JP2003047011A - Motion correcting circuit using motion vector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the distortion of an interpolated image caused by sensing erroneously motion vectors, when sensing the motion vectors of image signals and performing motion correction of the image by using the sensed motion vectors. SOLUTION: When dividing a digitized image signal into blocks each of which comprises m pixels × n lines (m and n are integers) to sense a motion vector by using the block as a unit, first and second motion vector sensing circuits 4, 6 are operated independently of each other by using a signal A of a certain field and a signal B of another field separated from the field by at least one field to perform the sensing of the motion vectors in two directions. In this case, the first motion vector sensing circuit 4 senses a motion vector by using the field of the signal A as a reference, and the second motion vector sensing circuit 6 senses a motion vector by using the field of the signal B as a reference. By using these two kinds of sensed motion vectors, an interpolating circuit 12 performs the interpolating processing of a dynamic image.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for detecting the magnitude and direction of motion of a moving image, that is, a motion vector in a television signal, and performing a motion correction using the detected motion vector.

[0002]

2. Description of the Related Art Motion vectors are used to improve interframe coding efficiency in high-efficiency coding of television signals and to reduce motion discontinuity due to conversion of the number of fields in television conversion. Has been.

A general motion vector detection method is a method of detecting a motion vector for each block after subdividing a television signal into blocks of m pixels × n lines (m and n are integers). The pattern matching method disclosed in JP-A-55-162683 and JP-A-55-162688, and the iterative gradient method described in JP-A-60-158786 are well known.

Further, in order to improve the detection accuracy of the motion vector using the iterative gradient method, the one using the initial displacement vector is described in, for example, Japanese Patent Application Laid-Open No. 62-206980.

[0005]

A typical example of the motion correction process using the motion vector described above is a motion interpolation process for converting the number of fields in the television system conversion. The most problematic point in the motion interpolation process using such a motion vector is image distortion due to the motion vector detection.

For example, in the case of detecting a motion vector between two consecutive fields, one of them is used as a reference field, and then the other field is searched for to detect the motion vector. Field F
For the signal of p, the current field Fc shown in FIG.
When detecting the motion vector of the signal of, the accurate motion vector cannot be detected in the region Z where the background B of the still image S has moved. The reason is that the motion vector can be detected only when the moving signal exists in the next field, and the signal appearing from the back of the still image S like the area Z does not exist in the previous field Fp. The vector cannot be detected.

Therefore, in the conversion of the television system,
When the interpolation process for correcting the motion is performed using the motion vector, image distortion occurs in the interpolated signal in the region where the motion vector cannot be detected.

FIG. 7 is an explanatory diagram of this phenomenon. When the motion vector detected with respect to the previous field Fp with respect to the current field Fc is V and the field interpolation ratio in the television system conversion is γ, The field Fi obtained by interpolation for correcting the motion is shown. The interpolation field Fi exists on an axis obtained by dividing the current field Fc and the previous field Fp into γ: 1-γ. The motion vector V is normally detected between a certain block A1 of the current field Fc and the corresponding block A2 of the previous field Fp, and in the interpolation field Fi,
The block A0 is properly obtained from these blocks A1 and A2 and the motion vector V.

However, in the blocks B1 and C1 of the current field Fc, the corresponding blocks B2 and C2 in the previous field Fp are located behind the still image area indicated by diagonal lines, so that the motion vector cannot be detected and the interpolation is performed. Since there is no optimum image in the corresponding block B0 of the field Fi, image distortion will occur. The block of the current field Fc corresponding to the block B0 of the interpolation field Fi is the block B1, but since the block B2 corresponding to this block B1 does not exist in the previous field Fp, the motion vector is not detected, and the block B1 is not detected.
0 is not detected either.

That is, the block B0 of the interpolation field Fi
Since the block C1 of the current field Fc and the block A2 of the previous field Fp on the same coordinate axis are moving images, the images are not the same, and the optimum block for the block B0 cannot be detected. As a result, the image distortion is generated in the block B0. Will occur.

An object of the present invention is to solve the above problems,
An object of the present invention is to provide a motion correction method using a motion vector that can reduce the correction image distortion due to the false detection of the motion vector.

[0012]

A motion correction method using a motion vector according to the present invention for achieving the above object is
Digitized video signal is m pixels x n lines (m, n
Is a whole number) and the signal A of a certain field is used when detecting a motion vector in units of these blocks.
And a signal B of a field at least one field away from this field, and a first motion detection circuit for detecting a motion vector with the field of the signal A as a reference, and a motion vector with the field of the signal B as a reference. The second motion vector detection circuit for detection is operated independently of each other to detect the motion vector in both directions,
It is characterized in that the motion interpolation processing is performed using these two types of detected motion vectors.

[0013]

1 is a block circuit diagram showing the configuration of an example of an apparatus for carrying out a motion correction method using a motion vector according to the present invention. For convenience of explanation, the input signal applied to the video input terminal 1 is simply the luminance signal Y. This input signal is supplied to the prefilter 2 to limit the band. The pre-filter 2 can usually be composed of a two-dimensional or three-dimensional low pass filter (LPF) and is provided to reduce a motion vector detection error. You can also delete it.

A first motion vector detection circuit 4, which detects a motion vector of the output signal of the prefilter 2 through a 1-field delay circuit 3 with the previous field as a reference,
The signal is supplied to the delay circuit 5 and also supplied to the second motion vector detection circuit 6 and the delay circuit 7 which detect the motion vector with the current field as a reference. In this embodiment, since the motion vector is detected between successive fields, the 1-field delay circuit 3 is provided. However, when detecting the motion vector between frames, the 1-field delay circuit 3 is replaced by 1 field delay circuit 3. A frame delay circuit may be provided.

First and second motion vector detection circuits 4 and 6
The method of detecting the motion vector in (1) may be the block matching method, the iterative gradient method, or the phase correlation method, and any method can be used as long as it detects the motion vector for each block. Further, the motion vector may be detected for each pixel.

First and second motion vector detection circuits 4 and 6
Is supplied to the first and second shift circuits 8 and 9 and the motion determination circuit 10, respectively. These first and second shift circuits 8 and 9 are supplied to the previous fields supplied via the delay circuits 5 and 7, respectively, by the motion vectors detected by the first and second motion vector detection circuits 4 and 6, respectively. And the image signal of the current field are shifted.
The delay circuits 5 and 7 delay the image signal by the time required for the motion vector detection by the first and second motion vector detection circuits 4 and 6.

Further, the output signals of the first and second shift circuits 8 and 9 are supplied to the motion judging circuit 10. The video signal output from the delay circuits 5 and 7 is also supplied to the motion determination circuit 10. In the present embodiment, in the motion determination circuit 10, the sum of the absolute values of the differences between the output of the delay circuit 5 and the output of the first shift circuit 8 and the output of the delay circuit 7 and the second shift circuit 9 are provided. The output is compared with the sum of absolute values of the differences, and the motion vector when the sum is the smallest is output as a true motion vector. The difference value in pixel units between fields is converted into an absolute value, and the accumulated value for the size of the block is used to determine the motion vector. When the motion vector is accurately detected, the difference between the motion correction fields is calculated. Since the value is "0", this is made an absolute value, and the cumulative total is also "0". As for the reason for accumulating the block size, since the influence of noise and the like becomes large when judging for each pixel, the value accumulated for the block size is used because of the meaning of smoothing. However, it is desirable that the size of this block is smaller than the block for motion vector detection. According to such a motion judging method, the structure of the motion judging circuit 10 becomes the simplest.

The image signal from the video input terminal 1 is further supplied to the motion interpolation circuit 12 via the delay circuit 11 and to the motion interpolation circuit 12 via the 1-field delay circuit 13 and the delay circuit 14. This one-field delay circuit 13 is not necessary when the prefilter 2 is not a three-dimensional filter but a two-dimensional filter. The delay circuits 11 and 14 delay the image signal by the processing time required for the motion correction determination. The motion interpolation circuit 12 performs motion correction interpolation processing based on the optimum motion vector calculated by the motion determination circuit 10, and supplies an interpolated image signal to the video output terminal 15.

In the present embodiment, as described above, in order to reduce the interpolated image distortion due to the motion vector detection error that occurs when there is no moving image in either the previous field or the current field, the first motion vector Two types of motion vectors are detected: a motion vector detected by the detection circuit 4 with the previous field as a reference and a motion vector detected by the second motion vector detection circuit 6 with the current field as a reference. A true motion-corrected image is obtained by using the two types of motion vectors thus detected and the motion-correction signal that has been shifted by the magnitudes of the motion vectors by the first and second shift circuits 8 and 9 as evaluation parameters. To generate. As the simplest evaluation, the magnitude of the inter-field difference value, which is the cumulative value in the block of the absolute value of the difference value between the fields whose coordinates are shifted by the motion vector, is used, and the smaller value is selected.

FIG. 2 is an explanatory diagram of a motion correction process of a method of reducing discontinuous distortion of a moving image due to conversion of the number of fields in the conversion of the television system. The labeling method of FIG. 2 is basically the same as that of FIG. 7, but is arranged in order to make the effects of the present invention easier to understand. It shows a situation in which the signal of the interpolation field Fi is generated from the signal of the current field Fc and the signal of the previous field Fp by the interpolation processing based on the interpolation ratio γ. Further, as the screen, a case where there is a still image area in the moving image background as shown in FIG. 6 is shown. Further, FIG. 2 shows the motion vector detected by each motion vector detection circuit and the motion correction signals A1 and A2.

As shown in FIG. 2, in the motion vector detected based on the current field Fc, blocks B1 and C
1 cannot be detected because there is no corresponding signal in the previous field Fp. However, in the motion vector detection based on the previous field Fp, the motion vector of the block A2 can be detected, and the motion vectors of the blocks B2 and C2 can also be detected. Further, with the motion vector detected based on the current field Fc, the motion vector of the block after the block J1 can be detected, but the previous field Fp
In the motion vector detected based on
The motion vector of 2, G2 cannot be detected.

FIG. 3 collectively shows the above results, and the detected motion vector is V. In FIG. 3A, "1" indicates a motion vector detectable area when the current field Fc is used as a reference, and "0" indicates a motion vector undetectable area. Further, “1” in FIG. 3C indicates a motion vector detectable area based on the previous field Fp, “0” indicates a motion vector undetectable area, and FIG. The signal on the insertion field is shown.

From FIG. 3, in the area a, the previous field F
It can be seen that the image distortion in the interpolation field Fi can be reduced by using the motion vector detected based on p and the motion vector detected based on the current field Fc in the region b.

Next, the parameters of the adaptive interpolation will be described. For ease of explanation, the motion vector is one-dimensional and the fields used are the current field signal and the previous field signal in the television conversion device.

As parameters, the total sum of absolute values of two types of inter-field difference values based on the interpolated block, the magnitude of the detected motion vector, and the distribution state are used. This interpolation size is a block size smaller than the size of the motion vector detection block, and is 4 pixels × 2 lines here, but it is of course also possible in pixel units. Further, the sum of absolute values of two types of inter-field difference values is the following two quantities.

(1) Sum of absolute values of inter-field difference signals corrected by the motion vector detected with the current field Fc as a reference.

(2) Sum of absolute values of inter-field difference signals with zero motion.

The motion vector distribution, as shown in FIG. 4, has a motion vector detection block size as a unit, and a distribution of motion vectors detected on the basis of the current field Fc on the current field Fc and a previous field Fp. It is the distribution of the motion vector detected as the reference on the previous field.

In the present embodiment, such a motion vector distribution and the magnitude of the motion vector are adaptively combined so that an optimum interpolation signal can be obtained.

First, the simplest example will be described with reference to FIG. The interpolation block is C0, and the motion vector detected with reference to the block A0 of the current field Fc is V
A0, the motion vector detected with reference to the block B0 of the previous field Fp is VB0, the sum total of the absolute values of the motion compensation inter-field difference values is DFD0, and the sum total of the absolute values of the inter-field difference values of motion zero is FD0. Therefore, D
FD0 = Σ | A0-B0 |, and FD0 = Σ | A01
-B01 |.

Generally, when DFD0 <FD0, it is determined that the motion vector is normal, and when DFD0> FD0, it is determined that the motion vector is not normal or a still image. In this case, the motion vector is " 0 ”.

In the present embodiment, a motion vector distribution condition is added to such a condition. The movement is uniform,
When the motion vector is normally detected, the motion vector VA0 detected based on the block A0 of the current field Fc and the motion vector VB0 detected based on the block B0 of the previous field Fp are equal to each other,
The signs of these motion vectors are opposite to each other, so
| VA0 + VB0 | ≈0.

Therefore, under such a condition, the DFD
When 0 is compared with FD0, if a fixed amount or the variable δ is added to FD0, even if DFD0> FD0, DFD0 <FD0 + δ, so the motion vector is selected.

Next, when | VA0 + VB0 |> α2 (α2 is an integer of 4 or more, for example), it is determined that the motion vector cannot be detected normally, and if a fixed amount or variable δ is subtracted from FD0, even if DFD0> Even with FD0,
Since DFD0> FD0-γ, the motion vector is selected.

As described above, in the present embodiment, not only the conventional inter-field difference value is used for evaluation, but the motion vector distribution is added to the evaluation condition, whereby more accurate motion correction can be performed. .

Further, generalizing such a vector distribution evaluation method, the motion vector detected with one field as a reference is stored in the first memory, and the motion vector detected with the other field as a reference is stored in the first memory. 2 and stores the motion vectors of the blocks whose coordinates are shifted from the first and second memories with reference to the interpolated block by VAf and VBf. Let FD be the value and DFD be the difference value between motion compensation fields, α1 be an integer close to 0, α2 be an integer close to 0 and be less than or equal to α1, and α3> α1
, (A) | VAf |> α1, | VBf |> α
In the case of 1, VAf + VBf <α2, the motion correction threshold value is lowered to make it easier to select the motion correction, and (b) | VAf
In the case of |> α1, | VBf |> α1, and VAf + VBf> α3, the motion correction threshold value is raised to make it difficult to select the motion correction.

Next, another embodiment of the motion correction method using the motion vector according to the present invention will be described. In this embodiment, the following three inter-field difference values are used.

(1) Sum of absolute values of inter-field difference signals corrected by motion vectors detected with reference to the current field: DFDA

(2) Sum of absolute values of inter-field difference signals corrected by the motion vector detected with reference to the previous field: DFDB

(3) Sum of absolute values of inter-field difference signals with zero motion: FD0

Then, as a judgment condition, DFDA <D
When FDB and DFDB <FD0, the motion vector detected based on the current field is used.

When FDA> DFDB and DFDB <FD0, the motion vector detected based on the previous field is used.

When FDA> DFDB and DFDB> FD0, it is determined that there is no motion and the motion vector zero is selected.

By adding the above-mentioned motion vector distribution condition to these conditions, a more accurate correction signal can be obtained.

Basically, the present invention can be applied to all image processing for performing interpolation processing using motion vectors. In the above-described embodiments, the television system conversion device has been shown as an example, but the invention can be applied to the following other applications.

(A) An interpolation device that uses a motion vector for inter-field interpolation processing when converting an interlaced signal into a non-interlaced signal.

(B) A noise reduction device using a motion vector.

(C) Generation of field interpolated image in high-efficiency encoder.

[0049]

As described above, according to the motion correction method using the motion vector according to the present invention, an error of the motion vector occurs in principle only by detecting the motion vector in one direction, and thus the erroneous detection is performed. Image distortion is induced when interpolation processing is performed using the generated motion vector, but by detecting multiple motion vectors, motion vector detection error is reduced and image distortion due to motion interpolation processing is also reduced. It

[Brief description of drawings]

FIG. 1 is a block circuit configuration diagram of an apparatus that implements a motion correction method using a motion vector.

FIG. 2 is an explanatory diagram of motion correction processing by the motion correction method of the present invention in television format conversion.

FIG. 3 is an explanatory diagram of a detectable area and an undetectable area of a motion vector.

FIG. 4 is an explanatory diagram of motion vector distribution.

FIG. 5 is an explanatory diagram of motion correction parameters.

FIG. 6 is an explanatory diagram of occurrence of a motion vector detection error.

FIG. 7 is an explanatory diagram of motion correction processing in a conventional television system conversion device.

[Explanation of symbols]

1 Video input terminal 2 Pre-filter 3,13 1-field delay circuit 4, 6 Motion vector detection circuit 5, 7, 11, 14 Delay circuit 8, 9 shift circuit 10 Motion determination circuit 12 Motion interpolation circuit 15 Video output terminal

   ─────────────────────────────────────────────────── ─── Continued front page    F-term (reference) 5C059 KK01 LB18 NN02 NN21 NN28                       NN40 PP04 TA65 TA66 TB08                       TC12 TD05 TD12 UA12 UA34                 5C063 BA08 BA12 CA05 CA38                 5J064 AA01 BA15 BB01 BB03 BB08                       BB14 BC24 BC25 BC27 BD02

Claims (7)

[Claims] 1. A digital video signal is represented by m pixels × n.
The signal is divided into blocks of lines (m and n are integers), and when detecting a motion vector in units of these blocks, a signal A of a certain field and a signal B of a field at least one field away from this field are used, The signal A
To detect a motion vector based on the field of
Motion detection circuit and a second motion vector detection circuit that detects a motion vector with reference to the field of the signal B are operated independently of each other to detect a motion vector in both directions. A motion correction method using a motion vector, characterized in that the motion interpolation process is performed using the motion vector. 2. The signal A is a signal of the current field,
2. The motion correction method using a motion vector according to claim 1, wherein the signal B is a signal of a field one field before the current field. 3. The motion using the motion vector according to claim 2, wherein the detected motion vector is evaluated using a vector distribution and a cumulative value of inter-field differences obtained by shifting the vector coordinate as parameters. Correction method. 4. As a method of evaluating the vector distribution, a motion vector detected with the field of the signal A as a reference is stored in a first memory, and a motion vector detected with the field of the signal B as a second is stored. In the memory, and the motion vectors of the blocks whose motion vector amount coordinates have been shifted from the first and second memories with reference to the interpolated block are set as VAf and VBf, respectively, and the inter-field difference value in which the motion is zero. Is FD, the difference value between motion compensation fields is DFD, α1 is an integer close to 0, and α2 is 0
When α3 is an integer close to or less than α1 and α3> α1, (a) | VAf |> α1, | VBf |> α1, VAf +
When VBf <α2, the motion correction threshold value is lowered so that the motion correction can be easily selected. (B) | VAf |> α1, | VBf |> α1, VAf +
4. When VBf> α3, the motion correction threshold value is increased so that it is difficult to select motion correction.
A motion correction method using the motion vector described in. 5. As a method of evaluating the vector distribution, an inter-field difference value that is motion-compensated with a motion vector detected based on the field of the signal A is DFDA, and a motion vector detected based on the field of the signal B is used. When the motion-corrected inter-field difference value is DFDB, (a) when DFDA <DFDB, the motion vector detected based on the field of the signal A is selected, and (b) when DFDA> DFDB. The motion compensation method using a motion vector according to claim 3, wherein the motion vector detected based on the field of the signal B is selected. 6. The motion correction method using a motion vector according to claim 1, wherein the signal A is a signal of a current frame and the signal B is a signal of a frame one frame before the current frame. . 7. A digitized television signal,
It is divided into blocks of m pixels × n lines (m and n are integers), the motion vector is detected for each block, the position of the image is corrected using the detected motion vector, and the motion-corrected inter-field When generating an interpolated image using the signal, the size of the detected motion vector,
Motion compensation using a motion vector characterized by changing the interpolation ratio of temporal linear interpolation of a signal between two fields or between two frames using a distribution, a motion compensation inter-field difference value, and an inter-field difference value of zero motion as parameters. Method.