JP2001061152A - Motion detection method and motion detector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To decrease the problems of a plurality of motions by detecting a large motion vector in the unit of one pixel or a small block, without increasing the calculation amount and preventing deterioration in accuracy of motion detection, so as to reduce errors due to repetitive pattern. SOLUTION: A representative point matching processing section 102 conducts matching processing between an image of a current frame and an image consisting of representative points of a preceding frame. The value of each pixel of a block of the same position of the current frame is subtracted from representative point data of a block of the preceding frame, absolute values of a subtraction output are integrated by one block and integral values of each block are integrated by one frame. The integrated values are fed to an evaluation value table generating section 103. The evaluation value table generating section 103 stores the integrated value obtained at each position within a search range to a memory, where an evaluation table is generated. An object vector extract section 104 extracts one or a plurality of object vectors, by referring to the evaluation value table. A motion vector detection section 105 detects a motion vector in the unit of one pixel through matching processing, using the object vector and outputs the detected motion vector to an output terminal 107.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for detecting a motion of an image signal.

[0002]

2. Description of the Related Art Conventionally, in the field of moving image processing, motion, that is, the moving direction and the size (or speed) of an object in an image which is temporally different are used. For example, motion is used for motion-compensated inter-frame coding in high-efficiency coding of an image, parameter control by motion in a television noise reduction device using an inter-frame time domain filter, and the like. As a motion detection method for obtaining a motion, a block matching method is known.

FIG. 9 shows an outline of the block matching method.
This will be described with reference to FIG. First, one screen is divided into blocks each having an appropriate number of pixels (for simplicity, a size of 2 × 2 pixels). Subsequently, in order to search for the image data thus blocked and an area in which the image data has moved, a search range in which image data of a screen different in time (for example, a screen one frame before) is blocked. Is set, and an evaluation value is calculated between the reference block of the current frame and the candidate block of the previous frame within the search range.

In FIG. 9, a region indicated by a broken line is a search range. In the example of FIG. 9, a search range of ± 4 in the horizontal (x) direction and ± 4 in the vertical (y) direction is set. An evaluation value is calculated between a candidate block moved by one pixel in each of the x direction and the y direction within the search range and a fixed reference block. As the evaluation value, a value obtained by totaling the absolute values of the differences between the values of pixels at the same position between the reference block and the candidate block in block units is used. FIG.
0 indicates a table of the evaluation values Ex, y thus obtained. Evaluation values corresponding to some examples of the candidate blocks are shown in FIG. In all, (9 × 9) evaluation values are obtained.

[0005] Among these evaluation values, the smallest one is obtained.
The position of the candidate block that produces the minimum evaluation value is the motion vector MV. For example, if E0,3 is the minimum value,
The motion vector is MV0,3. Thus, the motion of the image in block units can be detected.

[0006]

In the block matching method described above, it is necessary to set a relatively large search range in order to detect a large motion vector. However, there is a problem that the amount of calculation for calculating the evaluation value indicating the degree of matching increases, and that the process of determining a motion vector from the evaluation value increases.

In the above-described block matching method, a motion is detected in block units. Therefore, when a plurality of motions are mixed in a block, it is difficult to obtain an accurate motion. For example, when a block includes a moving part and a stationary part, even if the movement of the moving part can be detected, the movement is not exactly the movement of the block. When you make a block bigger,
In addition to an increase in the amount of calculation, a problem of a plurality of motions in a block is likely to occur.

In order to reduce such a problem, the size of the block may be reduced so that a plurality of motions are not included in the block. However, when the block size is reduced, the matching determination area becomes smaller, and thus a problem arises in that the accuracy of motion detection is reduced. That is, when performing the matching, a candidate block similar to the reference block easily appears due to no movement. Also,
When the block size is reduced, the pattern of the same image is likely to be repeated. For example, when a character telop moves in a horizontal or vertical direction, the effect of a repetitive pattern tends to appear. In the case of a kanji character pattern, the same character often becomes the same pattern when divided into small parts. Therefore, a plurality of candidate blocks having the same pattern as the reference block appear in the search range,
It is difficult to detect movement accurately.

As described above, when the block size is increased, a plurality of motions are likely to occur, and the amount of calculation for obtaining a motion vector increases. As described above, increasing the block size and decreasing the block size are contradictory, and it has been considered that it is difficult to solve all the above-mentioned problems.

For example, an image including a moving character telop is
This is a general image (scene) in television broadcasting. Such scenes are a typical example of having multiple motion and repetitive patterns simultaneously. When a character telop is included in a block, a plurality of movements are often mixed in the block. Also, character patterns often consist of geometric elements, often producing repetitive patterns. When such an image is targeted, it is difficult to detect a motion accurately by the conventional block matching method. Furthermore, a moving character telop has a high degree of visual attention, and the deterioration of an image caused by a motion detection error occurring here becomes very conspicuous.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to solve the above-mentioned problem, without increasing the amount of calculation.
It can detect large motion vectors,
An object of the present invention is to provide a motion detection method and a motion detection device capable of detecting a motion with high accuracy even when a repetitive pattern exists.

[0012]

According to a first aspect of the present invention, there is provided a motion detection method for detecting a motion in an image signal, comprising the steps of: For each relatively large divided block, an integrated value table is generated by a matching method. Using the integrated value table, one or a plurality of candidate vectors are generated for each whole screen or each relatively large block obtained by dividing one screen into a plurality. And a step (b) of performing matching on only the candidate vector and detecting a motion vector for each pixel or each relatively small block.

According to a fourth aspect of the present invention, there is provided a motion detecting apparatus for detecting a motion in an image signal, wherein an integrated value table is generated by a matching method for one entire screen or for each relatively large block obtained by dividing one screen into a plurality. Means for extracting one or a plurality of candidate vectors for each relatively large block obtained by dividing the entire screen or one screen by using the integrated value table; Alternatively, there is provided a motion detecting device comprising: means for detecting a motion vector for each relatively small block.

According to the present invention, in the preprocessing, a matching process is performed on one entire screen to extract one or a plurality of candidate vectors. By detecting a motion vector in units of one pixel or a small block with respect to the extracted candidate vector, a large motion vector can be detected without increasing the amount of calculation, and a problem of a plurality of motions in a block hardly occurs. In addition, the motion detection accuracy can be reduced and the effect of the repetition pattern can be reduced.

[0015]

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, one embodiment of the present invention will be described.
This will be described with reference to the drawings. FIG. 1 shows an overall configuration of an embodiment of the present invention. A digital video signal is supplied to an input terminal denoted by reference numeral 101. The digital video signal is obtained, for example, by sampling a luminance signal at a predetermined frequency and converting each sample (pixel) into 8 bits. The digital video signal is supplied to the representative point matching processing unit 102.

The representative point matching processing unit 102 converts the image of the previous frame into an image composed of representative points by thinning out the image of the previous frame, and performs block conversion between the image of the current frame and the image composed of the representative points of the previous frame. Similar to the matching, a matching process is performed. The representative points are shown in FIG.
As shown in FIG. 5, data representing each block when an image of one screen, for example, one frame is divided into a plurality of (m pixels × n lines) blocks. As representative point data,
The value of the pixel at the center position of the block, the average value of the values of the pixels in the block, the intermediate value of the pixels in the block, and the like are used.

The representative point matching processing unit 102 includes: a reference frame image of the current frame within the set search range;
An inter-frame difference is calculated with respect to a previous frame candidate frame image composed of representative point data. That is, the value of each of the m × n pixels of the block at the same position in the current frame is subtracted from the representative point data of a certain block in the previous frame, the absolute value of the subtracted output is integrated in one block, and Are integrated in one frame. This integrated value is supplied to the evaluation value table forming unit 103.
The evaluation value table forming unit 103 stores the integrated value obtained at each position within the search range in a memory, and forms an evaluation value table on the memory.

The candidate vector extracting unit 104 extracts one or a plurality of candidate vectors with reference to the evaluation value table formed by the evaluation value table forming unit 103. The extracted candidate vector is supplied to the motion vector detection unit 105. The motion vector detection unit 105 includes a delay circuit 106
The input video data is supplied via. Delay circuit 10
6 delays the input video data by the time required to delay the candidate vector. When the input video data is to be read from the memory, the read video data may be provided to the motion vector detection unit 105, and thus there is no need to provide the delay circuit 106.

The motion vector detecting section 105 detects a motion vector for each pixel by a matching process using a candidate vector, and outputs the detected motion vector to an output terminal 107. Here, detection of a motion vector per pixel will be described. Even in detection of a motion vector in units of one pixel, a block is formed in order to obtain an evaluation value. As shown in FIG. 3, for example, 3
A × 3 block B1 is configured. The block B1 is, for example, a reference block of the current frame. The candidate block of the previous frame is similarly configured. Then, an evaluation value table within the search range is formed by block matching. A motion vector is detected for the pixel P1 corresponding to the minimum evaluation value in the evaluation value table. Next, a reference block is similarly formed for the adjacent pixel P2. Then, similarly, an evaluation value table is formed by block matching, and a motion vector related to the pixel P2 is detected based on the evaluation value table. Thus, a motion vector is detected for each pixel.

In general, the processing for detecting a motion vector for each pixel greatly increases the amount of calculation compared to the processing for detecting a motion vector for each block, and complicates the processing. However, in this embodiment, since the motion vector is detected only for one or a plurality of candidate vectors extracted by the candidate vector extracting unit 104, it is possible to prevent an increase in the amount of calculation and a complicated process.

FIG. 4 is a diagram for explaining the processing of the candidate vector extraction unit 104. Evaluation value table forming unit 1
03, the evaluation value table formed in x and y
An evaluation value exists as the z coordinate of the search range defined by the coordinates. Figure 4A conceptually illustrates an example of the x-direction distribution of the evaluation value table, for example, a horizontal distribution through the y = y _1. FIG. 4B shows y in the evaluation value table.
Direction of an example conceptual shows the distribution, for example, a vertical distribution through x = x _1. From the evaluation value tables of FIGS. 4A and 4B, the origin on the coordinates and (x = x ₁ , y =
It can be seen that a minimum value exists at each of the points y ₁ ). Such an evaluation value table is obtained for an image in which an object that moves obliquely with respect to the background (still image) exists.

The candidate vector extracting unit 104 searches for a local minimum value from such an evaluation value table and extracts a candidate vector corresponding to the local minimum value. In the example of FIG. 4, (x = 0, y =
0) and a motion vector (x = x ₁ , y = y ₁ ) are extracted as candidate vectors. Here, for simplicity, an example of the evaluation value table in which two minimum values clearly exist has been described. However, in reality, there are more local minimum values, the sizes of the local minimum values are different, and the shapes of the curves drawn by the local minimum values and the peripheral evaluation values are different. In such a case, the candidate vector extracting unit 104 reduces the number of candidates so as to extract an appropriate candidate vector. That is, the minimum value is compared with the threshold value, and the minimum value larger than the threshold value is processed as a candidate vector, and the steepness of the curve drawn by the minimum value and the peripheral evaluation value is detected. Then, the detected steepness is compared with the threshold value, and a process in which the steepness is small is not set as a candidate vector.

The motion vector detecting section 105 determines only the candidate vector extracted in this manner as the best motion vector from the candidate vector for each pixel. As in the example described above, when two candidate vectors are given, two evaluation values are formed for the target pixel by matching processing as shown in FIG. One evaluation value E0,0 corresponds to a motion vector of (x = 0, y = 0), and the other evaluation value Ex1, y1 is (x =
x ₁ , y = y ₁ ).

An example of a method for determining the best motion vector based on the evaluation value will be described. The smallest evaluation value among the evaluation values obtained corresponding to the candidate vectors and the one having a sufficiently small magnitude are selected as the best motion vectors. For example, when the evaluation value E0,0 is the minimum and the evaluation value E0,0 is a sufficiently small value, (x = 0, y = 0) is selected as the motion vector MV. On the other hand, when the evaluation value Ex1, y1 is the minimum and the evaluation value Ex1, y1 is a sufficiently small value, (x = x ₁ , y = y ₁ ) as the motion vector MV of the target pixel.
Is selected. If a pixel does not satisfy these conditions, the motion vector is undefined. Other methods for determining the best motion vector from the candidate vectors are possible.

In the above description, the representative point matching is used, but the evaluation value table for the entire screen may be formed by all pixel matching. Also, a method of dividing one screen into relatively large blocks such as four divisions, twelve divisions, sixteen divisions, etc., instead of the entire screen, and creating an evaluation value table for each block can be adopted.

The motion vector detecting section 105 performs a matching process for each pixel with respect to the candidate vector. As the one that can be used as the motion vector detecting unit 105, one related to the proposal of the present applicant will be described below. In FIG. 6, reference numeral 1 denotes the overall configuration of the motion vector detecting device. In the following description, an example in which a motion vector is detected in block units will be described. This block is a relatively small block (2 × 2 pixels or the like). The motion vector detection device 1 can also detect a motion vector in pixel units, not in small block units.

The motion detecting apparatus 1 cuts out a block area (reference block) of image data of a frame (hereinafter referred to as (n-2) frame) located two frames before the current frame, and extracts the current frame (hereinafter referred to as n). A block area (candidate block) for detecting a motion from image data in a frame (noted as a frame), and a motion is detected from image data of a frame (hereinafter referred to as an (n-1) frame) located one frame before the current frame. Block area (candidate block) for detection, frame located three frames before the current frame (hereinafter referred to as (n-3) frame)
, And a block area (candidate block) for detecting a motion is extracted from the image data of a frame located four frames before the current frame (hereinafter referred to as (n-4) frame). Then, by moving the candidate block within the predetermined search range, a motion vector is detected between the reference block and the candidate block.

Further, the motion detecting device 1 will be described in more detail. In FIG. 6, reference numeral 2 denotes an input terminal, and image data Sn of the current frame is supplied from the input terminal 2.
The image data is, for example, a luminance signal in a component signal of a color video signal. The memory 3 and the delay unit 4 are connected to the input terminal 2. Delay units 6, 8, and 10 are connected in series to delay unit 4. Each of the delay units 4, 6, 8, and 10 stores image data for one frame and outputs image data delayed by one frame period. The image data delayed by the delay units 4, 6, 8, and 10 is supplied to the memories 5, 7, 9, and 11.
The output Sn-1 of the delay unit 4 is image data of (n-1) frame, the output Sn-2 of the delay unit 6 is image data of (n-2) frame, and the output Sn-3 of the delay unit 8 is. Is (n-
3) The image data of the frame, the output S of the delay unit 10
n-4 is the image data of (n-4).

The operation of the memories 3, 5, 7, 9, and 11 is controlled by the memory control unit 12, and blocks of a predetermined size and a predetermined position are cut out from the video signal of each frame. From the memory control unit 12, the block address information of the (n-2) th frame whose motion vector information is to be calculated is sent to the memory 7, and as a result, a predetermined block of the image data Sn-2 is cut out from the memory 7. The block of the extracted image data Sn-2 is supplied to the motion detection unit 13.

The candidate vector extracted by the candidate vector extracting unit 104 is supplied to the memory control unit 12. As described above, one of the candidate vectors is selected as the best motion vector, except for the case where the motion vector is indeterminate. Therefore, the address for moving the candidate block output from the memory control unit 12 is selected. The information corresponds to the candidate vector.

Further, address information of the (n-3) frame for moving the candidate block within a predetermined search range is sent from the memory control section 12 to the memory 9, and the memory 9 stores the address information at the given address. The corresponding (n-
3) The data of the candidate block of the frame is
Output to The movement amount of the (n-3) frame candidate block within the search range is represented by (x, y).

The address information of the (n-1) frame for moving the candidate block within the predetermined search range is sent from the memory control unit 12 to the memory 5, and the memory 5 corresponds to the given address. The data of the (n−1) frame candidate block is output to the motion detection unit 13. The moving amount of the (n-1) frame candidate block within the search range is ((-1) × x, (-1) × y). This corresponds to (n−2) frames,
3) While the frame is one frame past, (n)
-1) The frame is one frame in the future.

From the memory control unit 12 to the memory 11
Sends address information of an (n-4) frame for moving a candidate block within a predetermined search range,
The memory 11 corresponds to the given address (n-
4) The data of the candidate block of the frame is
Output to The moving amount of the (n-4) frame candidate block within the search range is (2 × x, 2 × y).
This is because the (n-3) frame is one frame before the (n-2) frame, whereas the (n-4) frame is two frames before.

The address information of n frames for moving the candidate block within a predetermined search range is sent from the memory control unit 12 to the memory 3.
Outputs the data of the nth frame candidate block corresponding to the given address to the motion detection unit 13. The amount of movement that the n-frame candidate block has within the search range is
((−2) × x, (−2) × y). This is (n
This is because the (n-4) frame is two frames ahead of the (-2) frame.

FIG. 8 schematically shows the positional relationship between a reference block and a candidate block. For simplicity, the movement of the candidate block within the search range is limited to the x direction. As shown in FIG. 8A, assuming that the movement amount of the (n-3) frame candidate block is (a, 0) with respect to the (n-2) frame reference block, the (n-4) frame candidate block Is (2a, 0), the moving amount of the (n-1) th frame candidate block is (-a, 0), and the moving amount of the nth frame candidate block is (-2a, 0). Is done. Further, as shown in FIG. 8B, assuming that the movement amount of the candidate block of the (n-3) frame is (-b, 0) with respect to the reference block of the (n-2) frame, the (n-4) frame Is set to (−2b, 0),
(N-1) The moving amount of the frame candidate block is (b,
0), and the moving amount of the n-frame candidate block is (2)
b, 0). That is, the motion is regarded as a constant speed motion, the moving amount of the candidate block is set in proportion to the time interval between frames, and the moving direction (polarity) is set in accordance with the time order.

The motion detector 13 detects a motion vector MV by a block matching method. This motion vector MV represents the motion of the reference block belonging to the (n-2) frame. The motion detection unit 13 integrates the absolute values of the differences between the pixel values at the same position between the n-th frame candidate block and the (n−2) -th frame reference block, and calculates the sum of absolute values. Similarly, the sum of the absolute value of the difference between the (n-1) frame candidate block and the (n-2) frame reference block, the (n-3) frame candidate block and the (n-2) frame reference Sum of absolute values of differences between blocks, (n-4) frame candidate blocks and (n-2)
The absolute sum of the difference between the frame and the reference block is calculated. As described above, the positional offset (movement amount and moving direction) of these candidate blocks with respect to the reference block depends on the time interval between the frame to which the reference block belongs and the frame to which the candidate block belongs, and the temporal context of the two frames. Is set by

As an example, the sum of the absolute values of the four differences is obtained in accordance with the movement corresponding to one candidate vector within the search range. The sum of the absolute values of the differences is integrated to obtain an evaluation value corresponding to the movement. The motion vector in the reference block is determined by detecting the positional offset of the candidate block that produces the minimum value among the evaluation values.

Evaluation value E _{(x, y) at} moving amount _{(x, y)}
Is calculated by the following equation. In this equation,
Each pixel level of the n frame is _{Zn (i, j)} , each pixel level of the (n-1) frame is _{Zn-1 (i, j)} , and each pixel level of the (n-2) frame is _{Zn- 2 (i, j)} , each pixel level of the (n-3) frame is _expressed as Zn _{-3 (i, j)} , and each pixel level of the (n-4) frame is expressed as _{Zn-4 (i, j).} . Also, if the block size is M
Pixels × N lines.

[0039]

(Equation 1)

In the above description, when calculating the sum of absolute differences between two frames, the sum of absolute differences between the (n-2) frame and another frame is calculated. The evaluation value may be obtained by calculating the absolute value sum of the difference between the two frames and integrating the absolute value sum of the difference. The formula for calculating the evaluation value in this case is shown below.

[0041]

(Equation 2)

An example of the motion detector 13 will be described with reference to FIG. (N-2) Frame image data Sn-2
Is supplied to the register 31 from the memory 7, and four identical data are output from the register 31 in parallel. The four pixel data from the register 31 are respectively inverted and added to an adder 36,
37, 38, and 39. Adders 36, 37, 3
8 and 39 have a function as a subtractor, and are hereinafter referred to as subtracters.

An output signal (pixel data) from the memory 3 in which the n-frame image data Sn is stored is supplied to the subtractor 36 via the register 32. An output signal (pixel data) from the memory 5 in which the (n-1) frame image data Sn-1 is stored is supplied to the subtractor 37 via the register 33. Similarly, an output signal (pixel data) from the memory 9 storing (n-3) frame image data Sn-3, and a memory storing (n-4) frame image data Sn-4. The output signal (pixel data) from 11 is supplied to subtracters 38 and 39 via registers 34 and 35, respectively.

The subtractor 36 calculates the difference between the pixel data of the reference block of the (n-2) frame and the pixel data of the corresponding candidate block of the n frame. This difference value is converted into an absolute value by the absolute value conversion circuit 40. An accumulator including an adder 44 and a register 48 is connected to the absolute value conversion circuit 40. When the calculation of the difference values for the M × N pixels in the block is completed, an integrated value (sum of absolute values) is generated in the register 48. This sum of absolute values is supplied to the evaluation value memory 52.

In the subtracter 37, the pixel data of the reference block of the (n-2) frame and the corresponding (n-1)
The difference from the pixel data of the frame candidate block is calculated, the difference output is converted into an absolute value by the absolute value conversion circuit 41, and the absolute value of the difference is cumulatively added by the adder 45 and the register 49. The sum of absolute values is supplied to the evaluation value memory 52.

The absolute value of the difference between the pixel data of the reference block of the (n−2) frame and the corresponding pixel data of the candidate block of the (n−3) frame is calculated by the subtractor 38 and the absolute value conversion circuit 42. Generated. Adder 4
6 and the register 50 generate the absolute value sum, and the absolute value sum is supplied to the evaluation value memory 52. Further, the pixel data of the reference block of the (n−2) frame and the corresponding pixel data of the candidate block of the (n−4) frame are calculated by the subtractor 39, the absolute value conversion circuit 43, the adder 47, and the register 51. Is generated. This sum of absolute values is supplied to the evaluation value memory 52.

The evaluation value memory 52 is provided with an integrating circuit for summing the absolute value sums of the differences generated in the above-described four combinations of two frames. Is stored in the evaluation value memory 52 as the evaluation value E (x, y). The evaluation value memory 52 stores the evaluation value at the origin of the search range and at a position offset by one pixel in the x and / or y directions from the origin. Thereby, an evaluation value table is created. The write operation and the read operation of the evaluation value memory 52 are controlled by the memory control unit 53. When the evaluation value corresponding to the movement amount (x, y) is stored in the evaluation value memory 52, the contents of the registers 48, 49, 50, and 51 are cleared.

The evaluation value is read according to the address specified by the memory control unit 53. The read evaluation value is supplied to the comparison circuit 54 and the register 55. The output of the register 55 is supplied to the comparison circuit 54. The read address generated by the memory control unit 53 is supplied to the register 56. Register 55
And 56, the output of the comparison circuit 54 is supplied as a write (input) enable.

The comparison circuit 54 outputs the smaller value of the two inputs. That is, when the evaluation value read from the evaluation value memory 52 is smaller than the previous minimum evaluation value stored in the register 55, a comparison output as a write enable to the registers 55 and 56 is generated. Thus, the new minimum evaluation value is fetched into the register 55, and the address of the new evaluation value is fetched into the register 56. In the opposite case, no output as a write enable is generated from the comparison circuit 54, and the contents of the registers 55 and 56 are not updated.

When the minimum value of all the evaluation values corresponding to the candidate vectors is detected, the address of the minimum evaluation value is output from the register 56 as the motion vector MV to the output terminal 1.
It is taken out to 4.

In the motion detecting device 1 described above, the sum of the absolute values of the differences is calculated as the evaluation value. However, the sum of the squares of the differences, the number of the absolute values of the differences that are equal to or larger than the threshold or equal to or smaller than the threshold are calculated. , The extreme value of the difference, etc. can be used. Also,
Five temporally consecutive frames are used, and the movement of the central frame is detected. However, the number of frames used for motion detection is not limited to five frames and may be three or more frames. The interval between two frames to be matched is not limited to one frame, and matching between two or more frames may be performed.

The present invention is not limited to the hardware configuration, and a part or all of the processing may be realized by software processing.

[0053]

According to the present invention, without increasing the amount of calculation,
A large motion vector can be detected in a unit of one pixel or a small block, a problem of a plurality of motions in a block is hardly generated, and a motion detection accuracy is reduced and it is hard to be affected by a repetitive pattern.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an overall configuration of an embodiment of the present invention.

FIG. 2 is a schematic diagram for explaining representative point matching in one embodiment of the present invention.

FIG. 3 is a schematic diagram illustrating a motion vector detection process in units of one pixel according to an embodiment of the present invention.

FIG. 4 is a schematic diagram illustrating an example of a distribution of evaluation values according to an embodiment of the present invention.

FIG. 5 is a schematic diagram illustrating an example of an evaluation value table according to an embodiment of the present invention.

FIG. 6 is a block diagram illustrating a configuration of an example of a motion vector detection unit according to an embodiment of the present invention.

FIG. 7 is a block diagram illustrating a configuration of a motion detection unit in an example of the motion vector detection unit.

FIG. 8 is a schematic diagram used to describe an example of a motion vector detection unit.

FIG. 9 is a schematic diagram used for describing a motion detection method by block matching.

FIG. 10 is a schematic diagram used for describing a motion detection method by block matching.

[Explanation of symbols]

103: evaluation value table forming unit; 104: candidate vector detecting unit; 105: motion vector detecting unit

Continuation of front page (72) Inventor Takeshi Miyai 6-35 Kita Shinagawa, Shinagawa-ku, Tokyo Sony Corporation F-term (reference) 5C059 KK10 KK19 LC01 MA05 NN03 NN21 NN27 NN28 NN29 NN31 TA64 TB08 TC12 TD02 TD05 TD15 UA37 5L096 AA06 CA04 DA02 FA32 GA08 GA17 GA19 GA26 HA04 JA03 JA11 LA14

Claims (6)

[Claims] 1. A motion detection method for detecting a motion in an image signal, comprising: (a) generating an integrated value table by a matching method for each whole screen or each relatively large block obtained by dividing one screen into a plurality of blocks; (B) extracting one or more candidate vectors for each relatively large block obtained by dividing the entire screen or one screen into a plurality using the integrated value table; And detecting a motion vector for each pixel or each relatively small block. 2. The method according to claim 1, wherein, in the step (a), when there are a plurality of integrated values whose values are small in the integrated value table, each integrated value and its neighboring A motion detection method comprising: selecting an integrated value from the plurality of integrated values based on a steep change of the integrated value to the integrated value; and extracting the candidate vector based on the selected integrated value. 3. The method according to claim 1, wherein the step (b) uses at least three temporally different frame images,
The first pixel or the first block is fixed, the second pixel or the second block is moved within a predetermined range, and the first pixel or the first block and the second pixel or the second Generating an evaluation value table indicating the degree of matching between the blocks; and detecting a motion vector based on the evaluation value table. In the step of generating the evaluation value table, the second The amount of movement of the pixel or the second block is determined by the first
A motion detection method characterized by making a time interval between a frame to which the pixel or the first block belongs and a frame to which the second pixel or the second block belongs. 4. A motion detection device for detecting a motion in an image signal, wherein an integrated value table is generated by a matching method for one whole screen or for each relatively large block obtained by dividing one screen into a plurality of blocks. Means for extracting one or a plurality of candidate vectors for each of the entire screen or each relatively large block obtained by dividing the screen into a plurality of pieces, and performing matching only on the candidate vectors,
Means for detecting a motion vector for each pixel or each relatively small block. 5. The method according to claim 4, wherein the processing of extracting the candidate vector is performed when a plurality of integrated values having small values exist in the integrated value table.
Selecting an integrated value from among the plurality of integrated values based on each integrated value and a steep change from each integrated value to a peripheral integrated value, and extracting the candidate vector based on the selected integrated value A motion detection device, characterized in that: 6. The method according to claim 4, wherein the means for detecting a motion vector for each pixel or each relatively small block uses at least three temporally different frame images, and uses a first pixel or a first block. Is fixed, and the second pixel or the second block is moved within a predetermined range to indicate the degree of matching between the first pixel or the first block and the second pixel or the second block. Means for generating an evaluation value table; and means for detecting a motion vector based on the evaluation value table. In the means for generating the evaluation value table, the second
The movement amount of the pixel or the second block is determined by comparing the frame to which the first pixel or the first block belongs to the second pixel or the second block.
A motion detection device that makes the ratio proportional to a time interval between frames to which the pixel or the second block belongs.