JPH07327209A - Detection of motion vector - Google Patents

Abstract

PURPOSE:To obtain a sequential scan conversion output of high picture quality by properly obtaining the motion vector used for picture processing in accordance with the classification of a picture. CONSTITUTION:A motion vector detection circuit 1 detects an optimum vector for each block of pixel without being limited at all by an input signal P. A vector V1 (horizontal component Vx1 and vertical component Vy1) is obtained as the output of the circuit 1. A motion vector detection circuit 2 detects an optimum vector V2 for each block on the condition that the vertical component of the detected vector is Vy2=2n0 (n0 is an integer). A right oblique component extraction circuit 4 detects a quantity Sob1 of right oblique components of the picture in the pixel block. A left oblique component extraction circuit 5 aetects a quantity Sob2 of left oblique components in the block. A switch 3 is controlled, and the vector V1 is selected when the picture has no remarkable oblique directivities, and the vector V2 is selected when the picture has a remarkable oblique directivity.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image motion vector detecting method, and more particularly to a motion vector used for image processing for applying image motion compensation to a 2: 1 interlaced scanning television image signal. It relates to a detection method.

[0002]

2. Description of the Related Art A block matching method, a gradient method, a phase correlation method and the like have been proposed so far as a principle method for detecting a motion vector, and these are all well-known methods. Further, as a method for removing image quality interference (interlace interference) in the current 2: 1 interlaced scanning television image signal and improving the image quality, Japanese Patent Application No. 04-283120 (conventional example 1), which is the applicant of the present invention. And there is an invention described in Japanese Patent Application No. 04-270342 (conventional example 2). The invention "scan conversion device for image signal" of the prior art example 1 is a motion compensation type progressive scan conversion for converting an interlaced scanning television image signal into a progressive scan, and the invention "motion vector detection method" of the prior art example 2 is: The present invention relates to a method of detecting a motion vector used for such scan conversion.

[0003]

In the motion-compensated progressive scan conversion of Conventional Example 1, the image quality can be effectively improved only when the vertical component V _y of the motion vector has an even number of motions of the number of scanning lines in a frame per field. Therefore, the invention of Conventional Example 2 was made. In the invention of Conventional Example 2, V _y is limited to an even number between diagonal edges and vertical edges. However, as will be described later, only diagonal edges have a positive meaning to limit the number to even numbers, and only vertical edges may be limited to even numbers.

On the other hand, the interpolation selection judgment circuit, which is a constituent element of the first conventional example, is not perfect at present, and a judgment error occurs depending on the image to deteriorate the image quality. The higher the reliability of the detected motion vector, the smaller the error. Limiting V _y to an even number in motion vector detection leads to a reduction in reliability, since not all possibilities have been verified. For this reason, although the image quality is improved at the diagonal edges in the detection method of the second conventional example, the other images result in a large number of conversion artifacts (image quality interference peculiar to the conversion device due to a determination error or the like). Was there.

Therefore, the object of the present invention is such a conventional example 2
SUMMARY OF THE INVENTION It is an object of the present invention to provide a motion vector detecting method which improves the problems of the detecting method of (1) and provides progressive scan conversion output with higher image quality.

[0006]

In order to achieve this object, the first motion vector detecting method of the present invention is for image processing in which image motion compensation is applied to a television image signal of 2: 1 interlaced scanning, A motion vector detection circuit 1 for detecting a motion vector from a 2: 1 interlaced scan signal without limiting motion vector candidates when detecting a motion vector of an image from the interlaced scan signal.
A motion vector detection circuit 2 for detecting the motion vector by limiting the vertical component of the motion vector to an even number in the unit of the number of scanning lines in a frame / field, and a vector V ₁ detected by the detection circuit 1 and the detection circuit 2, respectively. A selection switch for selecting a required vector from V ₂ is provided, and the selection switch selects the vector V ₁ when either the left or right diagonal direction component is not significant in the image portion where the motion vector is detected. In the remarkable case, the vector V ₂ is selected.

The second motion vector detecting method of the present invention
For 2: 1 interlaced scanning television image signals
For image processing applying image motion compensation, the interface
For detecting the motion vector of the image from the race scan signal
Or 2: 1 in without limiting the motion vector candidates
Motion vector for detecting motion vector from the tare scan signal
Toll detection circuit 1 and the vertical component of the motion vector
Limited to an even number of scan lines / field units
Motion vector detection circuit 2 for detecting a motion vector
And detected by the detection circuit 1 and the detection circuit 2, respectively.
Vector V₁And V ₂Choose the required vector from
A selection switch, and the selection switch
The left or right side of the image where the vector is detected
The diagonal component is not remarkable and the vector V₂Detection of
Error E₂And vector V₁Detection error E₁Ratio E₂/ E₁
Or the difference E₂-E₁Is large, vector V₁The above
Select with the select switch, otherwise vector V₂
It is characterized in that is selected.

Furthermore, in a preferred embodiment of the second motion vector detection method of the present invention, the detection circuit 1 is composed of a circuit for detecting a motion vector which gives the smallest inter-field difference, and the detection circuit 2 is the smallest. It is characterized in that it is composed of a circuit for detecting a motion vector that gives a difference between frames.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below with reference to the accompanying drawings. FIG. 1 shows the configuration of an embodiment of the motion vector detecting device according to the method of the present invention.
The technical background of the present invention will be described again in some detail with reference to FIGS.

FIG. 2 shows the prior art example 1 described previously (Japanese Patent Application No. 4-2).
No. 83120), a motion compensation type progressive scan conversion device.
The block diagram of a structure is shown. The image input signal p of this device is, for example, 525 scanning lines, a field frequency of 60 Hz,
It is a 2: 1 interlaced scanning signal, and the output signal at this time is a 525 lines / 60 Hz / 1: 1 progressive scanning signal.

In this device, the signal q of the interpolation scanning line necessary for the progressive scanning conversion is converted into the output signal qm of the motion compensation interpolation circuit 13.
It is obtained by mixing c and the output signal qvf of the intra-field interpolation circuit 14 in the mixer 15 according to the mixing ratio K _m . p and q are time-axis converted by the time-axis conversion circuit 16 and then combined to form an output signal. The motion vector detection unit 11 detects the motion vector V _c of the image from the input signal p for each block of the image. In circuit 13, V _C
Is used to perform motion-compensated interpolation of the scanning line, and the interpolation selection determination circuit 12 determines for each pixel whether V _c is accurate or not and determines K _m . In this way, the motion vector V _c affects the operation of each unit, and its content greatly affects the converted image quality.

FIG. 3 is a view of the input image p as viewed on the time-vertical axis, and the white circles in the figure show the positions where the scanning lines exist.
Focusing on the scanning line a, it is necessary to create the interpolation scanning line q in the middle of the scanning line b immediately above. Motion vector V
_When the vertical component of _c is 2n (the number of scanning lines in a frame / field, n is an integer), interpolation can be performed by directly shifting the scanning line c or d of the previous field according to the value of V _y as shown in the figure. However, if this is not the case, interpolation that is effective in terms of image quality cannot be performed (1 in FIG.
There is no signal in front of the field. For a more detailed description of FIGS. 2 and 3, refer to the publication of Conventional Example 1). Motion vector detection is generally performed for each small block of an image,
When the images are viewed locally, they can be roughly classified as shown in FIG.
4A, 4B, and 4C are edge images, and FIG. 4D is an example of other images. Images other than the edges include a lattice image as shown in FIG. 4D, a random image such as noise, a point image, and the like. Table 1 shows whether or not interlace interference can be removed by motion-compensated progressive scan conversion. When the motion of the image is V _y = 2n, the removal is possible (◯) in any case, but when V _y ≠ 2n, the removal is impossible (x) in some cases.

When the motion of the image is other than V _y = 2n, the diagonal edge image cannot be removed if the vertical component V _yc of the vector V _c used for motion compensation is V _yc = V _y , but V _yc ≠ Let V _y be V _yc = 2n _o (n _o is an integer other than n)
Then, it can be removed (the reason is the local simplicity of the edge image, for details, refer to the publication of the above-mentioned conventional example 2).
Vertical edges are always removable, and horizontal edges and other images (except uniform images with little change) are always non-removable.

[0014]

[Table 1]

From the above, the motion vector detecting method of the second conventional example is such that V _yc = 2n _o is always maintained except for the lateral edge. However, the operation of the interpolation selection determination circuit 13 is not perfect at present, and a determination error occurs depending on the image. For this reason, artifacts (image quality interference peculiar to the conversion device) occur in the interpolation of the scanning lines, and the converted image quality deteriorates. This determination error is due to the motion vector V _c
The higher the reliability of, the less. Limiting V _c to V _yc = 2n _o leads to a decrease in reliability because the optimum one is not detected from all motion vector candidates in motion vector detection. Therefore, in the motion vector detecting method of the present invention, V is used for images other than the diagonal edge image.
without a limit of _yc = 2n _o, it is intended to detect more reliable motion vector.

Next, examples of the method of the present invention will be described. The entire circuit shown in FIG. 1 can be used as the motion vector detection unit 11 shown in FIG. The motion vector detection circuit 1 detects the optimum vector for each block of pixels from the input signal p without any limitation. This is a state called full search in the block matching method which is one of the motion vector detecting methods, for example. At this time, the vector detection accuracy may be an integer both horizontally and vertically, or may be a finer accuracy. The output of circuit 1 is the vector V ₁ (horizontal component V
_x1 and vertical component V _y1 ).

On the other hand, the motion vector detection circuit 2 detects the optimum vector V ₂ for each block under the condition that the vertical component V _y2 of the detected vector is V _y2 = 2n _o (n _o is an integer). To do. It is possible to impose such a condition in motion vector detection even in an actual circuit. For example, in the block matching method, a motion vector with the smallest matching error is searched, but the search candidate may be limited to a vector of V _y2 = 2n _o . The vectors V ₁ and V ₂ are selected by the switch 3 according to the signal K _ob described later, and become the output vector V _c (V _xc , V _yc ).

The right diagonal component extraction circuit 4 is arranged in the pixel block.
The amount S of the right diagonal component of the image_ob1To detect. For example, in FIG.
(B) is a right diagonal edge image.
S_ob1Is a large value. Conversely, the left diagonal component extraction circuit 5
The amount S of the left diagonal component in the block _ob2To detect. Direction
The determination circuit 6 is S_ob1And S_ob2And compare both values
If there is no difference, it is assumed that there is no directivity and the control signal K_obTo
Set to "1". K if there is a big difference_obIs set to “0”
It This K_obControl switch 3 by_obBut
"1", that is, the left or right side of the image is significantly slanted
Vector V if there is no tropism₁If you select
Cutle V₂Select. These circuits 4, 5 and 6
The image directionality detecting circuit 7 is configured.

A more specific circuit example of the circuits 4, 5 and 6 will be shown below. FIG. 5 is a specific example of the circuits 4 and 5. The image input signal p is 1H delay (1 line delay) H21, H
Delayed by 22. Input signals p and H21,
The horizontal delay 23-27 of the image in clock units is connected to the output of H22. Here, for example, Delay 2
(LS) D of 4 represents an (LS) clock delay. The same applies to the other cases. L is a fixed value and S is a circuit parameter. The value of S is determined by cut and try in an actual circuit. These delays 23,
And the output signals of the delays 24-27 are the signals a and e, respectively.
.. h, the arrangement on these screens is as shown in FIG.
However, this figure shows the case of S = 2. In the adder 28, the signals e and h are added, and the output is added to 1 /
Multiply by 2. The subtractor 30 subtracts the output of the coefficient unit 29 from the signal a to obtain the signal obd1. That is, obd1 = a− (e + h) / 2 (1), and as shown in FIG. 6, the signal obd1 is the output of the high-pass filter in the diagonal left direction. Therefore, the signal obd1 has a large absolute value when the image has an oblique right edge or the like.

Adder 31, 1/2 coefficient unit 32, subtractor 3
3 similarly forms a high-pass filter in the right diagonal direction, and its output obd2 = a- (g + f) / 2 (2) has a large absolute value in a left diagonal edge image or the like. These signals obd1 and obd2 are calculated for each pixel. These are further converted into squared values by the squarers 34 and 35, and are integrated in the pixel block (B) by the integrators 36 and 37. The output signals of the integrators 36 and 37 are the amount S _ob1 of the right diagonal component and the amount S _{ob2 of} the left diagonal component in the _block.
Is.

FIG. 7 shows a specific example of the directionality determining circuit 6.
The divider 41 calculates (S _ob1 / S _ob2 ), and the divider 42 calculates the reciprocal (S _ob2 / S _ob1 ). In the maximum value detection circuit (MAX) 43, the larger one is selected. The output of the circuit 43 takes a value of 1 or more, which is close to 1 when S _ob1 and S _ob2 have similar values, but becomes a value sufficiently larger than 1 if there is a large difference between the two values. For example, in each of the images shown in FIGS. 4A, 4C, and 4D, S
_{Both ob1} and S _ob2 have large values, and the output of the circuit 43 is close to 1. However, in the right oblique edge image of FIG. 7B, S _ob1 is large and S _ob2 is small, and the output of the circuit 43 is small. Has a large value. In this way, it is possible to determine whether or not the image is a diagonal edge image based on the magnitude of the output of the circuit 43. In the comparator 44, the output of the circuit 43 is compared with the threshold TH _ob, and if it is smaller than the value TH _{ob, the} output K _ob is “1”,
Otherwise, it is set to "0". TH _ob is determined by cut and try in the actual circuit.

FIG. 8 relates to claim 2 of the present invention.
It is another embodiment. Circuits 1, 2, 3, 7 are the same as in FIG.
is there. However, the vectors detected from circuits 1 and 2 respectively
Le V ₁, V₂Error E₁, E₂Are output at the same time
ing. This error is one measure of vector reliability
The normal motion vector detection algorithm
When detecting, a numerical value similar to some error is sought
There is. For example, in the block matching method
It is the difference. Error E in the divider 51₂And E₁Ratio of (E₂/ E
₁) Is obtained, and the ratio is determined by the threshold value in the comparator 52.
Compared to THEr and only compared if greater than THEr
The output Ker of the container 52 is set to "1". Otherwise K
er becomes “0”. Ie E₂Is E₁Enough compared to
Ker becomes “1” only when it is larger. These 51,5
2 as a whole constitutes an error comparison circuit. Threshold
THEr is determined by cut and try. Ke
r and K which is the output of the circuit 7_obAND gate 5 is the logical product of
4, the output of gate 54 controls switch 3
It

The meaning of these circuits is as follows.
The features of the vectors V ₁ and V ₂ are summarized as follows. (Interlace interference removal capacity) V _1: V ₂ less than V _2: large (especially diagonal edge image) (reliability vector) V _1: High V _2: less than V ₁ equal to or V ₁

[0024] Therefore, the circuit of Figure 1 is to the outside of a large diagonal edge image in comparison with the particular vector V ₁ interference removal ability of V ₂ tries to select a reliable vector V _1. However, the determination as to whether or not the image is a diagonal edge image cannot always be said to be complete in an actual complex image, and the vector V ₁ may be selected even in the image in which the vector V ₂ should be selected. On the other hand, in terms of reliability, the vector V ₁
Is always more reliable than or equal to V _2, but the vector V ₁
In many cases, there is not much difference in reliability (value of error) between V ₂ and V ₂ . In such a case, the interference removal capability of the vector V ₂ should be expected rather than the selection of the vector V ₁ . Therefore, the circuit shown in FIG. 8 attempts to select the vector V ₁ only when the image has no directionality and the reliability of the vectors V ₁ and V ₂ is different.

Next, an embodiment according to claim 3 of the present invention will be described. The block matching method used for the motion vector detection method is the most widely used motion vector detection method at present due to its performance and simplicity of the algorithm. When the motion vector is detected from the 2: 1 interlaced scanning signal by the block matching method, two methods can be considered. That is, there are a method of configuring a circuit so as to give a minimum inter-frame difference as a vector and a method of configuring a circuit so as to give a minimum inter-field difference as a vector. However, each of them has different drawbacks.

The former is the vertical component V _{y of the} motion vector.
Can be easily detected in both even and odd cases. However, since a vector is detected based on a difference between images temporally separated by two fields, a detection error is likely to occur especially in an image having a periodic structure such as a striped pattern, and an incorrect vector is used for inter-field motion compensation. When applied, it causes visually noticeable image disturbance.

The latter has no problem of such an error because it detects a vector based on the difference between one field which is the minimum unit in terms of time. However, as described in the above-mentioned conventional example 1, it is a vector with an even V _y that the image quality is great in the inter-field motion compensation.
The latter cannot easily detect this V _y even vector. If spatial interpolation of the image signal is performed before obtaining the inter-field difference, even number of vectors can be detected. However, the interpolated signal contains many folds in the spatial frequency spectrum, and accurate even-numbered motion vector detection cannot be expected despite the large circuit scale required for the interpolation.

In any case, in the third aspect, the detection circuit 2 is required to detect the motion vector by limiting V _y to an even number in the unit of the number of scanning lines in the frame / field.
With regard to the above, the above-mentioned detection circuit 1 which does not necessarily need to detect even-numbered motions is configured by a block matching circuit that detects a vector that gives the smallest inter-frame difference.
Is to configure a circuit by block matching for detecting a vector that gives a minimum inter-field difference, thereby further reducing the error in vector detection according to claim 2 and attempting to accurately detect a motion vector. . In claim 2, the above consideration is not made.

FIG. 9 shows a motion vector detection circuit of an embodiment according to claim 3, but the same reference is given to the block (circuit) which has the same operation as the block explained in FIG. 8 or an operation similar to them. Numbered. The signal p which is an input of the circuit is, for example, a television image signal of 525 scanning lines, field frequency 60 Hz, 2: 1 interlaced scanning.

The broken line block 1 in the figure shows the components 61 to 6
It is composed of 6 blocks, and each block of an image is appropriately set (for example, horizontal 16 pixels ×
The motion vector detection circuit 1 detects a motion vector in every 8 lines (vertical line). The circuit 1 is constructed based on the block matching method. In the block matching method, when an object is moving in parallel at a speed of V _T (pixels / frame) in one frame period, if the image one frame before is shifted by V _T pixels, it matches the image of the current frame. The principle is that the intra-block cumulative value of the inter-frame difference shows the minimum value.

In the figure, the signal p is delayed in the delay 61 by (1 field + V _T1 ). V
_T1 (V _xT1, V _yT1 ) is a candidate vector that is the target of motion vector detection, and all candidate vectors within the range preset by the specifications are sequentially generated from the candidate vector generation circuit 62. Here, V _{xT1 and} V _yT1 are the horizontal and vertical components of V _T1 , respectively. Element 6
The operation of 1, 62 will be described in more detail later. In the subtractor 63, the difference between the signal p and the output of the delay 61 is calculated to obtain the motion compensation inter-field difference signal dfd1 (displace
d field difference) is required. The block area setting circuit 64 generates a block area signal B indicating whether or not the pixel is a pixel in a preset block. Element 65, the absolute value of the block in the accumulated value of dfdl, Σ _B | dfd1 | a seek. This value is the candidate vector V _T1
And is represented by DFD1 (V _T1 ). DFD
1 (V _T1 ) is the matching error of V _T1 for that block. The minimum value detection circuit 66 finds the minimum value of DFD1 and outputs V _T1 that produces the minimum value as the motion vector V ₁ (V _x1, V _y1 ) of the block. Also, DF
The minimum value of D1 is output as the error E ₁ of the vector V ₁ .

The broken line block 2 is composed of components 71 to 76, and limits the vertical component of the motion vector to an even number in the unit of (scan line / field),
This is a motion vector detection circuit that detects a motion vector for each block that is the same as the circuit 1. Like the circuit 1, the circuit 2 is also constructed based on the block matching method.

In the figure, the signal p is delayed by (1 frame + 2V _T2 ) in the delay 71. V
_T2 (V _xT2 , V _yT2 ) is a candidate vector that is the target of motion vector detection, and the candidate vector generation circuit 72 sequentially generates preset candidate vectors. Here, V _{xT2 and} V _yT2 are the horizontal and vertical components of V _T2 , respectively. The operation of elements 71 and 72 will be described in more detail below. In the subtractor 73, the difference between the signal p and the output of the delay 71 is calculated to obtain the motion compensation interframe difference signal dfd2 (displaced frame difference). The operation of elements 74-76 is exactly the same as circuits 64-66. The element 76 is a motion vector V ₂ (V
_x2, V _y2 ), and its error E ₂ are output. Circuit 7,
53, 54, and 3, a more accurate vector suitable for inter-field motion compensation of the image p is selected from V ₁ and V ₂ and output as the output vector V _c (V _xc, V _yc ). These circuits are the same as those shown in FIG.

Next, referring to FIG. 10, elements 61, 62, 7
The operations of 1, 72 will be described. The figure is a view of the scanning of the image as seen in the time-vertical plane. The horizontal axis of the figure is time t,
The vertical axis represents the vertical direction y of the screen. A, B, C,
---, a, b, c, --- etc. are symbols attached to each scanning line. Hereinafter, the vertical component V _y of the candidate vector will be considered (the suffixes T1 and T2 will be omitted below unless otherwise necessary for the sake of simplicity of description). here,
The unit of V _y is scanline / field (in frame). As for the horizontal component, there are no special circumstances regarding odd-even as follows.

First, regarding the difference between frames, V_yof
The relationship between the value and the position of the scanning line is as shown in FIG.
In the circuit 2 based on the difference between frames, the delay 71
Delay amount of 525H (H is 1 line (1 field
Inner scanning line)), the signal p scans the scanning line R.
Sometimes the output of delay 71 is just one frame
Before, that is, the scanning line E. Therefore, the difference between E and R is V
_yIt is an inter-frame difference corresponding to = 0. The amount of delay
When it is decreased by one line and becomes 524H, the delay 71
The output is the scanning line F, and (F−R) is V_yCorresponds to = 1
It is the difference between the frames. 1 line is a scan line in the frame
The delay amount of delay 71 is
As shown (1 frame + 2V_T2) Can be expressed as Less than,
Similarly, V _y= ---, -4, -3, ---, 3, 4, ---
Then, the inter-frame difference as shown in FIG.
If the minimum value of these differences is detected, V_yAll about
All integer motion vectors can be detected. But the circuit
2 is V_yT2Element 72 to V
_yT2V is an even number_T2Output only sequentially.

Regarding the inter-field difference, the relationship between V _y and the scanning line is as shown in FIG. 10 (b). In the circuit 1 based on the inter-field difference, when the delay amount of the delay 61 is 262H and the signal p is scanning the scanning line R, the output of the delay 61 is the scanning line e. At this time, (e−R) is the inter-field difference corresponding to V _y = 1. When the delay amount is reduced by one line and is 261H, the output of the delay 61 is the scanning line f and (f-
R) is the inter-field difference corresponding to V _y = 3. Therefore, like the delay 71, the delay amount of the delay 61 can be expressed as (1 field + V _T1 ). Hereafter, similarly,
Differences between fields as shown in FIG. 10B correspond to V _y = ---, -5, -3, ---, 3, 5, ---. When the motion vector is detected by detecting the minimum value of these differences, only odd-numbered vectors can be detected with respect to V _y without limiting the motion vector candidates in terms of the circuit. The element 62 sequentially outputs all candidate vectors that can be detected by such inter-field difference.

As described above, the circuit 1 can detect only a vector with an odd V _y, but the operation is necessary and sufficient for the embodiment of FIG. In the circuit of FIG. 8 and the invention of claim 2, when the diagonal direction of the image is not remarkable, the errors E ₁ and E
Selection of V _1, V ₂ is made by _two large and small. Therefore, even if V _y can detect only odd-numbered motions as V ₁ , if diagonal directionality is not remarkable, if V t is an even vertical component of the true motion of the image, good matching can be achieved with respect to V _1. Inevitably, E ₁ > E ₂ , and V ₂ is selected, so that the correct vector can be detected in the entire circuit of FIG. 9.

As described above, in addition to the circuit of FIG. 9 operating correctly, by using the circuit 1 as the detection circuit based on the inter-field difference, the detection error of the motion vector causing the image quality disturbance is as follows. Can be reduced.

FIG. 11 shows an example of motion vector detection and motion compensation for an image. The image is two horizontal line images in which black and white are inverted for every two intra-field scanning lines, and the vertical component of the true movement is V _y = 1. The notation in the figure is the same as that in FIG. 10 except that black and white are added to the scanning lines.

In FIG. 11, if a vector is detected by the difference between frames, it corresponds to the motion of V _y = 1 (J−R), (K−R ¹ ), and the motion of V _y = 5 (L). Since all the differences such as −R) and (M−R ¹ ) are zero, V _y = 1 and 5 are both strong candidates for the vector. Which one is selected is determined for reasons unrelated to the original image content, such as noise contained in the image and preset circuit priorities. But,
When these vectors are used for inter-field motion compensation as described in Conventional Example 1, a large difference occurs in the result of motion compensation. That is, when a scanning line is interpolated at a point R _m between R and R ¹ by motion compensation, a white scanning line is correctly interpolated as an average value of the scanning lines j and k when V _y = 1 is detected. . However, when V _y = 5 is detected, l, m
The black scan line is interpolated as the average value of This is apparently a false signal, and is a noticeable image quality disturbance in an actual image. As described above, in the motion vector detection based on the inter-frame difference, it is not possible to exclude motion vector candidates that cause image quality obstruction.

On the other hand, when the vector is detected by the inter-field difference, the difference (1-R), (m-R ¹ ) has a large value, so that the motion vector of V _y = 5 is detected. There is no. Since the differences (s−R) and (t−R ¹ ) are also zero, the movement of V _y = 9 may be detected, but
In this case, the white scanning line is correctly interpolated as the average value of s and t, and no problem occurs in inter-field motion compensation. In this way, the circuit of FIG. 9 is always able to detect the correct motion vector for motion compensation, for motions with odd V _y .

Although the present invention has been described in detail with reference to the examples, the present invention is not limited to these examples.
It is obvious that various modifications and changes can be made within the scope of the invention. For example, although the block matching method is assumed as the basic motion vector detecting method in the embodiment of the present invention, it can be realized by using another detecting method (for example, the gradient method).

[0043]

As described above, according to the method of the present invention, the motion vector used for the image processing in which the motion compensation of the image is applied to the television image signal of the interlaced scanning is appropriately adjusted according to the type of the image. Therefore, it is possible to obtain progressive scan conversion output with higher image quality.

[Brief description of drawings]

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

FIG. 2 is a diagram showing a configuration example of a motion compensation type progressive scan conversion apparatus.

FIG. 3 is a diagram for explaining the operation of the apparatus shown in FIG.

FIG. 4 is a diagram showing types of images.

5 is a diagram showing a specific circuit example of circuits 4 and 5 of the apparatus shown in FIG.

FIG. 6 is a diagram for explaining the operation of the circuit shown in FIG.

7 is a diagram showing a specific circuit example of a circuit 6 of the apparatus shown in FIG.

FIG. 8 is a diagram showing a configuration of an embodiment according to claim 2 of the present invention.

FIG. 9 is a diagram showing a configuration of an embodiment according to claim 3 of the present invention.

FIG. 10 is a diagram illustrating an inter-frame difference and an inter-field difference.

FIG. 11 is a diagram illustrating an operation example of motion vector detection and motion compensation.

[Explanation of symbols]

 1 motion vector detection circuit 1 2 motion vector detection circuit 2 3 selection switch 4 right diagonal component extraction circuit 5 left diagonal component extraction circuit 6 directionality determination circuit 7 image directionality detection circuit 11 motion vector detection unit 12 interpolation selection determination circuit 13 motion Compensation interpolation circuit 14 In-field interpolation circuit 15 Mixer 16 Time axis conversion circuit 21, 22 1 line delay 23-27 Horizontal delay 28, 31 Adder 29, 32 1/2 multiplier 30, 33, 63, 73 Subtraction Unit 34, 35 Square unit 36, 37 Accumulator 41, 42, 51 Divider 43 Maximum value detection circuit 44, 52 Comparator 53 Error comparison circuit 54 AND gate 61, 71 Delay circuit 62, 72 Candidate vector generation circuit 64, 74 Block area setting circuit 65, 75 Absolute value in-block accumulation circuit 66, 76 Minimum value detection circuit

Claims (3)

[Claims] 1. Image processing for applying image motion compensation to a 2: 1 interlaced television image signal,
In detecting a motion vector of an image from the interlaced scanning signal, a motion vector detection circuit 1 for detecting a motion vector from a 2: 1 interlaced scanning signal without limiting motion vector candidates, and a motion vector vertical direction. It is necessary from the motion vector detection circuit 2 for detecting the motion vector by limiting the component to an even number in the unit of the number of scanning lines in a frame / field, and the vectors V ₁ and V ₂ detected by the detection circuit 1 and the detection circuit 2, respectively. And a selection switch for selecting a vector to be used, and in the selection switch, if the left or right diagonal component is not significant in the image portion where the motion vector is detected, the vector V
_A method of detecting a motion vector, characterized in that ₁ is selected, and when it is remarkable, the vector V ₂ is selected. 2. Image processing for applying image motion compensation to a 2: 1 interlaced scanning television image signal,
In detecting a motion vector of an image from the interlaced scanning signal, a motion vector detection circuit 1 for detecting a motion vector from a 2: 1 interlaced scanning signal without limiting motion vector candidates, and a motion vector vertical direction. It is necessary from the motion vector detection circuit 2 for detecting the motion vector by limiting the component to an even number in the unit of the number of scanning lines in a frame / field, and the vectors V ₁ and V ₂ detected by the detection circuit 1 and the detection circuit 2, respectively. A selection switch for selecting a vector to be set, and in the selection switch, one of the left and right diagonal components is not significant in the image portion where the motion vector is detected, and the vector V
_Second detection error E ₂ and the ratio of the detection error E ₁ vector V ₁ E
₂ / E ₁ or the difference E ₂ −E ₁ is large, vector V
A motion vector detecting method characterized in that ₁ is selected by the selection switch, and if not, the vector V ₂ is selected. 3. The method according to claim 2, wherein the detection circuit 1 is composed of a circuit that detects a motion vector that gives a minimum inter-field difference, and the detection circuit 2 gives a motion vector that gives a minimum inter-frame difference. A motion vector detection method comprising a detection circuit.