JP2003111080A - Motion vector detector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a motion vector detector capable of reducing mis-detection of motion vectors as isolation points. SOLUTION: An accumulation difference arithmetic section 3 sums differences pixels of a target block BT and pixels of a reference block BR to obtain a difference accumulated value DA by each reference block BR within a range of detection of a motion vector before two fields with respect to the target block BT from an input terminal 1. A difference comparison section 4 obtains a reference block BR where the difference accumulated value DA is minimized in a motion vector detecting range, uses the position vector of the reference block with respect to the target block BT for a motion vector MV of the target block BT, and outputs a position of a reference block closer to an average vector (field vector FV) of a field including the target block detected by a field vector detection section 5 among the reference blocks BR when a plurality of the reference blocks BR with the minimum difference accumulated value DA exist.

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 for detecting motion information from moving image data.

[0002]

2. Description of the Related Art The motion vector at each point on a moving image is
For example, it is important information that serves as a basis for various moving image processing such as motion compensation predictive coding processing in MPEG2, and various studies have been made in the past. One of the most widely used conventional motion vector detection methods is
One is a method called a block matching method. Hereinafter, this method will be described with reference to FIG.

FIG. 2 is a diagram showing the concept of the block matching method. 101 is a current field, 102 is a reference field (for example, a field two fields before the current field 101), 103 is a target block, 1
Reference numeral 04 is a reference block, 105 is a motion vector, 106 is a search range, and 107 is a subject (indicated by a smiling expression).

In the figure, a NicoNico mark is used as a subject 107 to be detected, and a screen in a reference field 102, for example, two fields before the current field 101 (hereinafter, this is also a reference field unless otherwise specified). A subject 107 located at a position illustrated in FIG. 102 has been moved to a position illustrated in the screen of the current field 101 (hereinafter also abbreviated to current field 101 unless otherwise specified). And In the same field, for example, the vector between the position of the subject 107 in the reference field 102 and the position of the subject 107 ′ obtained by projecting the subject 107 of the current field 101 onto the reference field 102 is the motion vector 1 of this subject 107.
05.

In the block matching method described above, the screen of the current field 101 is divided into mini blocks (for example, 8 pixels × 8).
A block of a line, which will be referred to as a target block 103 hereinafter, is divided into blocks of the same size in the reference field 102 (that is, the reference block 104), and while shifting the position in the current field 101, The correlation with the target block 103 is calculated. And the target block 1 with the highest correlation
The shift amount (distance and direction) of the reference block 104 that matches 03 is set as the motion vector 105. The calculation of this correlation is usually done by target block 103 and reference block 1
04, the absolute value of the difference between each corresponding pixel is taken, and these are cumulatively added, and the addition result (correlation value) is the smallest, that is, the shift amount of the reference block 104 having the highest correlation. Is the motion vector 105.

The shift range of the reference block 104 is determined to be, for example, 3 × 5 blocks, and this shift range is referred to as a motion vector search range 106. The size of the search range 106 determines the direction and size (speed) of the movement of the subject 107 that the movement vector 105 can deal with. That is, a wide search range is desirable so that all movements of the object on the screen can be distinguished.

As described above, the target block 103
The motion vector can be obtained by detecting the reference block 104 that has the highest correlation with. However, when detecting the reference block 104, in some cases,
As will be described later, a plurality of reference blocks within the search range 106 may have the maximum correlation with the same correlation value, and thus a plurality of motion vectors may be obtained. In this case, it is necessary to have a function of giving priority to motion vectors according to some method and selecting one of the motion vectors having the same correlation value.

A conventional example of such a function is described in Japanese Patent Laid-Open No. 6-30399.

FIG. 3 is a table showing the priorities of motion vectors described in Japanese Patent Laid-Open No. 6-30399. Hereinafter, FIG. 3 will be described.

In the figure, the numbers shown in the upper and left ends respectively indicate the horizontal and vertical components (horizontal and vertical vectors) of the motion vector, and the numerical values in the table are the horizontal vector above and the left side thereof. And the vertical vector shown in FIG. Further, the larger this value, the higher the priority. For example, the motion vector (horizontal vector, vertical vector) = (0,0), that is, the highest priority is 15 for the reference block at the same position as the target block. Also, the motion vectors (-7, -5), (-7,-)
Since the priorities for 7) are 3 and 1, respectively, these motion vector (-7, -5) and motion vector (-7,
If the two correlation values with respect to −) are the same and the correlation is the maximum in both cases, the motion vector (−7, −5) having a higher priority is adopted.

Further, there may be a case where a plurality of motion vectors having the same correlation value have the maximum correlation and the same priority value. For example, in both cases, the motion vector (-7, -5) and the motion vector (-5,-
7) In such a case, the left side of the table, that is, the horizontal vector of the motion vector is closer to the negative maximum value, in this case, the motion vector (-
7, -5) has priority. Further, in the case where both have a priority of 3 and the horizontal vectors are equal to each other, such as a motion vector (-7, -5) and a motion vector (-7, 5), the upper side of the table, that is, the motion The direction in which the vertical vector of the vector is closer to the negative maximum value is determined to have priority.

[0012]

However, according to the method of determining the priority order of the motion vectors as described above, the following problems occur.

That is, now, let us consider the window portion of the building 110 when the image of the building as shown in FIG. 4A is moving in the arrow direction (vertical direction). Here, it is assumed that the image of the window portion is a stripe-shaped image (for example, a repeating pattern of 5 lines of black and 3 lines of white) that repeats in several lines, and this image moves in the vertical direction.

FIG. 4B shows the search range of the motion vector in the reference field, and FIG. 4C shows the range corresponding to the search range of this motion vector in the current field two fields later. There is. Then, as shown in FIG.
In any of (c), one block is 8 pixels × 8 lines, and the motion vector search range is 3 × 5 blocks. Here, as an example, it is assumed that the image vertically moves from the top to the bottom by 6 lines between two fields, and the horizontal vector is not considered.

In this case, according to the block matching method, the correlation is highest with the target block C in the current field shown in FIG. 4C, that is, the difference value (correlation value) with the target block C is the minimum. Two reference blocks A and B obviously exist as reference blocks in the reference field shown in FIG. 4B.

Here, the vertical vector mvA of the motion vector for the reference block A is -6, the vertical vector mvB of the motion vector for the reference block B is 2, both of these horizontal vectors are 0, and their priority order is Looking at FIG. 4, when fitted to the table shown in FIG. 4, the priority of the motion vector for the reference block A is 19, the priority of the motion vector for the reference block B is 23, and the priority of the motion vector for the reference block B is 23. The motion vector for is to be selected.

However, as is clear from FIGS. 4B and 4C, the motion vector to be originally detected is the motion vector for the reference block A, but the motion vector for the reference block B is selected. As a result, a motion vector that has nothing to do with the motion of the subject (in this case, the window) is detected.

Using such an incorrect motion vector,
For example, when the scanning line interpolation process is performed, a non-interlaced signal in which a signal having completely different information is used as an interpolation signal is generated, which causes block-shaped distortion and deteriorates image quality.

Further, there is a possibility that the above case may occur due to the influence of noise contained in the image signal, and furthermore, there may be a motion vector having a correlation larger than that of the correct motion vector. In some cases, the vector may not accurately reflect the motion of the subject in the image.

It is an object of the present invention to provide a motion vector detecting device which solves such a problem and can reduce erroneous detection of a motion vector.

[0021]

In order to achieve the above object, the present invention provides a reference block having the highest correlation with a target block among reference blocks within a motion vector detection range two fields before the target block. A motion vector detection device that uses a block position vector as a motion vector of a target block, and has the highest correlation with a target block and first means for generating a field vector representing an average motion of the entire screen for each field. When there is only one reference block, the position vector of this reference block is used as the motion vector of the target block, and when there are a plurality of reference blocks that have the highest correlation with the target block, the first of the plurality of reference blocks is selected. 1 fee for target block generated by means In which a structure having a second means the distance between de previous field vector is selected as a motion vector a position vector of the closer the reference block.

Here, the plurality of reference blocks having the highest correlation with the target block are reference blocks whose correlation values are within a predetermined value range that can be arbitrarily set.

The first means divides the screen into a plurality of areas having blocks as constituent units, obtains an average vector of motion vectors obtained from the second means for each area, and calculates all areas. When the average vectors of 1 are equal to each other, or when the average vectors of a predetermined number of regions or more are equal to each other, this equal average vector is used as a field vector.

Further, the first means divides the screen into a plurality of areas each having a block as a structural unit, and makes a second division for each area.
When the average vector of the motion vectors obtained from the means is obtained, and the number of average vectors that are close to each other within a predetermined range that can be arbitrarily set is equal to or larger than an arbitrary value that can be arbitrarily set, these are approximated to each other. The vector obtained by averaging the average vector is used as the field vector.

Further, the number of target blocks of the motion vector within the first predetermined range in which the distance to the field vector can be set arbitrarily is calculated, and when this number is equal to or larger than a predetermined value that can be set arbitrarily, A third means is provided in which it is determined that the entire screen is moving at a constant speed, and this field vector is used as the motion vector of the target block on the entire screen.

[0026]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a first embodiment of a motion vector detecting device according to the present invention, in which 1 is a current field data input terminal, 2 is a reference field (for example, previous two fields) data input terminal, and 3 is a cumulative difference. Arithmetic unit,
Reference numeral 4 is a difference comparison unit, 5 is a field vector detection unit, 6 is a field vector valid / invalid determination unit, and 7 is a motion vector output terminal.

In the figure, from the input terminal 1, a block of the current field (hereinafter referred to as a target block) B
Data of T is input and, for example, a block two fields before (two fields before) the target block BT from the input terminal 2 (hereinafter referred to as a reference block).
BR data is input. By the block matching process described below, the motion vector MV of this target block BT is set to the reference block B within the motion vector detection range set for this target block BT.
It is detected as the position vector of the reference block selected from R with respect to this target block BT.

That is, in the cumulative difference calculation unit 3, the input terminal 2
The difference calculation between the corresponding pixels of the target block BT and the reference block BR is performed for each reference block BR input to the target block BT (that is, within the motion vector detection range set for the target block BT). As a result, the cumulative addition result of these differences (hereinafter referred to as difference cumulative value DA) is obtained and supplied to the difference comparison unit 4. The difference comparison unit 4 calculates the difference cumulative value DA of the first reference block with respect to the target block BT.
Is supplied as the minimum difference cumulative value DA and is held together with the position vector of this first reference block, and when the difference cumulative value DA of the next reference block is supplied, this and the held minimum The difference cumulative value DA is compared, the smaller difference cumulative value DA is held as the minimum cumulative value together with the position vector of the reference block, and so on. Similarly, the difference cumulative value DA within the same motion vector detection range is held. The reference block having the minimum DA is detected, and its position vector for this target block BT is output as the motion vector of this target block BT.

In this way, from the difference comparison unit 4, for each target block BT input from the input terminal 1,
The motion vector MV is obtained and output from the output terminal 7.

However, the difference comparison unit 4 may detect a plurality of reference blocks of the motion vector having the smallest difference accumulated value DA within the same motion vector search range.
In this case, the difference comparison unit 4 causes the field vector detection unit 5 to output, as a motion vector, the position vector of the reference block closer in distance to the field vector FV obtained from the previous field, as described later. .
For this distance determination, for example, the absolute difference value between horizontal and vertical vectors is used.

As will be described later, when the field vector FV is invalid, for example, the motion vector is determined according to the priority order described in FIGS. Further, even if the field vector FV is valid, even when there are a plurality of reference blocks of position vectors having the same distance from the field vector FV, the motion vector is calculated according to the priority order described in FIGS. 3 and 5, for example. Decide

Further, in the block matching process of the first field at the start of the search, the field vector FV cannot be obtained from the field vector detection unit 5. Therefore, in the first field, when there is a target block BT in which a plurality of reference blocks BR having the smallest difference cumulative value DA exist within the same motion vector search range, the difference comparison unit is executed for the target block BT. 4 (FIG. 1) determines one of the position vectors of the plurality of reference blocks as a motion vector based on the priority order shown in FIG. 3 or 5.

In this way, in this embodiment, when the same target block BT has a plurality of reference blocks BR of the position vector having the smallest difference cumulative value DA,
The field vector FV is used to determine the motion vector of this target block BT from the position vector of these reference blocks BR.
Is invalid or the plurality of position vectors have an equal distance from the field vector FV, the position vector of the reference block closest to the field vector FV is determined as the motion vector. However, for the target block of the first field, the motion vector is determined by the method described with reference to FIGS.

The field vector validity / invalidity determining unit 6 determines validity / invalidity of the field vector FV detected by the field vector detecting unit 5 based on the field vector valid / invalidity flag Vg1 obtained by the field vector detecting unit 5. When the field vector valid / invalid judgment unit 6 judges that the field vector FV is invalid, as described above, this field vector FV
Is not used by the difference comparison unit 4.

It should be noted that there is no problem even if the field vector validity / invalidity judgment unit 6 is incorporated into the field vector detection unit 5.

FIG. 6 is a block diagram showing a specific example of the field vector detection unit 5 in FIG. 1, in which 8 and 9 are input terminals, 10 is horizontal / vertical edge detection units, 11 is an AND gate, and 12 is cumulative. An addition unit, 13 is a field vector calculation unit, 14 is a field vector smoothing unit, 15 is a block counter with edges, 16 is a threshold value block control unit with edges, and 17 and 18 are output terminals. Also, FIG.
Is a flowchart showing the operation of this specific example.

In FIG. 6, the motion vector MV detected by the difference comparison unit 4 is input from the input terminal 8 for each target block BT, and the motion vector M is input from the input terminal 9.
The target block BT for V is input. Here, in the block matching process of the first field at the start of the search, the accumulated difference value is within the motion vector search range set for the target block BT input from the input terminal 9. When there are a plurality of reference blocks BR having the smallest DA, the difference comparison unit 4 (FIG. 1) determines and outputs a motion vector based on the priority order shown in FIG. 3 or 5. The output motion vector MV is input from the input terminal 8.

The horizontal / vertical edge detection unit 10 detects a target block BT having an image edge in both the horizontal and vertical directions among the target blocks BT input from the input terminal 9 (step F7 in FIG. 7). Based on the detection result of the horizontal / vertical edge detection unit 10, the AND gate 11 selects from among the motion vectors MV input from the input terminal 8 the motion vector MV of the target block BT having edges in both the horizontal and vertical directions. Is extracted and supplied to the cumulative addition unit 12. Cumulative adder 1
2 is a motion vector M supplied from the AND gate 11.
The values of V are sequentially added (step F3 in FIG. 7).

Here, the reason why the motion vector used for obtaining the field vector FV, that is, the motion vector MV sequentially added in the cumulative addition unit 12 is limited to the motion vector of the target block having horizontal / vertical edges, is as follows. explain.

When the motion vector is calculated by using the block matching process, the size of the motion vector is 0 when the target image is a homogeneous one that does not include a pattern.
However, in reality, there is a high possibility that the motion vector originally obtained cannot be obtained due to the influence of noise contained in the image signal. Therefore, when calculating the field average vector (that is, the field vector) FV, it is better to use only the motion vector of the portion (block including the pattern) having horizontal / vertical edges in terms of accuracy and calculation amount. In, it is effective. It should be noted that the target block having either one of the horizontal and vertical edges may be targeted without including both the horizontal and vertical edges.

On the other hand, the block counter with edge 15 has
The number of target blocks (hereinafter referred to as edge blocks) determined to have edges by the horizontal / vertical edge detection unit 10 is counted for each field (step F in FIG. 7).
8), the field vector calculation unit 13 calculates the field from the count value of the block with edge from the block with edge counter 15 and the cumulative addition value of the motion vector MV of the block with edge in the field from the cumulative addition unit 12 The average value of the motion vectors of the target blocks within the area (hereinafter referred to as the average motion vector) is calculated (step F4 in FIG. 7).

The field vector smoothing unit 14 smoothes the average motion vector from the field vector calculating unit 13 (step F5 in FIG. 7), and outputs the result as the field vector FV from the output terminal 17 to the valid field vector. / Output to the invalidity determination unit 6 (FIG. 1) (step F6 in FIG. 7). This field vector FV
Indicates the direction and size of the overall movement of the screen in that field. Field vector smoothing unit 1
As a method of smoothing the average motion vector by 4, the average value of the average motion vectors of 2 to several fields is calculated, for example.

Further, the block number threshold control unit 16 compares, for each field, the number N _{1 of} blocks with edges obtained by the block with edges counter 15 with a preset threshold value N _TH (step of FIG. 7). F9), when the number N _{1 of} blocks with edges is larger than the threshold value N _TH , the field vector valid / invalid flag Fg1 of “0” that validates the field vector FV output from the field vector smoothing unit 14 When the number N _{1 of} blocks with edges is less than or equal to the threshold value N _TH , the field vector FV calculated at that time is considered to have low reliability, and this field vector FV is invalidated as a “1” field vector. The valid / invalid flag Fg1 is output, and the field vector valid / invalid determination unit 6 is output from the output terminal 18, respectively.
(FIG. 1) (step F10 in FIG. 7). Although the threshold value N _TH can be set to an arbitrary value, it is, for example, 50% of the total number of target blocks in one field.

Field vector valid / invalid flag Fg
When 1 is “0”, in the difference comparison unit 4 (FIG. 1) of the block matching process in the next field, there are a plurality of reference blocks having the smallest difference cumulative value DA with the target block within the motion vector search range. As a motion vector determining means when it exists, the distance from the field vector FV is used to determine whether the field vector is valid /
When the invalid flag is “1”, for example, the priority order described in FIGS. 3 and 5 is followed (steps F1 and F in FIG. 7).
2).

Here, the position vector closer in distance to the field vector FV is the position vector most suited to the movement of the entire screen of the field. In the first embodiment, this is calculated from a plurality of position vectors. It is selected and used as the motion vector of the target block.

As described above, in the first embodiment,
In the case of an image in which the subject moves on the screen at a constant speed as shown in FIG. 4, an average vector (FV) indicating the movement of the entire screen is calculated, and the difference cumulative value DA with the target block in the block matching process is calculated. Since the position vector of the reference block having the shortest distance to the average vector FV is selected as the motion vector as a priority order when a plurality of reference blocks having the smallest size are present in the motion vector search range, there is no relation to the motion of the subject. It is possible to prevent a phenomenon such as the detection of a motion vector with no motion.

FIG. 8 is a block diagram showing a second embodiment of the motion vector detecting apparatus according to the present invention. Reference numeral 19 is a difference allowable value determining unit, and the portions corresponding to those in FIG. A duplicate description will be omitted.

In the second embodiment, when there are a plurality of motion vectors in the detection range including the motion vector whose correlation with the target block in the block matching processing is the maximum or a value equivalent thereto, the field vector F
This is to give a high priority as a motion vector to a position vector of a reference block closer to V.

In FIG. 8, the difference cumulative value DA of each reference block obtained by the cumulative difference calculation unit 3 is supplied to the difference allowable value determination unit 19 together with the difference comparison unit 4. The difference allowable value determination unit 19 determines, for example, the difference accumulated value DA for the two reference blocks that are the comparison targets in the difference comparison unit 4.
Absolute difference value is calculated, and when the absolute difference value is equal to or less than a preset allowable value, the difference cumulative value DA of these two reference blocks is considered to be the same value, and the difference value allowable flag Fg2 of “1” is calculated. Is output to the difference comparison unit 4. Otherwise, the difference value allowance flag Fg2 of “0” is output to the difference comparison unit 4. Here, this allowable value can be arbitrarily set.

The difference comparing section 4 is similar to the difference comparing section 4 in the first embodiment shown in FIG. 1, and the difference cumulative value DA held as the minimum difference cumulative value DA and the cumulative difference calculating section 3 are used. Although the operation of comparing the supplied cumulative difference value DA is repeated every time the cumulative difference calculation unit 3 supplies the cumulative difference value DA, the reference block having the minimum cumulative difference value DA is detected. While the operation is in progress, the difference value allowance flag Fg2 of “1” is output from the difference allowance value determination unit 19.
, The difference cumulative value DA of the two reference blocks being compared at this time becomes the same value, and these 2
Of the two reference blocks, the reference block whose distance from the position vector is closer to the field vector FV is set as a block having a high correlation with the target block BT, and the difference cumulative value DA and the position vector of this reference block are held. It is used as the minimum difference cumulative value DA for comparison with the difference cumulative value DA. When the difference value allowance flag Fg2 is “0”, the difference comparison unit 4 determines the difference accumulated value DA of the reference block having the smaller difference accumulated value DA.
Is held as the minimum difference cumulative value.

In this way, even if the difference cumulative value DA of the reference block indicating the motion vector to be originally detected is not affected by noise contained in the image signal and takes the minimum value within the detection range, The position vector of this reference block can be selected as the motion vector.

FIG. 9 is a block diagram showing a modification of the second embodiment shown in FIG. 8, in which 20 and 21 are line memories, 22 is a maximum / minimum difference calculation unit, and corresponds to FIG. The same reference numerals are given to the portions to be described, and the duplicated description will be omitted.

In the embodiment shown in FIGS. 1 and 8, the difference comparing section 4 compares the difference accumulated values DA of two preceding and following reference blocks to detect the reference block having the smallest difference accumulated value. However, the difference cumulative value DA of an arbitrary number of consecutive reference blocks of 3 or more may be compared and calculated. In the modification shown in FIG. 9, as an example, the difference cumulative value DA of three consecutive reference blocks is compared and calculated.

In FIG. 9A, the target block BT input from the input terminal 1 is supplied to the cumulative difference calculation unit 3 and the field vector detection unit 5. Further, the reference block BR1 input from the input terminal 2 is supplied to the cumulative difference calculation unit 3 and is also stored in the line memory 20 as 1
Reference block B delayed by one line (one horizontal scanning period)
It is supplied to the cumulative difference calculation unit 3 as R2. Further, the reference block BR2 from the line memory 20 is further delayed by one line in the line memory 21, and the reference block BR2 is delayed.
3 is supplied to the cumulative difference calculation unit 3.

Now, assuming that the target block BT and the reference block BR are blocks each having 8 pixels × 8 lines,
Reference blocks BR1, BR2, BR having the same timing as the target block BT input from the input terminal 1
3 has a positional relationship as shown in FIG. 9B, for example. Here, the reference blocks BR1, BR2, BR3 are
Although it is a block of a field two fields before the field of the target block BT, FIG. 9B shows it together with the target block BT within the vector detection range for the target block BT. here,
The vector detection range is composed of 24 pixels x 24 lines. Therefore, 3 x 3 is included in this vector detection range.
There will be number of reference blocks.

Therefore, the reference block BR1 input from the input terminal 2 at the same timing as the target block BT input from the input terminal 1 is the 17th vector detection range.
If the block from the second line (called 16H. The same applies below; however, the first line is 0H) to 23H, the reference block BR2 output from the line memory 20 at the same timing as this is from 8H to 15H. The line memory 21 is a block and at the same timing as this.
The reference block BR3 output from is a block of 0H to 7H. These reference blocks BR1 to BR3 are supplied to the cumulative difference calculation unit 3 together with the target block BT.

In the cumulative difference calculation unit 3, the cumulative difference value DA1 between the target block BT and the reference block BR1, the cumulative difference value DA2 between the target block BT and the reference block BR2, and the cumulative difference value between the target block BT and the reference block BR3. DA3 and DA3 are calculated and supplied to the difference comparison unit 4 and the maximum / minimum difference calculation unit 2
2 is supplied. The maximum / minimum difference calculation unit 22 obtains the maximum value and the minimum value in the difference accumulated values DA1 to DA3 of these three reference blocks, and calculates the difference value between these maximum values and the minimum values. Then, the difference allowable value determination unit 1
When the difference value calculated by the maximum / minimum difference calculation unit 22 is less than or equal to a preset allowable value, reference numeral 9 indicates cumulative difference values DA1 to DA3 of these three reference blocks BR1 to BR3.
Are the same value, the difference value allowance flag Fg2 of "1"
Is output to the difference comparison unit 4. If not, the difference allowable value determination unit 19 determines the difference value allowable flag Fg of “0”.
2 is output to the difference comparison unit 4.

Therefore, when the difference value allowance flag Fg2 of "1" is supplied, the difference comparison unit 4 makes three references that the difference cumulative value DA supplied from the cumulative difference calculation unit 3 at that time is the same value. Of the blocks, the field vector F
A reference block having a position vector closer to V is set as a block having the highest correlation with the target block BT,
The difference cumulative value DA is compared with the minimum cumulative value currently held. Also, the difference value allowance flag F of “0”
When g2 is supplied, the difference comparison unit 4 selects the difference accumulation value having the smallest value among the difference accumulation values DA supplied from the accumulated difference calculation unit 3 at that time and holds the minimum accumulation held at that time. Compare with the value. In any case, the difference cumulative value DA of the reference block that is smaller than the comparison is held as the minimum difference cumulative value together with the position vector.

Thus, also in this modification,
The difference cumulative value DA of the reference block indicating the motion vector to be originally detected is selected as the motion vector even if it does not take the minimum value within the detection range due to the influence of noise or the like contained in the image signal. Become.

The difference allowable value determining unit 19 may be provided in the difference comparing unit 4.

FIG. 10 is a block diagram showing a third embodiment of the motion vector detection device according to the present invention, in which 23 is a field vector detection / constant speed determination unit and 24 is a changeover switch, which corresponds to FIG. The same reference numerals are given to the parts, and the duplicated description will be omitted.

The third embodiment relates to a motion vector for a scrolling image,
If the screen is determined to be moving at a constant velocity by the constant velocity determining means for detecting the constant velocity motility of the entire screen such as a vertical scroll image for each field, the screen is displayed. The field vector FV is selected as the entire motion vector.

In FIG. 10, the field vector detection / constant velocity determination section 23 determines the field vector FV and determines whether or not the entire screen is moving at a constant velocity, and a constant velocity determination flag Fg3 representing the determination result. Is output. The constant velocity determination flag Fg3 is “1” when it is determined that the entire screen is moving at a constant velocity, and is “0” otherwise.

When the constant speed determination flag Fg3 of "1" is supplied, the change-over switch 24 receives the field vector F obtained by the field vector detection / constant speed determination unit 23.
When V is selected and the constant velocity determination flag Fg3 of "0" is supplied, the motion vector for each target block obtained by the difference comparison unit 4 is selected, and the output vector 7 is output as the motion vector for each target block. Output from. Therefore, when the entire screen is moving at a constant speed, the motion vector of the target block in the entire field is
The field vectors FV are all equal.

FIG. 11 is a block diagram showing a specific example of the field vector detection / constant velocity determination unit 23 in FIG. 10, 25 is a constant velocity determination unit, and 26 is an AND gate, which corresponds to FIG. Are denoted by the same reference symbols and redundant description will be omitted. FIG. 12 is a flowchart showing the operation of this specific example.

In FIG. 11, the operation of the portion corresponding to the specific example shown in FIG. 6 is similar to the operation shown in FIG. 7 (steps F1 to F10).

The constant velocity determination unit 25 is supplied with the motion vector of the block with edges output from the AND gate 11 and the unsmoothed field vector FV output from the field vector calculation unit 13 and operates as follows. To do.

That is, first, the distance from the field vector FV is calculated for each motion vector of the supplied block with edges (step F11). As the calculation of this distance, for example, the absolute value difference between the motion vector and the field vector FV is taken for each of these horizontal and vertical components, and these are added, and this added value is added to the field vector FV. As the distance parameter.

Next, the motion vector of the edged block whose distance parameter value is equal to or less than the preset predetermined value is counted, and the edged block whose distance parameter value in the field is equal to or less than the preset predetermined value is counted. The number (count value N ₂ ) of is calculated (step F12). The predetermined value can be set arbitrarily.

[0070] Then, the count value N ₂ of the predetermined proportions arbitrarily set with respect to the total number N ₁ of an edge perforated block in the field alpha (where, 0 <α <1) and comprising a threshold alpha N ₁
(Step F13 in FIG. 12), and if N ₂ ≧ αN ₁ , it is determined that the screen is moving at a constant velocity, and a constant velocity determination flag Fg3 of “1” is generated (step F1 in FIG. 12).
4). When N ₂ <αN ₁ , a constant speed determination flag Fg3 of “0” is generated (step F15 in FIG. 12).
However, when the field vector valid / invalid flag Fg is "0" indicating invalid by the AND gate 26, the constant velocity determination unit 25 sets the constant velocity determination flag Fg3 of "1".
Even if occurs, the constant velocity determination flag Fg3 is set to "0" (step F15).

As described above, in this embodiment, when the subject moves on the screen at a constant speed, the constant velocity motility is determined, and the field vector FV is applied as the motion vector of the entire screen to move the object. It is possible to reduce false detection such as isolated points of a vector.

FIG. 13 is a block diagram showing a specific example of the field vector detecting section 5 in the fourth embodiment of the motion vector detecting device according to the present invention, which is 11 ₁ , 11 ₂ , ...
…, 11 _m is an AND gate, 12 ₁ , 12 ₂ ,…, 12
₁₁ is a cumulative addition unit, 13 ₁ , 13 ₂ , ... 13 _m is an average vector (field vector) calculation unit, 27 is a screen division unit, 28 is an OR circuit, 29 is a determination unit, and corresponds to FIG. Are denoted by the same reference symbols and redundant description will be omitted. However, m is an integer of 2 or more. FIG. 14 is a flow chart showing the operation of the field vector detecting section 5.

In this specific example, the screen is divided into m areas, the average vector of motion vectors is obtained for each of these areas, and the field vector FV of this screen is obtained from these average vectors. An example of this screen division is shown in FIG.

In FIG. 13, the motion vector MV for each target block BT input from the input terminal 8 is supplied to the AND gates 11 ₁ , 11 ₂ , ..., 11 _m . In addition, the screen dividing unit 27 outputs signals (referred to as region signals) indicating the range of each region when the screen is divided into m, and the region signals are separately AND gates 11 ₁ , 1 1.
It is supplied to 1 ₂ , ..., 11 _m . Further, the detection output of the horizontal / vertical edge detection unit 10 is also supplied to the AND gates 11 ₁ , 11 ₂ , ..., 11 _m . This allows
In the AND gates 11 ₁ , 11 ₂ , ..., 11 _m , the motion vector of the block with edge is distributed for each area (step F101 in FIG. 14).

Here, when m = 4 and the screen division of FIG. 15A is taken as an example, the motion vector of the block with edges in the input motion vector MV is the area A, B,
It will be sorted by C and D.

AND gate 11 ₁ , 11 ₂ , ..., 11 _m
The motion vectors from are respectively added to the cumulative addition units 12 ₁ , 12 ₂ , ...
, 12 _m, and the cumulative addition value of the motion vector values is obtained for each area. These cumulative addition values are supplied to the respective average vector calculation units 13 ₁ , 13 ₂ , ..., 13 _m . On the other hand, in the horizontal / vertical edge detection unit 10, using the area signal from the screen division unit 27, the block with edges is sorted for each area from the target block BT input from the input terminal 9, and the block with edge counter 15 is provided.
Counts the number of blocks with edges for each region. Each average vector computing unit 13 _1, 13 _2, ..., 13 _m are each, accumulative adder 12 _1, 12 _2, ..., a cumulative addition value supplied from 12 _m was obtained in the edge perforated block counter 15 It is divided by the number of blocks with edges in the corresponding area to calculate the average vector BV of motion vectors for each area (step F102 in FIG. 14). These average vectors BV
Is supplied to the determination unit 29.

FIGS. 15 (b), 15 (c) and 15 (d) respectively show m.
= 4, the average vector B in each area A, B, C, D
It is a figure which shows the example of V.

The judging section 29 compares the supplied average vectors BV of the respective areas with each other (step F10 in FIG. 14).
3) When all the average vectors BV match (for example, FIG. 15B), or when the ratio of the number of the average vectors BV that match each other to the total average vector number is equal to or more than a preset threshold value. (For example, in FIG.
(C)), this matching average vector BV is output as the field vector FV (step F10 in FIG. 14).
4). When this is not satisfied, the field vector FV is not output (step F105 in FIG. 14). The threshold can be set arbitrarily.

The AND gates 11 ₁ , 11 ₂ , ...,
All the motion vectors from 11 _m are supplied to the constant velocity motion determination unit 25 via the OR circuit 28, and the same processing as the constant velocity motion determination unit 25 in FIG. 11 is performed.

FIG. 16 is a flow chart showing another example of the operation of the specific example of the field vector detecting section 5 shown in FIG.

In the figure, steps F201 and F202
14 is an operation of a configuration unit before the determination unit 29.
Is the same as steps F101 and 102 in
The operation part 9 will be described.

The determining unit 29 determines that all the supplied average vectors BV of the respective regions have the same directivity, or
As shown in FIG. 15D, when the ratio of the average vector BV having the same directionality is equal to or more than a preset threshold value (step F203 in FIG. 16), the average vector BV having the same directionality is calculated. Average value (hereinafter, average vector A
BV) is calculated (step F204 in FIG. 16),
This is output as the field vector FV (FIG. 16).
Step F206). The threshold value in this case can also be set arbitrarily.

Here, as shown in FIG. 17, when the vector value (0,0) with no motion is used as a reference and divided into four areas of 90 °, the same directionality is assigned to the same area. It is assumed that the average vector BV has.

As described above, according to this embodiment, it is possible to detect the field vector FV with higher accuracy.

Further, for example, with respect to the constant velocity motion of a "deep" image as shown in FIG. 18, since the velocity of the motion is different between the position closer to the camera and the position far from the camera, the detected motion vector is also The distant motion vector MVA and the motion vector MVB of the near video are different. Therefore, when it is determined that the screen as shown in FIG. 18 is constant velocity and the field vector FV is applied, the motion vector (field vector FV) of the entire screen is either one of the motion vectors MVA and MVB, or these Is not the correct motion vector. However, this embodiment can solve such a problem.

FIG. 19 is a block diagram showing an embodiment of a scanning line interpolating device using the motion vector detecting device according to the present invention, in which 30 is an input terminal, 31 and 32 are field memories for storing moving image signals, 33 is a motion vector detection circuit for detecting a motion vector in block units, MV is a motion vector, 34 is a motion compensation processing circuit, 35 is a non-interlaced signal output circuit, and 36 is a non-interlaced image signal output.

In this embodiment, a non-interlaced image signal is generated from an interlaced image signal by scanning line interpolation, and a motion vector is used to form a scanning line to be interpolated.

In the figure, an interlaced image signal (hereinafter referred to as post data) is input from the input terminal 30 and supplied to the motion vector detection circuit 33.
It is stored in the field memory 31 and delayed by one field, and further stored in the field memory 32 and delayed by one field.

Here, the field output from the field memory 31 is called the current field F _0, and therefore the field input from the input terminal 30 at the same timing as this current field is called the rear field F ₁ and this current field. A field output from the field memory 32 at the same timing as is referred to as a previous field F _-1 .

There is a time interval of two fields between the rear field F ₁ input from the input terminal 30 and the front field F _-1 output from the field memory 32. These are the motion vector detecting devices described in the above embodiment. 33, and processed as described above to detect the motion vector MV for each block. This motion vector MV is supplied to the motion compensation processing circuit 34.

In the motion compensation processing circuit 34, based on the motion vector MV between these two fields, the inter-field interpolation data (scanning line) of this current field F ₀ from the current field F ₀ and the previous field F _-1. Is formed. The motion vector between 1 fields is calculated by, for example, halving the motion vector between 2 fields. Therefore, in order to obtain the inter-field interpolated signal, the motion compensation processing circuit 34 halves the motion vector MV obtained by the motion vector detection device 33 to generate a motion vector for one field. Interfield interpolation data is generated by moving the data of the previous field F ₋₁ according to the motion vector of

In the non-interlaced signal output circuit 35, the motion compensation processing circuit 3 is provided between the scanning lines of the current field F _0.
The inter-field interpolation data obtained in 4 is inserted to generate a scanning line-interpolated image signal, that is, a non-interlaced image signal, which is output from the output terminal 36.

As described above, in this embodiment of the scanning line interpolating apparatus, since the interpolation signal is generated by the motion vector which suppresses the false detection due to the influence of the noise contained in the image signal, the motion compensation processing is performed. A non-interlaced image signal with less peculiar isolated point deterioration can be obtained.

[0094]

As described above, according to the present invention,
In the block matching process, if there are multiple reference blocks within the motion vector search range that have the smallest cumulative difference with the target block, the average vector (that is, the field vector) of the entire screen is obtained, and this field vector is set to this field vector. Since the motion vector with the shortest distance is selected, it is possible to prevent the phenomenon that a motion vector that has nothing to do with the motion of the subject is detected. Furthermore, due to the influence of noise contained in the image signal, it should be detected. Even if the difference cumulative value of the reference block indicating the motion vector does not take the minimum value within the motion vector detection range, this can be selected as the motion vector, and the motion vector can be detected with high accuracy.

Further, when the object moves at a constant speed on the screen, the constant speed is determined and the field vector is applied as the motion vector of the entire screen.
It is possible to detect a motion vector with high accuracy by eliminating false detection of a motion vector like an isolated point.

[Brief description of drawings]

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

FIG. 2 is a diagram for explaining a block matching method.

FIG. 3 is a diagram showing priorities of conventional motion vectors.

FIG. 4 is a diagram for explaining a problem of a conventional motion vector detection device.

FIG. 5 is a diagram in which the priorities of conventional motion vectors are expanded.

FIG. 6 is a block diagram showing a specific example of a field vector detection unit in FIG.

7 is a flowchart showing an operation of the field vector detection unit shown in FIG.

FIG. 8 is a block configuration diagram showing a second embodiment of the motion vector detection device according to the present invention.

FIG. 9 is a block configuration diagram showing a modification of the third embodiment shown in FIG.

FIG. 10 is a block diagram showing a third embodiment of the motion vector detection device according to the present invention.

11 is a block diagram showing a specific example of the field vector detection / constant velocity determination unit in FIG.

FIG. 12 is a flowchart showing the operation of the field vector detection / constant velocity determination unit shown in FIG.

FIG. 13 is a block diagram showing a specific example of a field vector detection / constant velocity determination unit in the fourth embodiment of the motion vector detection device according to the present invention.

FIG. 14 is a flowchart showing an operation example of the field vector detection / constant velocity determination unit shown in FIG.

15 is a diagram showing the concept of screen division and the method of forming field vectors in the field vector detection / constant velocity determination unit shown in FIG.

16 is a flowchart showing another operation example of the field vector detection / constant velocity determination unit shown in FIG.

FIG. 17 is a diagram showing a concept of the same directivity in the operation shown in FIG. 16.

FIG. 18 is a diagram showing an example of movement of a screen.

FIG. 19 is a block diagram showing an embodiment of a scanning line interpolating device using the motion vector detecting device according to the present invention.

[Explanation of symbols]

1 Target field input terminal 2 Reference field data input terminal 3 Cumulative difference calculation section 4 Difference comparison section 5 Field vector detection section 6 Field vector valid / invalid judgment section 7 Motion vector output terminal 8 Motion vector input terminal 9 Target block input Input terminal 10 Horizontal / vertical edge detection unit 11, 11 _{1 to} 11 _m AND gate 12, 12 _{1 to} 12 _m Cumulative addition unit 13, 13 _{1 to} 13 _m Field vector calculation unit 14 Field vector smoothing unit 15 Edged block counter 16 Edge existence block number threshold control unit 17 Field vector output terminal 18 Field vector valid / invalid flag output terminal 19 Difference allowable value determination unit 20, 21 Line memory 22 Maximum / minimum difference calculation unit 23 Field vector detection / Constant velocity determination Part 24 Changeover switch 25 Constant velocity motion determination unit 26 AND gate 27 Screen division unit 28 OR circuit 29 Determination unit 30 Interlaced image signal input terminals 31 and 32 Field memory 33 Motion vector detection device 34 Motion compensation processing circuit 35 Non-interlaced signal output circuit 36 Non-interlaced image Signal output terminal

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Mitsuo Nakajima             292 Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa             Ceremony Hitachi Digital Media Development Book             Department (72) Inventor Yasutaka Tsuru             292 Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa             Ceremony Hitachi Digital Media Development Book             Department (72) Inventor Yoshiaki Mizuhashi             292 Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa             Ceremony Hitachi Digital Media Development Book             Department (72) Inventor Kazuo Ishikura             5-20-1 Kamimizuhonmachi, Kodaira-shi, Tokyo Stock             Ceremony Hitachi System LSI Development Center             Within (72) Inventor Takaaki Matono             292 Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa             Ceremony Hitachi Digital Media System             Within the business division F-term (reference) 5C059 KK19 NN10 NN11 NN21 NN26                       NN28 NN38

Claims (5)

[Claims] 1. A motion vector in which a position vector of a reference block having the highest correlation with the target block among reference blocks within a motion vector detection range two fields before the target block is used as the motion vector of the target block. In the detection apparatus, there is provided a first means for generating a field vector representing an average motion of the entire screen for each field, and when there is only one reference block having the highest correlation with the target block, the reference block When the position vector of is the motion vector of the target block, and there are a plurality of reference blocks having the highest correlation with the target block, among the plurality of reference blocks,
Second means for selecting, as a motion vector, a position vector of the reference block that is closer to the field vector one field before with respect to the target block generated by the first means. Motion vector detection device. 2. The reference block according to claim 1, wherein the plurality of reference blocks that have the highest correlation with the target block are reference blocks whose correlation values are within a predetermined value range that can be set arbitrarily. Motion vector detection device. 3. The average vector of motion vectors obtained by the second means according to claim 1, wherein the first means divides the screen into a plurality of areas having a block as a constituent unit. And when the average vectors of all the regions are equal, or when the average vectors of a predetermined number or more of the regions are equal, the same average vector is set as the field vector. Vector detector. 4. The average vector of motion vectors obtained by the second means according to claim 1, wherein the first means divides the screen into a plurality of areas having a block as a unit. When there are a number of average vectors that are close to each other within a predetermined range that can be arbitrarily set and are equal to or larger than an arbitrary value that can be set arbitrarily, a vector obtained by averaging the average vectors that are close to each other is set in the field. A motion vector detection device characterized by being a vector. 5. The number of target blocks of the motion vector according to claim 1, wherein a distance from the field vector is within a first predetermined range that can be set arbitrarily, When the number is equal to or more than a predetermined value that can be set arbitrarily, it is determined that the entire screen is moving at a uniform speed, and a third means for setting the field vector as a motion vector of a target block of the entire screen is provided. A motion vector detection device characterized by the above.