaluation criterion, and a motion vector search apparatus that uses a vector to a candidate block most similar to the evaluation value as a motion vector by the evaluation value A block classification unit that classifies the small blocks of the current frame into two types, a first small block and a second small block; and a result of a motion vector search process for the first small block output from the block classification unit. A motion vector storage unit for storing the obtained motion vector, and a second small block output from the block classification unit. To the motion detection target small block, in the detection target small block at the same position-out animals, in terms of the past or in the future reference blocks are stored in the motion vector storage unit, the one of the first small block A second smaller than the first search range, centered on a point shifted by a motion vector with respect to a small block located near the motion detection target small block.
A motion vector search device, comprising: a search range setting unit that sets a search range.