JPH0715700A - Apparatus and method for movement corrected processing of video signal - Google Patents

Abstract

PURPOSE: To compensate the movement of a picture by detecting the maximum correlation ridge on the correlation face of an input picture, and generating a movement vector. CONSTITUTION: This is a movement correlation video signal processor in which a movement vector indicating the movement of a picture between a pair of input pictures of an input video signal is generated. This device is provided with a means 190 which compares one of a pair of input pictures with a retrieval area having plural blocks in the one side and in the other side of a pair of input pictures, and forms a correlation face having a series of correlation values indicating correlation between the retrieval blocks and the retrieval area, means 210 which detects the maximum correlation value in the correlation face, and means 230 which generates the movement vector corresponding to the position of the maximum correlation value in the correlation face (including a ridge detecting means which detects the correlation value of the correlation face within the total threshold values of the maximum correlation value indicating the maximum correlation value of a ridge).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion compensation video signal processing apparatus and method.

[0002]

2. Description of the Related Art A motion-compensated video signal processing apparatus uses a television standard conversion and a film standard conversion.
ersion) and various conversion devices such as a video system and a movie system.

In a motion compensated television standard converter described in GB-A-2231749, a motion compensating television standard system motion is processed to indicate the motion of an image between a pair of input images. I am trying to generate a set of vectors. Since this process is done in separate blocks of the image, each motion vector will indicate the motion between the images contained in the respective block.

In the motion vector estimating process, a correlation surface showing a spatial correlation between blocks of two input images is inspected to detect a maximum correlation point. (This correlation surface actually represents the difference between the two input images, which means that its maximum correlation point is at the true minimum point on the correlation surface (hereinafter called the minimum point. .)). When the minimum point is detected, a motion vector is generated from the spatial position that is the minimum point on the correlation surface. A check is made to determine if this minimum point represents a valid peak in the correlations associated with the remaining correlation surfaces. If this minimum passes this test,
Treat as "valid" and set the valid flag associated with that motion vector.

Next, each motion vector set is sent to a motion vector reducer, which extracts a subset from the (valid) motion vector set for each block, and a subset of this motion vector. To a motion vector selector, which assigns one of the subsets of motion vectors to each pixel in each block of the image. The motion vector selected for each pixel is sent to the motion correction interpolator, and the motion correction interpolator interpolates the output image from the input image according to the motion between the input images.

[0006]

However, when the image includes a moving object, and the moving object is longer in one direction than the length of the search block used for the block matching process, there is a problem. There is. In this case, since the moving image is out of the search block used for the block matching process, the component of the movement of the object along the vertical axis cannot be detected (in this search block). Therefore, a "one-dimensional" or ridge minimum (ridge minimum) (ridge correlation maximum) is detected.

FIG. 1 shows a correlation surface 5 having a minimum ridge 10.
For example: This minimum ridge defines the component of motion of the object in a direction perpendicular to the ridge direction (which is along the vertical axis of the object), but does not indicate the motion of the object along the ridge direction. Absent.

In the conventionally proposed motion vector estimator, the motion vector generated from the minimum ridge is
It was not used for interpolation of the output image because the aptitude test could not be performed. Instead, a zero motion vector was used by default. However, using a zero motion vector when interpolating pixels of an object with long motion can result in defects such as partial omission of the object. FIG. 2 shows this type of defect. Figure 2
At, interpolation of the output image with a vertically moving elongated object 15 is shown. The normal motion vector 20 is detected above and below the moving image 15, but the defective zero motion vector 25 is detected at the center of the moving image 15.
A gap 30 is created between the upper part, which is interpolated using the normal motion vector, and the central part, which is interpolated using the zero motion vector.

Therefore, an object of the present invention is to provide a means for detecting a maximum correlation ridge in a correlation plane in a motion correction video signal processing device for generating a motion vector indicating an image motion between a pair of input images of an input video signal. It is to provide.

[0010]

Specifically, the present invention relates to a motion-compensated video signal processing device for generating a motion vector indicating the motion of an image between a pair of input images of an input video signal. , Comparing the pair of input images with one of the pair of input images and a search region having a plurality of blocks in the other of the pair of input images, and comparing the pair of input images with each other between the search block and the search region. Means for generating a correlation surface having a series of correlation values showing correlation, means for detecting the maximum correlation value in the correlation surface, and generating a motion vector according to the position of the maximum correlation value in the correlation surface And ridge detecting means for detecting the correlation value of the correlation surface within the total threshold value of the maximum correlation value indicating the maximum correlation value of the ridge.

The present invention, in the above embodiment, further comprises means for detecting the maximum correlation ridge in the correlation plane. This detection may be followed by various steps such as not using the resulting motion vector, using a larger search block to avoid this problem, or re-qualifying the motion vector.

Further, the ridge detecting means used in the above embodiment of the present invention has means for detecting whether or not there is a required number of correlation values within the range of the total threshold value of the maximum correlation value.
The ridge detection means includes means for detecting whether or not there is a smaller second predetermined number of correlation values within the range of the total threshold value of the maximum correlation values so that appropriate measures can be taken according to the sharpness of the ridges. And the second predetermined number is greater than the first required number.

Further, in the above-mentioned embodiment of the present invention, the apparatus is arranged to detect the maximum correlation value of the ridge and detect the position of the correlation values of the two maximum intervals in the correlation plane within the range of the total threshold of the maximum correlation value. In accordance with the above, there is a means for generating an average motion vector.

Further, in the above embodiment of the present invention, the ridge detecting means detects the area of the correlation surface having these correlation values from within the range of the total threshold value of the maximum correlation value and the movement of the vertical image. A means for detecting the ratio of the length of the area in the direction shown to the length of the area in the direction showing the movement of the horizontal image; and a means for comparing this ratio with a first predetermined value in the direction showing the movement of the horizontal image. It has means for detecting the maximum correlation value of the ridge, and means for comparing this ratio with a second predetermined value to detect the maximum correlation value of the ridge that can be placed in the direction indicating the movement of the vertical image.

Also, in the above embodiment of the present invention, the apparatus includes means for setting a ridge flag indicative of the maximum correlation value of the ridge so that detection of the maximum correlation value of the ridge can be fed to other parts of the apparatus. Good to have

Further, in the above embodiment of the present invention, the ridge flag may indicate whether or not the maximum correlation value of the ridge is in the direction indicating the movement of the horizontal image or the direction indicating the movement of the vertical image.

Further, in the above embodiment of the present invention, the maximum correlation value of the ridge is detected so that the motion vector component generated from the minimum value of the ridge is assigned as an effective value and the motion vector can be used for interpolation of motion correction. Means for comparing the motion vector under test generated from the generated correlation surface with the next motion vector generated from the pair of input images, and the corresponding component of the next motion vector and the motion vector under test are actually the same. And a means for detecting whether or not the next motion vector is generated from the correlation surface in which the maximum correlation value of the ridge is detected, and the next motion vector is the maximum correlation of the ridge. Replaces the component of the motion vector under test in the ridge direction with the corresponding component of the next vector when it is detected that the value did not originate from the detected correlation surface And a means.

It is also preferable that the next motion vector is generated in the ridge direction from the correlation surface replaced with the motion vector under test.

The next motion vector is preferably generated in the ridge direction from a correlation surface which is vertically adjacent to the motion vector under test.

Also, in order to enable requalification of the motion vector generated from the correlation surface having the maximum correlation ridge which indicates the movement of the long object in the diagonal direction, the next motion vector is under examination in the ridge direction. It may be generated from the correlation surface that is diagonally adjacent to the motion vector.

Further, the threshold value is preferably an average value of correlation values on the correlation surface.

The correlation surface has a correlation value that increases as the correlation increases, but the correlation surface can be simplified by having a series of correlation values indicating the difference between the capacities of the search block and the search area. In this way, the maximum correlation value is actually indicated by the minimum value on the correlation surface.

The correlation value indicates the difference in chromaticity, or the chromaticity and brightness of the search block, and the component between the search block and the search area. The correlation value indicates the difference in the brightness component between the search block and the search area. Good to show.

The motion compensated video signal processing device according to the present invention is particularly advantageous when used as part of a motion compensated television standard converter.

The second embodiment of the present invention provides a motion-compensated video signal processing method for generating a motion vector indicating the motion of an image between a pair of input images of an input video signal, and comprises the following steps. . That is, a search block in one of the pair of input images is compared with a search area having a plurality of other blocks of the pair of input images, and a series of correlation values indicating the correlation between the search block and the search area. Forming a correlation surface having :, detecting a maximum correlation value in the correlation surface, generating a motion vector based on the position of the maximum correlation value in the correlation surface, the maximum correlation showing a maximum correlation ridge Detecting a correlation value on the correlation surface within a total threshold of values.

[0026]

Embodiments of the present invention will be described below with reference to FIGS. FIG. 3 is a schematic block diagram of a motion compensation television standard conversion device (hereinafter, simply referred to as “conversion device”). This conversion device takes in an input interlace-scanned (interlaced-scan) digital input video signal (for example, 1125/60 2: 1 high-definition video signal (HDVS)) 50, and interlaced-scan output digital video signal. (For example, 1250/50 2: 1 HDVS) 60 is output.

First, the input video signal 50 is supplied to the input buffer / packer 110. When the input video signal is a standard definition signal, the input buffer / packer 110
Sets the format of the image data to a high definition format (aspect ratio of 16: 9) by supplementing black pixels at necessary positions. When the input video signal is HDVS, the input buffer / packer 110 simply buffers the data.

This data is stored in the input buffer / packer 1
10 is sent to the matrix circuit 120. In the matrix circuit 120, the format of the input video signal is “CCIR recommended standard 601” (Y, Cr,
Converted to Cb) format.

The input video signal is sent from the matrix circuit 120 to the time axis conversion (TBC) / delay circuit 130, and is also sent to the subsample time axis conversion / delay circuit 180 via the subsampler 170. The time axis conversion / delay circuit 130 identifies the temporal position of each field of the output video signal, and when the output field is used for the interpolation processing, the two of the video input signals closest in time to this output field are used. Select a field. For each field of the output video signal, the two input fields selected by the time axis conversion / delay circuit 130 are appropriately delayed and sent to the interpolator 140, which interpolates the output field. Control signal t indicating the temporal position of each output field with respect to the two selected input fields
Are supplied from the time axis conversion / delay circuit 130 to the interpolator 140.

Sub-sample time base conversion / delay circuit 180
Operates similarly to the time base conversion / delay circuit 130,
The spatially subsampled video signal sent by the subsampler 170 is used. The sub-sampling time-axis conversion / delay circuit 180 selects from the sub-sampled video signals, and the pair of input fields is used to generate a motion vector.

Time axis conversion / delay circuits 130 and 180
Can operate according to a synchronization signal included in the input video signal, the output video signal, or both. When only one of the sync signals is supplied, the timing of the field of the video signal of the input video signal and the output video signal to which the sync signal is not supplied is set in the time axis conversion / delay circuits 130 and 180. It occurs deterministically.

Sub-sample time base conversion / delay circuit 180
The pair of fields of the video input signal, which are selected by and subsampled by, are sent to a motion processor 185, which direct block matcher 190,
Data stripper 200, motion vector estimator 21
0, motion vector reducer 220, motion vector selector 2
30 and a motion vector post-processing device 240. The input field pair is first the direct block matcher 19
0, this direct block matcher 190
Compute a correlation surface that shows the spatial correlation between the search block that is faster in time and the search block that is later (larger) in time of the two selected input fields. The data showing these correlation surfaces are data stripper 200
Is reformatted by the motion vector estimator 210
Sent to. The motion vector estimator 210 detects the maximum correlation point on the correlation surface. (In practice, these correlation surfaces are
It shows the difference between the blocks of the two input fields, which means that the maximum correlation value is at the true minimum value on the correlation surface, which maximum correlation point will be referred to below as the "minimum value". ) In order to detect the minimum value, interpolation processing is performed between the additional points on the correlation surface to compensate for the reduction in resolution that occurs when the correlation surface is generated using the sub-sampled video signal. ing. The motion vector estimator 210 generates a motion vector from the minimum value detected for each correlation surface, and sends this motion vector to the vector reducer 220.

The motion vector estimator 210 also performs a confidence check on each generated motion vector, determines whether or not the motion vector is greater than the total noise level, and indicates a confidence flag indicating the result of the confidence check. Is added to each motion vector. This confidence check is known as the "threshold" test, but it is (along with the other features of the device shown in Figure 3) GB-A-2 231749.
No.

Further, by the motion vector estimator 210,
A check is also performed to detect whether each motion vector is a false motion vector. In this test, the correlation surface (excluding the exclusion areas around the detected minimum) is examined and the second smallest minimum is detected. If the second minimum value does not exist at the boundary of the exclusion area, the motion vector generated from the first detected minimum value is set as a possibility that it is a false motion vector, and the flag is set.

Motion vector reducer 220 operates to reduce the range of motion vector selections that may occur for each pixel of the output field prior to providing the motion vector to motion vector selector 230. The output field is virtually divided into blocks of pixels, and the position of each block in the output field corresponds to the position of the search block in the preceding input field of the selected input fields. The motion vector reducer 220 composes a motion vector consisting of four motion vectors to be incorporated into each block of the output field into a group, and finally each pixel of each block is extracted from the group having four motion vectors. Select one and interpolate.

Even if the motion vectors have the flag set as "possibly false", they are not set in the adjacent block, that is, they are the same as the non-false motion vector. , Are requalified during the motion vector reduction process described above.

The motion vector reducer 220, as part of its function, is the frequency of occurrence of "normal" motion vectors (eg, those that have passed confidence and false checks, or have been requalified to be non-false). , But do not consider the block positions of the input fields used to obtain these motion vectors. Normal motion vectors are classified in order of frequency of occurrence. The most normal motion vectors having a large difference from each other are classified as wide area motion vectors. The three types of motion vectors that have passed the reliability check are selected for each block of output pixels, and are sent to the motion vector selector 230 including a motion vector of zero for the next processing.

These three kinds of selected motion vectors are selected from the following three motion vectors in a predetermined selection order. (I) Motion vector generated from corresponding search block (ii) Motion vector generated from surrounding search block (local motion vector) (iii) Wide area motion vector.

The motion vector selector 230 also receives the two input fields selected by the sub-sample time base conversion / delay circuit 180 and used to calculate the motion vector. These fields are appropriately delayed and input so that motion vectors can be supplied to the motion vector selector 230 at the same time as they are generated from these fields. The output provided by the motion vector selector 230 is one motion vector per pixel in the output field. This motion vector is selected from the four motion vectors of the corresponding block sent from the motion vector reducer 220.

The vector selection process includes the step of detecting the degree of correlation between the test blocks of the two input fields indicated by the motion vector under test. The motion vector having the highest degree of correlation between the inspection blocks is selected and used for interpolation of output pixels. Also, the motion vector selector 230
To generate a "movement flag". This motion flag is "still" if the degree of correlation between blocks indicated by a motion vector of zero is greater than a preset threshold.
Set as (no movement).

The vector post-processing unit 240 reformats the motion vector selected by the motion vector selector 230 to arbitrarily reproduce the vertical size of the pixel, and also interpolates the reformatted motion vector into the interpolator 140. Send to. By using the motion vector, the interpolator 140 converts the time axis according to the motion of the image indicated by the motion vector currently supplied to the interpolator.
The delay circuit 130 interpolates one output field from the two (non-subsampled) interlaced scanned input fields.

If the motion flag indicates that the current output pixel is in the moving part of the image, the pixel resulting from the two selected fields sent to the interpolator 140 is:
The two input fields are combined in a relative ratio depending on the temporal position of the output field (indicated by the control signal t). For this purpose, the closer input field with the larger proportion of these two input fields is used. Temporal weighting if the motion flag indicates that it is set to "still"
Does not. The output of the interpolator 140 is sent to an output buffer 150 that outputs an HDVS signal and a down converter (standard definition converter) 160 that generates a standard definition output signal 165 using a motion flag.

The sub-sampler 170 performs horizontal and vertical spatial sub-sampling of the input field sent from the matrix circuit 120. Subsampling in the horizontal direction is a simple process in which the input field is first pre-filtered by a half-bandwidth low-pass filter (2: 1 horizontal decimation process in this example), every other video on each video line. By removing the samples, the number of samples on each video line can be reduced by a factor of two.

Vertical subsampling of the input field is complicated by interlaced scanning of the input video signal. This is because the spacing between consecutive lines of video samples in each interlaced scanned field is actually the spacing of two completed video signal lines, and the position of the lines within each field is This is because one line is vertically displaced from the field line before or after the video line of the selected frame. The actual vertical sub-sampling method used is to first perform vertical low-pass filtering (to prevent aliasing distortion), then add each pixel to 1 of the video line.
Filtering processing is performed to displace downward (odd field) or upward (even field) by ½ of the space for one line. The displaced field obtained in this manner is approximately equal to the field of the progressive scanning frame which has been subjected to the half sub-sampling processing in the vertical direction.

FIG. 4 shows the motion vector estimator 21 of the present invention.
It is a block diagram which shows the detail of 0. In FIG. 4, a correlation surface having a series of correlation values is input from the data stripper 200 and stored in the correlation surface memory 300. Correlation values are stored in the correlation surface memory 300 from the “first” minimum value detector 3
Read by 10, this minimum detector 310 scans the correlation surface to detect one minimum from this. First
The minimum point detector 310 of 3 outputs the following three conditions. That is, the actual minimum correlation value (MV) 320, the average correlation value (AV) 330 of the correlation surface, and the minimum correlation value 32 of the correlation surface.
The coordinate value (or position) (MC) 340 is 0.

The first scan is performed and the minimum correlation value (MV)
After 320 is detected, a second inspection of the correlation surface is performed by the second minimum detector 360. The second minimum value detector 360 inputs the correlation value from the correlation surface memory 300 together with the coordinate values of these correlation values. Further, the coordinate value of the correlation value is sent to the exclusion signal generator 370, and the exclusion signal generator 370 outputs an exclusion flag 380 to surround the coordinate value 340 of the minimum correlation value 320 with, for example, 5 × 5. Mask exclusion areas with correlation values. Then, the second minimum detector 360 detects the minimum correlation value outside this exclusion area. The detected second minimum correlation value is a threshold value comparison process for comparing the difference between the second minimum correlation value and the first minimum correlation value 320 with a threshold value derived using the average value 330 of the correlation surface. Receive. The minimum correlation value 320 and the second (next)
If the difference from the minimum correlation value is greater than the threshold,
The "Valid Vector" flag (VV) 390 is set, indicating that the first minimum correlation value 320 has a large correlation peak value.

A third inspection of the correlation surface is performed by the "similarity" minimum detector 400. This similarity minimum value detector 4
00 detects all correlation values of the correlation surface that are within the threshold of the minimum correlation value 320. The threshold value used in the similarity minimum value detector 400 is again derived using the average correlation value 330. Each of these points is the minimum value detector 400 for similarity.
Minimum counter / comparator 410 when detected by
(Hereinafter, simply referred to as "minimum value counter".) The count value held in the counter is incremented.

The minimum value detector for similarity 400 detects the positions of two correlation values having the maximum interval on the correlation surface within the range of the threshold value of the minimum correlation value 320. The minimum coordinate value 420 of these maximum intervals is sent to the vector averager 440.

The vector averager 440 has a minimum value 420 of the maximum intervals detected by the similarity minimum value detector 400.
An output 450 indicating the motion vector generated from the average position of
Is output to the output vector selector 460. In addition, the vector averaging unit 440 uses the minimum value 42
Detect the spatial association of 0 and detect whether the ridge is actually in the horizontal or vertical direction. The result of this test is
It is output as a “ridge direction” flag 445.

The valid vector flag 390 and the minimum coordinate value 340 are sent to the vector generator 470, where (if the first minimum value is valid) the minimum coordinate value 340.
To generate a motion vector.

The count value held in the minimum value counter 410 is compared with two threshold values indicating the ridge region 480.
When the count value of the minimum value counter 410 indicates that the count value is larger than the first predetermined number and smaller than the second predetermined number with respect to the minimum value detected on the correlation surface, it indicates that the minimum ridge is detected. The "ridge" flag (VR) 490 and the "ridge" flag 495 are set. However, in reality,
Ridge vector 4 generated from vector averager 440
Use 50 as valid. In this case, the output vector selector 460 sends the ridge vector 450 as the output vector 465.

When the number of detected minimum values is larger than the second predetermined number, the effective ridge flag 490 is not set and the ridge flag 495 is set. This indicates that the ridge vector 450 should not be used and the motion vector should either be requalified (described below) or not used. In this case, the vector output selector 460 outputs the motion vector generated from the vector generator 470 as the output vector 465.

If the number of detected minimum values is less than the first predetermined number, neither the effective ridge flag 490 nor the ridge flag 495 is set. in this case,
The output vector selector 460 outputs the motion vector generated from the vector generator 470 as the output vector 465.

5a, 5b and 5c respectively show the operation of the motion vector estimator shown in FIG. 4 when the effective ridge flag 490 is set.

5a and 5b show a horizontal motion component 5
Each successive image is a vertical elongated object with 10 ( v ). In a and b of FIG. 5,
The image blocks 520 and 530 are compared and the correlation surface 540 is
It is shown that in forming (c in FIG. 5), a number of minimum values 550 with very similar correlation values occur.

When the correlation surface 540 is processed by the motion vector estimator 210 shown in FIG. 4, the two minimum values 560 and 570 of the maximum intervals are sent to the vector averager 440. The vector averager _{440, (v a + v b)} /
Generate a ridge vector 450 equal to 2, which is, for example, a motion vector indicating the average position of the two minima 560,570. When the number of minimum values 550 exceeds a predetermined number, the output vector selector 460 outputs the ridge vector 450 as the output vector 465.

FIG. 6 likewise shows a second method for detecting the minimum ridge. In the first inspection of the correlation surface 600, the minimum correlation value 610 is detected. All correlation values within the threshold of this minimum correlation value 610 are detected and form a threshold region 620. However, compared to the method described with reference to the motion vector estimator of FIG. 4, this method cannot detect the minimum number within the threshold of the minimum correlation value 610. Instead, the "aspect ratio", which is the ratio of the (horizontal) length to the (horizontal) width of the threshold region 620, is detected.

From this aspect ratio, the threshold region is determined as shown in FIG.
~ It is classified into one of the three categories shown in FIG. Figure 7
The 1st example of classification of a threshold area is shown. In FIG. 7, when the detected aspect ratio is between the first and second predetermined values and the threshold region 630 is not elongated, the minimum value is the effective coordinate in the horizontal direction (x) and the vertical direction (y). It is classified as the minimum point having a value.

FIG. 8 shows a second example of classification of the threshold regions. In FIG. 8, the minimum area 640 has an aspect ratio larger than the first predetermined value, and this area has a vertical direction 65.
It shows that it is elongated at 0. This minimum is partitioned as the vertical minimum ridge, and the resulting motion vector has a horizontally defined component 660, but a vertical 650 component is undefined.

Similarly, FIG. 9 shows a third example of classification of the threshold region. In FIG. 9, the threshold region has an aspect ratio smaller than the second predetermined value and indicates the minimum horizontal ridge. As a result, the motion vector has a fixed vertical component, but the horizontal component is undetermined.

From the classification of the threshold regions described above, a ridge flag is attached to each motion vector to indicate the classification of the minimum value at which the motion vector has occurred. That is, the ridge flag (which may be 2 bits) indicates whether the motion vector is generated from the minimum point of the minimum vertical ridge or the minimum horizontal ridge.

When the motion vector is generated from the minimum ridge, the component of the motion vector in the ridge direction (for example, the vertical direction in the case of the vertical minimum ridge) is undetermined.
However, as a part of the vector reduction processing operation of this example, the processing of re-qualifying the motion vector generated from the minimum ridge in the horizontal or vertical direction is performed by allocating the effective vector component in the ridge direction.

FIG. 10 outlines the requalification of motion vectors originating from the minimum ridge. Motion vector reducer 2
At 20, the motion vectors, along with their associated ridges and confidence flags, are stored in memory array 680 at a position corresponding to the relative position of the search block that generated this motion vector. When the motion vector generated from the minimum ridge appears during the motion vector inspection, the motion vector in the adjacent ridge direction is inspected,
It is detected whether a vector component of 90 degrees with respect to the ridge direction matches (for example, within a threshold range) the corresponding component of the motion vector generated from the minimum ridge.

If there is a match, adjacent vectors are considered to be associated with the same elongated moving object. If the adjacent vector was itself the one that originated from the smallest ridge, then testing in the ridge direction continues. Therefore, the motion vector must have an actual identical component of 90 degrees to the ridge direction, unless it originates from a minimum point (eg, this point has a defined component along the ridge direction). I understand. This motion vector represents the movement of the edge of the elongated object, and this motion vector component in the ridge direction is considered to be equal to the previously undetermined component of the motion vector under examination in the ridge direction.

FIG. 10 shows two examples of the above processing (one in the horizontal direction and the other in the vertical direction). In FIG. 10, in the case of the vertical direction, the motion vector 700 is
It has a ridge flag indicating that it originated from the minimum vertical ridge, and has a normal horizontal (x) component except for the undetermined vertical (y) component. Next, the adjacent motion vector 710 in the ridge direction is inspected and it is detected that it has a matching x component except for the undetermined y component (this motion vector 710 is also generated from the vertical minimum ridge). . Similarly, the following vectors 720 and 73
The search for 0 also represents the matching x component except for the undetermined y component. In this way, a motion vector 740 having a matching x component and a determined y component is detected (this motion vector originates from the minimum point). Motion vector 740
Represents the motion of the edge of the elongated object, so the y component of motion vector 740 is considered to be assigned to motion vector 700. The ridge flag associated with this motion vector 700 is modified to indicate that the motion vector 700 is requalified (by setting the ridge flag to indicate that it originated from a minimum point), resulting in output pixel It can be used to interpolate.

A corresponding process is also shown for requalification of motion vectors generated from horizontal minimum ridges. In this case, the first search from left to right is not valid. This is because it hits the end of the memory array 680 that stores the motion vector. Next, a second search is performed from the right to the left, and three motion vectors 760 having ay component that matches the motion vector 750 are detected except for the undetermined x component.

As a result, the motion vector 7 having the y component and the determined x component that match the components of the motion vector 750.
70 is detected. The x component of motion vector 770 is assigned to motion vector 750 and the ridge flag is modified as described above.

FIG. 11 is a schematic block diagram showing a part of the motion vector reducer 220 (FIG. 3) that performs the above processing. In FIG. 11, the motion vector input from the motion vector estimator 210 (FIG. 3), along with their confidence and ridge flags, is input vector memory 80.
Written to zero. The vector processing device 810 reads out the stored motion vector and performs the requalification process described with reference to FIG. Before performing the next vector reduction process, the motion vector including the re-qualified one output from the vector processing device 810 is output to the output vector memory 8
Written in 20. Input and output vector memory 80
The separate use of 0 and 820 is to prevent the increase in error in the series of vectors resulting from the requalification process.

FIG. 12 shows the vector processing device 81 of FIG.
It is a flowchart which shows the process by 0. Step 9
At 00, the motion vector under test (along with the ridge flag) is read from the input vector memory 800. In step 910, the ridge flag associated with the vector under test is examined to detect if the vector under test originated from the minimum horizontal or vertical ridge. Then, in step 920, if the vector under test originates from the minimum point, the vector under test and its accompanying flag are written to the output vector 820. Return to step 900 to read the next motion vector from the input vector memory 800.

Next, in step 930, if the motion vector under test originates from the smallest ridge, the adjacent motion vector in the ridge direction is read. In step 940, adjacent 90 ° motion vector components with respect to the ridge direction are compared to the corresponding motion vector component under test. If there is a difference in these values as a result of the comparison and there is more than the total threshold value, the process proceeds to step 950, and the vector under inspection is written to the output vector memory 820 without changing (without requalification). After writing, step 9
00, the next motion vector under inspection is read from the input vector memory 800.

As a result of the inspection in step 940,
If the adjoining vector has almost the same component of 90 degrees with respect to the ridge direction, the process proceeds to step 960 to inspect the ridge flag associated with this adjoining vector. Step 930 if the adjacency vector itself originates from the minimum ridge.
Return to and check the next adjacency vector.

If the test at step 960 indicates that the adjacent vector did not originate from the minimum ridge, the process proceeds to step 970 and the adjacent vector component is assigned to the motion vector under test. Then, in step 980, the ridge flag attached to the motion vector under test is changed to indicate that the motion vector under test has been requalified. This process is accomplished by setting the ridge flag for the motion vector under test so that it indicates that it originated from the minimum point.

Finally, in step 990, the changed value of the motion vector under inspection is written (along with the changed flag) in the output vector memory 820, and the process returns to step 900.

In the above-mentioned processing, it is possible to perform inspection so that a vector oblique to the vector derived from the minimum ridge is also improved so that the same vector component of 90 degrees in the ridge direction can be detected. In each step of FIG. 12, the adjoining vector along the ridge direction is examined in steps 930, 940 and 960 so that it follows the vector on either side (front or back) of this adjoining vector. In addition, this improved example can obtain better results even for a long object moving diagonally, which may rarely appear, regardless of an increase in device cost.

Specific examples of the present invention have been described above in detail with reference to the accompanying drawings. However, the present invention is not limited to these detailed specific examples, and is defined by the appended claims. It will be apparent to those skilled in the art that various changes and modifications can be made without departing from the scope and spirit of the invention.

[0076]

As described above, according to the present invention,
Since the maximum correlation ridge on the correlation surface is detected and the motion vector indicating the motion of the image between the pair of input images of the input video signal is generated to perform the motion correction of the image, it is possible to reduce image defects. it can.

[Brief description of drawings]

FIG. 1 is a diagram showing a minimum value of a ridge (one-dimensional) of a correlation surface of an image used for explaining the present invention.

FIG. 2 is a schematic diagram showing an image defect due to a conventional zero motion vector.

FIG. 3 is a block diagram of a motion compensation television standard system conversion device used in the present invention.

4 is a block diagram illustrating details of a motion vector estimator according to the present invention for use in the apparatus of FIG.

5 is an explanatory diagram showing an operation of the motion vector estimator of FIG. 4. FIG.

FIG. 6 is a diagram showing detection of a ridge minimum value according to a second method of the present invention.

FIG. 7 is a diagram showing a first example of three threshold regions classified.

FIG. 8 is a diagram showing a second example of the threshold region.

FIG. 9 is a diagram showing a third example of the threshold region.

FIG. 10 is a schematic diagram illustrating requalification of a motion vector generated from a ridge minimum value according to the present invention.

11 is a block diagram showing a part of the motion vector reducer of FIG. 3. FIG.

FIG. 12 is a flowchart showing the process of FIG. 11.

[Explanation of symbols]

200 data stripper (correlation surface generation means) 210 vector estimator 310, 360, 400 minimum value detector (maximum correlation value detection means) 410 minimum value counter / comparator (ridge detection means) 470 vector generator (motion vector generation means) )

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor James Edward Burns United Kingdom Hampshire, Basingstoke, Old Basing, Hatchrain 81 (72) Inventor Martin Rex Dricot United Kingdom Hampshire, Basingstalk, Basing, Ringfield Cross 6 (72) Inventor Sima Love The Berthani United Kingdom Hampshire, Basingstalk, Hook, For Acres Copis 5

Claims (16)

[Claims] 1. A motion vector correction video signal processing device for generating a motion vector indicating a motion of an image between a pair of input images of an input video signal, wherein a plurality of search blocks are provided in one of the pair of input images. Means for generating a correlation surface having a series of correlation values indicating the correlation between the search block and the search area by comparing with the other search area of the pair of input images having the block of A means for detecting a value, a means for generating a motion vector according to the position of the maximum correlation value on the correlation surface, and a correlation value for the correlation surface within a total threshold of the maximum correlation values indicating a maximum correlation ridge And a ridge detecting means. 2. The ridge detecting means, wherein the correlation value included in the range of the total threshold value of the maximum correlation value is the first of the correlation values.
2. The apparatus of claim 1 including means for detecting if more than a predetermined number of. 3. The ridge detecting means detects whether a correlation value included in the range of the total threshold value of the maximum correlation value is smaller than a second predetermined number larger than the first predetermined number of the correlation values. The apparatus of claim 2 including means. 4. Means for generating an average motion vector according to the position of the two correlation values with the maximum interval on the correlation surface within the range of the total threshold value of the maximum correlation value in response to the detection of the maximum correlation ridge. The apparatus of claim 1 having. 5. The ridge detecting means detects the area of the correlation surface having a correlation value within the range of the total threshold value of the maximum correlation value, and the length of the area in the direction showing the movement of a vertical image. A means for detecting the ratio of the length of the area in the direction showing the movement of the horizontal image, and the ratio is compared with a first predetermined value to detect the maximum correlation ridge in the direction showing the movement of the horizontal image. 2. The apparatus of claim 1 including means and means for comparing the ratio to a second predetermined value to detect a maximum correlation ridge in the direction indicative of vertical image motion. 6. The apparatus of claim 1 including means for setting a ridge flag indicating a maximum correlation ridge. 7. A means for generating a ridge flag indicating whether said maximum correlation ridge is in a direction indicative of horizontal image motion or in said direction indicative of vertical image motion. Equipment. 8. A means for comparing a motion vector under inspection generated from a correlation surface in which the maximum correlation ridge is detected with a next motion vector generated from the pair of input images, and a motion vector under test with the next motion vector. Means for detecting whether the corresponding components of the motion vector are substantially the same, a means for detecting whether the next motion vector is generated from the correlation surface in which the maximum correlation ridge is detected, and the next motion vector Means for replacing the component of the motion vector under test in the ridge direction with the next motion vector if the maximum correlation ridge is not generated from the correlation surface detected. 9. The apparatus of claim 8, wherein the next motion vector is generated in the ridge direction from a correlation surface that is replaced with the motion vector under test. 10. The apparatus of claim 9, wherein the next motion vector is generated in the ridge direction from a correlation surface that is vertically adjacent to the motion vector under test. 11. The apparatus according to claim 9, wherein the next motion vector is generated from a correlation surface that is diagonally adjacent to the motion vector in the ridge direction. 12. The apparatus according to claim 1, wherein the total threshold value is an average value of correlation values of the correlation surface. 13. The apparatus of claim 1, wherein the correlation surface comprises a series of correlation values indicative of a difference in capacity between the search block and the search area. 14. The apparatus of claim 13, wherein the correlation value indicates a difference in brightness between the search block and the search area. 15. A motion compensated television standard converter comprising the device according to claim 1. 16. A method of processing a motion-compensated video signal for generating a motion vector indicating the motion of an image between a pair of input images of an input video signal, wherein one of the pair of input images has a plurality of blocks. Comparing with the other search area of the pair of input images, generating a correlation surface having a series of correlation values indicating the correlation between the search block and the search area, detecting the maximum correlation value of the correlation surface, Generating a motion vector in response to the position of the maximum correlation value on the correlation surface, and detecting the correlation value of the correlation surface within a total threshold of the maximum correlation value indicating a maximum correlation ridge.