JP2000201328A - Method and circuit for detecting motion vector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To highly accurately and efficiently detect a motion vector suitable for motion correcting processing by performing search processing to a mini block for which a block is horizontally/vertically fractionized, based on this motion vector. SOLUTION: Among detected reference vectors RMV, one of least predictive error is set as a representative vector TMV (S2). Concerning a block having the representative vector TMV of predictive error more than a prescribed value, a search area determined from a search setting signal MOD is searched again by block matching processing and a motion vector MV1 is generated (S3). A special vector is detected from the correlation of the motion vector with a block adjacent to the target block and replaced with the motion vector of high correlation and an erroneous detection vector is removed. Then, an accurate motion vector MV is generated (S4). A motion vector is generated for each mini block fractionized horizontally/vertically and in this output, a motion vector PV for the unit of pixel suitable for motion correcting processing is obtained (S5).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and a circuit for detecting a motion vector per unit time of an image, and is particularly suitable for signal processing such as frame number conversion and interlaced to sequential scan conversion using a motion correction process. The present invention relates to a method and a circuit for detecting a simple motion vector.

[0002]

2. Description of the Related Art In recent years, television receivers and PCs as information home appliances have required various format conversion functions for image signals, such as frame number conversion, interlaced to sequential scan conversion, and pixel number conversion. Has become. In order to perform these format conversion functions with high image quality, interest in signal processing using motion correction processing is increasing.

In the field of high-efficiency coding of images, predictive coding for motion compensation has become mainstream. This encodes an error component between a signal of a predicted frame generated by moving a position of an image by a motion vector detected in a block unit (for example, 8 pixels × 8 lines) and a signal of a current frame. Therefore, the motion vector may be a block unit, and even if the detection accuracy is low, the coding efficiency is slightly lowered, but the image quality is not significantly degraded.

[0004] On the other hand, in signal processing for improving image quality using the motion correction processing, an inaccurate motion vector causes an unsightly image quality degradation such that a part of an image is replaced by an unrelated image. Therefore, high accuracy is required for detecting a motion vector. In addition, since the signal processing is performed for each pixel, a motion vector for each pixel is required.

Many methods have been devised so far as a method of detecting a motion vector suitable for a motion correction process. For example, JP-A-7-170496 discloses a method for efficiently searching for a motion vector. However, these prior arts have a problem in the accuracy of detecting a motion vector, which is a major obstacle in realizing signal processing using motion correction processing.

[0006]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned problems, and provides a method and a circuit for detecting a motion vector which efficiently and efficiently detects a motion vector suitable for a motion correction process. With the goal.

[0007]

In order to achieve the above object, the present invention employs the following technical means.

The detection of the motion vector is performed stepwise. First, the detection of the motion vector for each block is performed based on the block matching processing. , And finally a motion vector for each pixel is detected. Then, in each stage, a detection process that makes full use of the information of the motion vector that has been detected and the features of the motion of the image is performed.

Focusing on the property that the motion vector has a high correlation between the blocks, the detection of a motion vector in units of a block is based on the motion vector having the smallest prediction error component among the motion vectors already detected in the blocks adjacent to the target block. Set to the representative vector of the block. When the prediction error caused by the representative vector exceeds a predetermined value, a search area defined by the average speed and the direction calculated from the already detected motion vectors in the predetermined area is searched for by a block matching process. Find the vector.

Next, a motion vector having a low correlation with a motion vector of an adjacent block is detected as a singular vector, and correction is performed by replacing the motion vector of the adjacent block with a motion vector having a minimum prediction error.

With the above technical means, it is possible to detect a motion vector in block units with high accuracy and efficiency.

The detection of a motion vector for each pixel, which is the final target, is performed for a mini-block obtained by subdividing a block horizontally and vertically, when the prediction error is smaller than a set value, the motion vector of the target block, and In the mini-block, among the motion vectors of the target block and blocks adjacent thereto, the one with the smallest prediction error is assigned to the motion vector of each pixel in the mini-block. At this time, if the size of the mini-block is set to, for example, about 2 pixels × 2 lines, it is extremely likely that the pixels in the mini-block will have almost the same motion vector, and the motion vector for each pixel can be efficiently detected. Can be.

In detecting a motion vector in block units, a still block and a moving image block are determined based on a difference signal component between frames, and the detection of a motion vector is performed only for a moving image block.

Further, the motion vector information used in the motion compensation predictive coding is utilized, and a representative vector is set by using the converted motion vector information per unit time.

With these technical means, it becomes possible to detect a motion vector suitable for the motion correction processing with higher accuracy and more efficiently.

[0016]

(Embodiment 1) A first embodiment of the present invention will be described with reference to FIGS. FIG. 1 shows this block configuration of the motion vector detection circuit according to the first embodiment of the present invention. The delay unit 1, the representative vector setting unit 2, the re-search unit 3, the singular vector detection unit 4, the mini-block search unit 5,
It comprises a memory unit 6, a vector analysis unit 7, and a control unit 8.

One of the input image signals S1 (the scanning mode is interlaced scanning or sequential scanning, and the signal component can be a luminance signal or any combination of a luminance signal and a chrominance signal) is input to the delay unit 1 and is delayed by a predetermined unit time. Signal S2
Generate For example, when used in the frame number conversion process, the signal S2 is a signal delayed by one frame period, and when used in interlaced to sequential scan conversion, the signal S2 is a signal delayed by several frame periods.

The representative vector setting unit 2 performs the signal processing of the first and second steps in the flowchart of the motion vector detection shown in FIG. That is, it is determined whether the target block is a still / moving block. If the target block is a still block, 0 is added to the representative vector TMV. Set to TMV. The size of the block is
For example, there are 8 horizontal pixels × 8 vertical lines.

The re-search section 3 performs the signal processing of the third step in the flowchart of FIG. That is, for a block in which the prediction error of the representative vector TMV is equal to or larger than a predetermined value, the search area determined by the search setting signal MOD is re-searched by the block matching process to generate the motion vector MV1.

The singular vector detector 4 performs the signal processing of the fourth step in the flowchart of FIG. That is, a singular vector is detected from the correlation between the motion vector of the target block and a block adjacent thereto, and a correction process of replacing the singular vector with a motion vector having a high correlation is performed. ) Is removed. Then, a highly accurate motion vector MV is generated.

One of the motion vectors MV is stored in the memory unit 6 and outputs a detected reference vector RMV. Further, the vector analysis unit 7 extracts motion vector information MVIF such as an average speed and a direction of a motion vector in a predetermined area based on the output.

The mini-block search unit 5 performs the signal processing of the fifth step in the flowchart of FIG. That is, a motion vector is generated for each mini-block (the size is, for example, two horizontal pixels × two vertical lines) obtained by subdividing the block in the horizontal and vertical directions. Then, a motion vector PV in pixel units suitable for the motion correction process is obtained from this output.

The control unit 8 generates a search setting signal MOD from the motion vector information MVIF. Although not explicitly shown in the drawings, signals necessary for signal processing operation in each block are generated.

The description of the overall configuration is now complete.
The configuration and operation of each block will be described in detail.

First, the representative vector setting unit 2 will be described with reference to FIGS. FIG. 3 shows an example of the configuration of the representative vector setting unit 2, which includes a still and moving image block determining unit 9, a calculating unit 10, and a determining unit 11.

First, the still / moving block discriminating section 9 extracts a difference signal component between frames (for example, one to several frame periods) by subtraction processing of the signals S1 and S2. Then, the ratio of pixels whose difference components are smaller than the threshold value is, for example, 10
The blocks equal to or more than% are determined to be still blocks, and the following blocks are determined to be moving image blocks.

The arithmetic units 10-1 to 10-n calculate the following equation 1 with respect to the detected reference vector RMVi (i = 1,..., N).
Of the prediction error ERi shown in FIG. Then, the detected reference vectors RMV1 to RMVn and their prediction errors ER
1 to ERn.

[0028]

ERi = Σ | S1 (x, y) −S2 (x + Vxi, y + Vyi) | (1) (where S1 (x, y) is the signal of pixel (x, y) in the target block, S2 (x + Vxi, y + Vyi) uses pixel (x, y) as a reference vector RMVi (x-direction component V
xi, y-direction component Vyi) The signal of the point whose position has been moved, the symbol || indicates its absolute value, and Σ indicates the sum of the pixels in the target block. FIG. 4 shows an example of this detected reference vector. is there. A motion vector of a block adjacent to the target block is used as a reference vector. The motion vector detected in the current frame before the target block is used as a reference vector, and the motion vector detected in the previous frame is used as a reference vector thereafter (in the hatched area in the drawing).

Returning to FIG. 3, the determination unit 11 determines that the stationary block whose signal SMF is 0 is the representative vector TMV and the prediction error ER.
Outputs 0 to T. On the other hand, in the moving image block in which the signal SMF is 1, the smallest one of the prediction errors ER1 to ERn is detected, and the corresponding reference vector is represented by the representative vector TMV,
The prediction error is output to ERT.

Next, the re-search section 3 will be described with reference to FIGS. FIG. 5 shows an example of the configuration of the re-search section 3.
First search area setting unit 12, block matching processing unit 1
3, 16, second search area setting unit 15, setting units 14, 17
Consists of

The first search area setting section 12 controls a control signal SR for designating an area to be searched in the block matching process with the characteristics shown in FIG. 6 according to the search setting signal MOD from the control section.
1 is generated.

That is, when the MOD is 1 to 3 and the direction of the average motion vector is ang (ang = ave | Vy | / ave | Vx
│, ave│Vx│, ave│Vy│ are average horizontal speeds and average vertical speeds), and when the horizontal movement is １ ／ or less, the control signal SR1 indicates that the average horizontal speed ave│Vx│ With a search range setting coefficient Sx of a value of 0.5 to 2 proportional to the magnitude, the horizontal search area DX is as wide as 12 * Sx pixels, and the vertical search area DY is a narrow horizontally long area as 1 to 4 lines. specify.

The vector direction of MOD = 4 is ４
When <ang <2, the control signal SR1 specifies a rectangular area in which the horizontal maximum search area DX is 12 * Sx pixels and the vertical maximum search area DY is 12 * Sy lines. Note that the search range setting coefficient Sy is an average vertical speed.
ave | Vy | takes a value of 0.5 to 2 in proportion to the magnitude of ave | Vy |.

On the other hand, if the MOD is 5 to 7 and the vector direction is 2
When the above vertical movement is the main, the control signal SR1
Specifies a vertically long area where the horizontal search area DX is as narrow as 1 to 4 pixels and the vertical search area DY is as wide as 12 * Sy.

By performing the processing described above, the search area can be set to an area that matches the motion of the image.

Returning to FIG. 5, the block matching processing unit 1
3 is for the search area specified by the control signal SR1
A block matching process is sequentially performed, and a motion vector M
VBi and its prediction error ER are output.

The setting section 14 sequentially compares the magnitudes of the prediction errors ER and detects the one with the smallest prediction error. Then, the motion vector MVB and the prediction error E
Outputs Rmin.

Next, focusing on the fact that there is a fairly high correlation between the magnitude of the prediction error and the accuracy of the motion vector,
As illustrated in FIG. 7A, the search area setting unit 15 performs a second block matching process using the motion vector MVB as a starting point.
If the prediction error is small, a solid line, a dotted line,
In a large case, a control signal SR2 for specifying a region indicated by a thick line is generated. An example of this characteristic is shown in FIG. First, the prediction error E
When Rmin is less than the set value th, it is determined that the accuracy of the motion vector is high. The control signal SR2 specifies an area where the search setting coefficient Sc is 0 and both the horizontal search area DXS and the vertical search area DYS of the second search are 0. Therefore, at this time, the second search is not performed.

On the other hand, when the prediction error ERmin is equal to or larger than the set value th, it is determined that the accuracy of the motion vector is low, and the accuracy is improved in the second search. Therefore, the control signal SR2 is
The horizontal search area DXS = DX * Sc and the vertical search area DYS = DY * Sc are specified. Note that DX and DY are specified by the first search area setting unit. Further, the search setting coefficient Sc is set to 0.2 when the prediction error is from th to th1, set to 0.4 for th1 to th2, and set to 0.6 for th2 or more, and the search area is enlarged in proportion to the prediction error.

The block matching processing unit 16 in FIG.
The search area specified by the control signal SR2 is sequentially subjected to block matching processing starting from the motion vector MVB, and the motion vector MVBm and its prediction error ERR are output.

The setting unit 17 sequentially compares the prediction errors ERR in magnitude, and detects the one with the smallest prediction error. Then, the motion vector MV1 is output.

Next, the singular vector detecting section 4 will be described with reference to FIGS. FIG. 8 shows an example of the configuration of the singular vector detection unit 4, which includes a prediction error calculation unit 18, a vector correlation detection unit 19, and a determination unit 20.

As shown in FIG. 9, the vector correlation detecting section 19 examines the correlation between the motion vectors of the upper, left, right, and lower blocks adjacent to the target block, and detects a singular vector having a small correlation. Specifically, the motion vector MV of the target block is the motion vector M of the four adjacent blocks.
When it matches any of Vu, MVl, MVr, and MVd, it is determined that the correlation is high, and 0 is output to the signal SVF.
On the other hand, when the MV is different from any of the four motion vectors, the MV is determined to be a singular vector, and 1 is output to the signal SVF.
Most of the singular vectors correspond to erroneous detection vectors (vectors different from the motion of the image as described above).

The prediction error calculator 18 shown in FIG. 8 sequentially calculates the prediction error shown in the above equation (1) on the target block with respect to the above four motion vectors, and calculates the motion vector MV.
1i (MVu, MVl, MVr, MVd in FIG. 9) and a prediction error ERR.

When the signal SVF is 0, the setting unit 20 sets the motion vector MV of the target block, its prediction error,
When VF is 1, the motion vector having the smallest prediction error among the four motion vectors and the prediction error are signal MV, ER
Output to min. This signal processing greatly suppresses the generation of erroneous detection vectors.

Next, regarding the mini-block search section 5, FIG.
This will be described with reference to FIGS. FIG. 10 shows an example of the configuration of the mini-block search unit 5, which includes a division search determination unit 21, an adjacent vector generation unit 22, a calculation unit 23, and a selection unit 24.

The divided search judging section 21 calculates the prediction error ERmin
A signal CT for controlling the operation of the mini-block division search is generated in accordance with the magnitude of. Specifically, the prediction error ERmin
Is smaller than the set value, the accuracy of the motion vector is determined to be high, and 0 is output to the signal CT. On the other hand, for a block whose prediction error is equal to or larger than the set value, the motion vector is determined to be incorrect, and 1 is output to the signal CT.

As shown in FIG. 11, the adjacent vector generation section 22 outputs reference vectors MVu, MVr,..., MVd used in the mini-block division search. That is, the motion vector of the target block and the upper left, upper, upper right, left, right, lower left, lower, and lower right blocks of the periphery are output as reference vectors.

As shown in FIG. 11, the calculation unit 23 calculates a prediction error based on a reference vector for each mini-block (the size is, for example, two horizontal pixels × two vertical lines) obtained by subdividing the block in the horizontal and vertical directions. , A reference vector and a prediction error.

The selector 24 selects the motion vectors of the blocks for all the mini-blocks in the block where the signal CT is 0, and selects the reference vector with the smallest prediction error for each mini-block in the block where the signal CT is 1 Then, a highly accurate motion vector PV in pixel units is output.

According to the signal processing of the mini-block searching unit 5, as shown in FIG. 11, when there is a different motion in the block (the object above the dotted line in the drawing is in the upper right direction,
Even if the object below the bold dotted line moves leftward), the miniblock contained in the upper object has the smallest prediction error, for example, the motion vector MVu of the upper block, and the miniblock contained in the lower object has the smallest prediction error. For example, it is the motion vector MVrd of the lower right block. Therefore, the motion vector of the pixel can be detected with high accuracy in the mini-block division search processing.

It should be noted that all the pixels in the mini-block have the same motion vector, but by reducing the size of the mini-block, it is possible to obtain a motion vector having no problem in signal processing by the motion correction processing.

Finally, regarding the vector analysis unit 7, FIG.
This will be described with reference to FIG. FIG. 12 shows an example of the configuration of the above-mentioned vector analysis unit 7, and | Vx | accumulation unit 25 and | Vy | accumulation unit 2
6. Motion direction calculator 27, | Vx | average speed calculator 2
8, a | Vy | average speed calculation unit 29.

| Vx | accumulator 25 and | Vy | accumulator 26
Is the absolute value | Vx | of the horizontal component of the motion vector MV,
The absolute value | Vy | of the vertical component is accumulated between predetermined regions, and the accumulated value is output to signals S3 and S4. In the predetermined area, various areas such as one screen of an image or an N divided screen area obtained by dividing one screen into N pieces can be set.

The motion direction calculation section 27 calculates the vector direction ang | Vy | / | Vx | by performing an operation of dividing S4 by S3, and outputs it to the motion vector information MVIF1.

The | Vx | average velocity calculation unit 28 calculates the average horizontal velocity ave | Vx | by calculating S3 by dividing the number of blocks although the motion vector in the predetermined area is not 0, and outputs the result to the motion vector information MVIF2. .

The | Vy | average speed calculation unit 29 calculates the average vertical speed ave | Vy | by calculating S4 by dividing the S4 by the number of blocks although the motion vector in the predetermined area is not 0, and outputs it to the motion vector information MVIF3. .

As described above, according to the first embodiment of the present invention, it is possible to realize a motion vector detection circuit for detecting a motion vector suitable for motion correction processing with high accuracy and efficiency.

(Embodiment 2) Next, a second embodiment of the present invention will be described with reference to FIGS. The present embodiment is suitable for detecting a motion vector with higher accuracy by using information of a motion vector used in motion compensation prediction coding.

FIG. 13 shows an example of a block configuration of this embodiment, which is configured by adding a vector conversion unit 30 to the first embodiment.
Among them, the configurations and operations of the delay unit 1, representative vector setting unit 2, re-search unit 3, singular vector detection unit 4, mini-block search unit 5, memory unit 6, vector analysis unit 7, and control unit 8 are the first. This is the same as the embodiment, and the vector conversion unit 30 will be described below with reference to FIGS.

FIG. 14 shows an example of the configuration of the above-mentioned vector conversion unit 30, in which a P-vector conversion unit 31, a B-vector conversion unit 32,
It comprises a selection unit 33.

Among the motion vector information MVD obtained by decoding a signal sequence of motion compensation prediction coding (for example, MPEG video coding) by a predetermined decoding process, a motion vector used for P picture coding is a P vector conversion. The unit 31 converts the motion vector into a motion vector MTV1 per unit time.

FIG. 15A schematically shows this operation. In the P picture coding of the motion compensation prediction coding, a signal of a frame indicated by symbols I and P in an image signal sequence is coded by a motion compensation process of one-way prediction. That is, the P vector P
The difference component between the predicted frame generated by moving the position by V and the current frame is encoded. The frame of the symbol P to be encoded is taken in a plurality of frame periods (three frame periods in the figure). Therefore, this P vector PV
Of the motion vector MVT1 (N in the figure) for a predetermined unit time, for example, one frame period.
= 3).

On the other hand, the B vector conversion unit 32 converts a motion vector used for encoding a B picture into a motion vector MTV2 per a predetermined unit time. This operation is schematically shown in FIG. In the B picture encoding, a signal of a frame indicated by a symbol B in an image signal sequence is encoded by a bidirectional motion compensation process. Therefore, the B vector BV used for this encoding is two types of vectors having different directions. Therefore, a predetermined unit time, for example, one frame period is selected from these B vectors, and a motion vector MVT2 is generated.

Referring back to FIG. 14, the selection unit 33 selects a motion vector having a high reliability from the two motion vectors, in other words, a motion vector having a small prediction error, and outputs the selected motion vector to the reference vector RMV1.

The representative vector setting unit 2 sets the representative vector by using the reference vector RMV and the reference vector RMV1 together.

As described above, according to the second embodiment of the present invention, it is possible to realize a motion vector detecting circuit for detecting a motion vector suitable for motion correction processing with higher accuracy and efficiency.

Embodiment 3 Next, a third embodiment of the present invention will be described with reference to FIG. This embodiment is suitable for suppressing the amount of calculation required for detecting a motion vector in a region where a scene change has occurred.

In this embodiment, a scene change detection section 34 is added to the first embodiment, and the control section 35 controls to stop the motion vector detection operation in the scene change occurrence area. The configurations and operations of the delay unit 1, the representative vector setting unit 2, the re-search unit 3, the singular vector detection unit 4, the mini-block search unit 5, the memory unit 6, and the vector analysis unit 7 are the same as in the first embodiment. is there.

The scene change detecting section 34 extracts a frame difference signal component by a subtraction process between the signals S1 and S2, and sets a case where the number of pixels exceeding the threshold value is, for example, half or more of one screen as a scene change occurrence area. To detect. Then, 1 is output to the signal SCF. Otherwise, 0 is output. Since the level of the frame difference signal component is considerably large in the scene change area, it is desirable to set the threshold value to a considerably higher value than in the case of normal motion detection.

When the signal SCF is 1, the control unit 35 generates a signal for stopping the operation of detecting each block motion vector.

Although not explicitly shown in the drawings, when the signal SCF is 1 in the high-quality signal processing by the motion correction processing, it can be used for operation control such as stopping the motion correction processing.

(Embodiment 4) Finally, FIG. 17 shows a fourth embodiment of the present invention. In the present embodiment, a scene change detecting section 34 is added to the second embodiment, and a control section 35 controls to stop detecting a motion vector in a scene change occurrence area. Since this configuration and operation can be easily understood from the above description, the description is omitted.

As described above, according to the third and fourth embodiments of the present invention, a motion vector detecting circuit for detecting a motion vector suitable for motion correction processing with higher accuracy and more efficiently. Can be realized.

Although no particular mention has been made in the description of the embodiment, in the present invention, the calculation of the prediction error can be performed using a luminance signal or any one of a luminance signal and a color difference signal.

In the present invention, the present invention can be applied to signals in any of interlaced scanning and sequential scanning.

[0077]

According to the present invention, it is possible to realize a method and a circuit for detecting a motion vector that accurately and efficiently detect a motion vector suitable for a motion correction process for improving image quality. . As a result, a remarkable effect can be obtained in reducing the cost of the motion detection circuit and further improving the image quality of the information home appliance terminal.

[Brief description of the drawings]

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

FIG. 2 is a flowchart of motion vector detection.

FIG. 3 is a diagram illustrating a configuration example of a representative vector setting unit according to the first embodiment.

FIG. 4 is an example diagram of a reference vector.

FIG. 5 is a diagram illustrating a configuration example of a re-search unit according to the first embodiment;

FIG. 6 is a characteristic example diagram of a first search area setting unit.

FIG. 7 is a characteristic example diagram of a second search area setting unit.

FIG. 8 is a diagram illustrating a configuration example of a singular vector detection unit according to the first embodiment.

FIG. 9 is an operation schematic diagram of vector correlation detection.

FIG. 10 is a diagram illustrating a configuration example of a mini-block searching unit according to the first embodiment;

FIG. 11 is an operation schematic diagram of a mini-block search unit.

FIG. 12 is a diagram illustrating a configuration example of a vector analysis unit.

FIG. 13 is a block diagram of a motion vector detection circuit according to a second embodiment of the present invention.

FIG. 14 is a diagram illustrating a configuration example of a vector conversion unit according to the second embodiment.

FIG. 15 is an operation schematic diagram of a vector conversion unit.

FIG. 16 is a block diagram of a motion vector detection circuit according to a third embodiment of the present invention.

FIG. 17 is a block diagram of a motion vector detection circuit according to a fourth embodiment of the present invention.

[Explanation of symbols]

1 ... delay section, 2 ... representative vector setting section, 3 ... re-search section,
4: singular vector detection unit, 5: mini-block search unit, 6
... memory unit, 7 ... vector analysis unit, 8, 35 ... control unit,
9 Still and moving image block discriminating unit, 10, 23 ... computing unit, 11
... Judgment unit, 12 first search area setting unit, 13, 16 block matching processing unit, 14, 17 setting unit, 15
2nd search area setting unit, 18 prediction error calculation unit, 19 vector correlation detection unit, 20 determination unit, 21 division search determination unit, 22 adjacent vector generation unit, 24, 33 ... selection unit,
25 ... | Vx | accumulator, 26 ... | Vy | accumulator, 27 ...
Motion direction calculator, 28... | Vx | average speed calculator, 29
... | Vy | average speed calculation unit, 30 ... vector conversion unit, 3
1 ... P vector converter, 32 ... B vector converter, 34
Scene change detection unit.

Continued on the front page (72) Inventor Mitsuo Nakajima 292 Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture Inside the Multimedia Systems Development Division of Hitachi, Ltd. (72) Inventor Yasutaka Tsuru 292 Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa, Japan Hitachi, Ltd. Multimedia System Development Division (72) Inventor Takaaki Matino 292, Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa Pref. Hitachi, Ltd. AV Business Unit (72) Inventor Haruki Takada Yoshida, Totsuka-ku, Yokohama, Kanagawa No. 292, Hitachi, Ltd. In the AV Division of Hitachi, Ltd. (72) Inventor Kazuo Ishikura 5-20-1, Kamisumihonmachi, Kodaira-shi, Tokyo F-term in the System LSI Development Center of Hitachi, Ltd. 5C059 KK00 LA00 NN02 NN03 NN11 NN21 NN28 NN29 PP05 PP06 PP07 UA34 5C063 BA12

Claims (20)

[Claims] 1. A motion vector detecting method for detecting a motion vector per unit time of an image, comprising: a motion vector analyzing means for extracting an average speed and a direction of the detected motion vector; Means for setting a representative motion vector of the target block from the already detected motion vector; and a block in which a prediction error of the representative motion vector exceeds a predetermined value is searched for a predetermined search area determined by the motion vector analysis, and a motion vector is determined. Re-search means to detect, singular vector detection means to detect motion vectors with low correlation with the motion vectors of adjacent blocks and correct them to motion vectors with high correlation, and mini-blocks into which blocks are subdivided horizontally and vertically On the other hand, referring to the motion vector of the target block and the block adjacent thereto, Detection method of a motion vector, characterized by comprising a means of mini-block division search for detecting a motion vector of the click. Means for determining whether the target block is a stationary block or a moving image block. In the stationary block, the operations of the means for setting a representative motion vector and the means for re-searching are stopped. The motion vector detecting method according to claim 1, wherein 0 is assigned to the motion vector. 3. The singular vector detecting means includes:
The motion according to claim 1 or 2, wherein a correction is performed using, as a motion vector of the target block, a motion vector having a minimum prediction error in the target block among motion vectors of blocks adjacent to the target block. Vector detection method. 4. The mini-block division search means, wherein when the prediction error of the target block is smaller than a set value, the motion vectors of the target block are assigned to all mini blocks in the target block, and 4. The method according to claim 1, wherein a motion vector having a minimum prediction error in the mini-block among the motion vectors of the target block and adjacent blocks is set as the motion vector of the mini-block. The method for detecting a motion vector described in Crab. 5. The motion vector detecting method according to claim 1, wherein said prediction error is calculated by using a luminance signal component of an image signal. 6. The motion vector detecting method according to claim 1, wherein the calculation of the prediction error uses a luminance signal component and a color difference signal component of the image signal. 7. A motion vector detecting apparatus according to claim 1, further comprising means for detecting a scene change of the image, wherein the operation of detecting the motion vector is stopped in a section of the scene change. Method. 8. The motion vector detecting method according to claim 1, wherein the image signal used for detecting the motion vector is an interlaced scanning signal. 9. The motion vector detecting method according to claim 1, wherein the image signal used for detecting the motion vector is a signal in a form of progressive scanning. 10. A vector conversion means for converting a motion vector used for motion compensated prediction encoding into a motion vector per a predetermined unit time, and in the target block, the converted motion vector or the converted motion vector The method for detecting a motion vector according to claim 1, wherein the representative motion vector is set by using both the vector and the already detected motion vector. 11. A motion vector detecting unit for detecting a motion vector per unit time of an image, a motion vector analyzing unit for extracting an average speed and a direction of the detected motion vector; A representative vector setting unit that sets a representative motion vector of the target block from the detected motion vector; and a block in which a prediction error of the representative motion vector exceeds a predetermined value searches a predetermined search area determined by the motion vector analysis. A re-searching unit that detects a motion vector, a singular vector detecting unit that detects a motion vector with low correlation with the motion vector of an adjacent block and corrects it to a motion vector with high correlation, and a mini-block that subdivides blocks horizontally and vertically The motion vector of the target block and its neighboring blocks. Detection circuit of a motion vector, characterized by comprising a mini-block searching section for detecting a motion vector of the block. 12. A still / moving block determining unit for determining whether a target block is a still block or a moving block,
12. The motion vector detecting circuit according to claim 11, wherein the operation of setting a representative motion vector and re-searching is stopped in a still block, and a zero motion vector is assigned to the target block. 13. The singular vector detecting section corrects a motion vector having a minimum prediction error in the target block among motion vectors of blocks adjacent to the target block as a motion vector of the target block. The motion vector detection circuit according to claim 11, wherein: 14. The mini-block searching unit assigns motion vectors of the target block to all mini-blocks in the target block when the prediction error of the target block is less than a set value, 14. The method according to claim 11, wherein, among the motion vectors of the target block and blocks adjacent thereto, a motion vector having a minimum prediction error in a mini-block is used as the motion vector of the mini-block. The described motion vector detection circuit. 15. The method according to claim 11, wherein said prediction error is calculated by using a luminance signal component of an image signal.
15. The motion vector detection circuit according to any one of claims 14 to 14. 16. The motion vector detecting circuit according to claim 11, wherein the calculation of the prediction error uses a luminance signal component and a color difference signal component of the image signal. 17. The motion according to claim 11, further comprising a scene change detecting section for detecting a scene change of the image, wherein an operation of detecting a motion vector is stopped during a scene change. Vector detection circuit. 18. The motion vector detecting circuit according to claim 11, wherein the image signal used for detecting the motion vector is a signal in the form of interlaced scanning. 19. The motion vector detecting circuit according to claim 11, wherein the image signal used for detecting the motion vector is a signal of a progressive scanning form. 20. A vector conversion unit for converting a motion vector used for motion compensated prediction encoding into a motion vector per a predetermined unit time, and in the target block, the converted motion vector or the converted motion vector A representative motion vector is set by using a combination of a motion vector and an already detected motion vector.
9. The motion vector detection circuit according to any one of 9.