JP4179089B2 - Motion estimation method for motion image interpolation and motion estimation device for motion image interpolation - Google Patents

Description

  The present invention relates to a motion estimation method for motion image interpolation and a motion estimation device for motion image interpolation, and in particular, when a motion image is displayed with high quality with natural motion, an image rate different from the image rate of an incoming motion image signal. In order to obtain sequentially scanned images from moving image signals and interlaced scanned images, frames, fields, and scanning lines that do not exist in the original moving image signal are formed by interpolating from the incoming moving image signal, and an effective frame of the moving image is formed. The present invention relates to a motion estimation method for motion image interpolation and a motion estimation device for motion image interpolation that change a rate and a scanning line structure.

  First, the time axis interpolation and scanning line structure conversion of moving images will be described. Since a normal NTSC television signal, which is a moving image signal, is 60 fields per second in interlace scanning, there is not much problem in the smoothness of the moving image. On the other hand, film images of 24 to 30 frames per second such as movies and computer graphics images created by sequential scanning at 30 frames per second have unnatural movement (judder and jerseyness).

  On the other hand, since PAL and SECAM television signals are 50 fields per second by interlace scanning, it is necessary to convert them to 60 fields per second for broadcasting and display on NTSC televisions. In addition, since the PAL method and the SECAM method have a low field frequency, flickering on the entire screen becomes a problem, and it is desired that the display is converted to 100 fields instead of 50 fields.

  As described above, when an incoming moving image signal is converted into a moving image signal having a different image rate, it is desired that the motion of the image is natural. Therefore, a moving image rate conversion method (moving image interpolation motion estimation method) that performs motion compensation that forms a frame or field that does not exist in time by motion compensation interpolation and changes the effective image rate with a smooth motion is a conventional method. (For example, refer to Patent Document 1).

  That is, in Patent Document 1, a block unit motion vector is searched based on an image signal, a pixel unit motion vector is generated based on the block unit motion vector, and an interpolation of the image signal generated based on the pixel unit motion vector is performed. A method of converting the number of frames of an image signal using an interpolation frame of an image signal generated from a frame or a frame signal of an image signal is disclosed.

  In addition, in order to display an image on a liquid crystal or plasma display, convert a moving image format, or obtain high efficiency by high-efficiency encoding, an interlaced scanned image is converted into a progressively scanned image. In this case, the scanning lines thinned out by the interlace scanning are interpolated from the scanning lines of the fields that are temporally forward and backward, and the upper and lower scanning lines, thereby forming image signals for sequential scanning.

  Next, the conventional motion estimation for moving image interpolation will be described. FIG. 9 is a block diagram showing an example of a conventional motion estimation apparatus for moving image interpolation. In the figure, the image signal coming from the image input terminal 1 is supplied to the motion compensator 2 and simultaneously supplied to the motion compensator 4 after being delayed by about one frame by the frame delay unit 3.

  The motion compensator 2 spatially moves the input image signal according to the temporary MV from the temporary motion vector (MV) setting unit 5. The motion compensator 4 spatially moves the image signal delayed by approximately one frame, which is the output of the frame delay unit 3, according to the temporary MV from the temporary MV setting unit 5. Here, the provisional MV supplied to both motion compensators 2 and 4 is common, but the moving direction is reversed between the motion compensator 2 and the motion compensator 4. The provisional MV setting unit 5 gives information on the provisional MV set in advance to the motion compensators 2 and 4 and the MV determination unit 8.

  The input image signal motion-compensated by the motion compensator 2 and the 1-frame delayed image signal motion-compensated in the reverse direction by the motion compensator 4 are subtracted by the subtractor 6 to become a differential signal. This difference signal is converted into an absolute value by the matching detector 7 and then added in units of a predetermined block to become a block matching value in the temporary MV. Here, the block is 8 × 8 pixels or the like, but it may be a finer block or may not be a square.

  The block matching value output from the matching detector 7 is input to the MV determination unit 8 together with the preset temporary MV output from the temporary MV setting unit 5 for each temporary MV, and the matching for each temporary MV in the block. The values are compared, and the temporary MV having the smallest value is finally determined as MV.

Japanese Patent Application Laid-Open No. 11-112939

  However, in the conventional motion image interpolation motion estimation, the image for which matching is to be determined changes as the motion estimation search range is expanded, so that the matching value is small for a motion vector in which the target for matching is a flat image. On the other hand, in a motion vector that is an image that changes spatially, the matching value tends to be large, and it is difficult to perform proper motion estimation. In particular, when there is a small or thin object in front of a flat image, there is a problem that a flat background has better matching than the object, and an interpolated image is formed with the background.

  In image coding and the like, since the block of the frame to be encoded is the same regardless of the temporary MV, such a problem does not occur. However, in interpolation, both images to be matched change depending on the temporary MV. Cause problems.

  The present invention has been made in view of the above points, and by obtaining the spatial activity of the corresponding part of the interpolation reference image with each temporary motion vector of motion estimation, and comparing the comparison after normalizing the matching between the images with the activity. Another object of the present invention is to provide a motion estimation method for motion image interpolation and a motion estimation device for motion image interpolation in which an appropriate motion vector is obtained even for an image with a flat background.

  In order to achieve the above object, the motion estimation method for motion image interpolation according to the present invention performs motion compensation on an incoming motion image for each predetermined region to form a motion vector for forming a time axis interpolation image. In the moving image interpolation estimation method to be generated, a first step of setting a plurality of temporary motion vectors in a search range in each predetermined region and a plurality of reference images used for motion compensation interpolation in each of the plurality of temporary motion vectors A second step of calculating a matching value between the predetermined regions in the second step, and a third step of calculating a spatial activity of the predetermined region for performing motion compensation in each of the reference images used for motion compensation interpolation in each of a plurality of temporary motion vectors. A fourth step of normalizing the matching value based on the spatial activity of the predetermined area to obtain a normalized matching value, Each provisional motion vectors, by comparing the normalized matching values, characterized in that it comprises a fifth step of the motion vector of the provisional motion vectors to provide a minimum value in a predetermined region.

  In order to achieve the above object, the moving image interpolation estimating apparatus according to the present invention performs motion compensation on an incoming moving image for each predetermined area to form a time axis interpolated image. In the estimation apparatus for moving image interpolation that generates a plurality of temporary motion vector setting means for setting a plurality of temporary motion vectors in a search range in each predetermined region, and a plurality of references used for motion compensation interpolation in each of the plurality of temporary motion vectors A matching value calculation means for calculating a matching value between predetermined regions between images, and a spatial activity of a predetermined region for performing motion compensation in each reference image used for motion compensation interpolation in each of a plurality of temporary motion vectors. Spatial activity calculation means and a normalized map that obtains a normalized matching value by normalizing the matching value based on the spatial activity of a predetermined region. And a motion vector determination unit that compares the normalized matching value for each temporary motion vector in a predetermined region and outputs a temporary motion vector that gives a minimum value as a motion vector in the predetermined region. It is what.

  In the estimation method and apparatus for moving picture interpolation according to the present invention, the spatial activity of a predetermined area of an interpolated reference image is obtained by each temporary motion vector of motion estimation in motion compensated interpolation of the moving picture, and matching between the predetermined areas between the reference pictures is performed. After normalizing the value with spatial activity, each temporary motion vector is compared and the temporary motion vector that gives the minimum value is output as the motion vector in the specified area, so the spatial activity is small and normalized matching It is difficult to select a flat part of a predetermined area of the interpolation reference image having a large value as a motion vector, while a part having a large spatial change in the predetermined area of the interpolation reference image having a large spatial activity and a small normalized matching value. It can be easily selected as a motion vector.

  According to the present invention, when the interpolated reference image is a flat portion due to the temporary motion vector, the spatial activity is small, so the normalized matching value is large, and when the spatial change of the reference portion is large, the spatial activity is Since the normalized matching value is small, the normalized matching value can appropriately represent the difference in matching due to the motion of the image. For each temporary motion vector, the normalized matching value is compared, and the minimum Since the temporary motion vector that gives a value is output as a motion vector in a predetermined region, more appropriate motion estimation becomes possible.

  Therefore, according to the present invention, if there is a small but thin object in front of a flat image, the matching value between the flat backgrounds is larger than that of the object because the matching value is increased by normalization, so that it is difficult to select the background. This avoids the problem of forming an interpolated image. As described above, according to the present invention, since a motion vector corresponding to the actual motion of an image is obtained, an appropriate interpolated image and interpolated scanning line are formed.

  Next, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram of a first embodiment of a motion estimation apparatus for moving picture interpolation according to the present invention. In the figure, the same components as those in FIG. 9 are denoted by the same reference numerals. In the first embodiment shown in FIG. 1, activity detectors 11 and 12 and a matching normalizer 13 are added as compared with the conventional apparatus shown in FIG.

  The operation of this embodiment will be described by taking as an example the case of converting a progressive scan image signal of 30 frames per second (30 fps) into a progressive scan image signal of 60 frames per second (60 fps). In FIG. 1, a 30 fps progressively scanned image signal coming from the image input terminal 1 is supplied to the motion compensator 2 and simultaneously supplied to the motion compensator 4 after being delayed by about one frame by the frame delay unit 3. Is done.

  The motion compensator 2 spatially moves the input image signal according to the temporary MV from the temporary motion vector (MV) setting unit 5. The motion compensator 4 spatially moves the image signal delayed by approximately one frame, which is the output of the frame delay unit 3, according to the temporary MV from the temporary MV setting unit 5. Here, the provisional MV supplied to both motion compensators 2 and 4 is common, but the moving direction is reversed between the motion compensator 2 and the motion compensator 4. The temporary MV setting unit 5 gives the temporary MV within the preset search range to the motion compensators 2 and 4 and the MV determination unit 8 in the motion estimation of each block.

  The input image signal motion-compensated by the motion compensator 2 and the 1-frame delayed image signal motion-compensated in the reverse direction by the motion compensator 4 are subtracted by the subtractor 6 to become a differential signal. This difference signal is converted into an absolute value by the matching detector 7 and then added in units of a predetermined block to become a block matching value in the temporary MV. Each matching value is given to a matching normalizer 13.

  On the other hand, the respective image signals subjected to motion compensation in the motion compensators 2 and 4 are also supplied to the activity detectors 11 and 12. The activity detectors 11 and 12 obtain the spatial activity (activity) of the pixels in the block subjected to motion compensation in the motion compensators 2 and 4. This degree of activity is an absolute value of a standard value difference or a neighborhood difference of pixel values in a block.

  As a method of taking the neighborhood difference, there are a difference between adjacent pixels (one pixel difference) as shown in FIG. 2A, a difference between two pixels apart as shown in FIG. 2B (two pixel difference), and the like. Conceivable. Further, since this process must be obtained for each provisional MV, it may be simplified to reduce the processing amount, and thinning as shown in FIG.

  Furthermore, instead of a simple difference, a high frequency component may be extracted by an HPF (High Pass Filter). Any difference value or high frequency component signal is converted into an absolute value and added in the block, and the degree of activity in the provisional MV is obtained.

  The matching normalizer 13 calculates the activity Ab of the input image given from the activity detector 11 and the activity Af of the one-frame delayed image given from the activity detector 12 as shown in the equation (1). Addition to obtain the spatial activity A ′ of the temporary MV.

A ′ = Af + Ab (1)
Here, the spatial activity A as a denominator of normalization is set to a value of A ′ when the spatial activity A ′ of the temporary MV is larger than the lower limit AL, but the spatial activity A ′ of the temporary MV is a lower limit. When it is below AL, it is limited by the lower limit AL. Alternatively, the space activity A may be obtained by adding the lower limit AL to the space activity A ′ of the temporary MV. This is to prevent the normalization process from becoming unstable when the activity is low and the value of the spatial activity A ′ of the temporary MV is about the noise component. The above can be expressed as follows.

A = AL (when AL ≥ A ')
A = A '(when AL <A')
Or
A = A '+ AL

Subsequently, the matching normalizer 13 normalizes the matching value C with the spatial activity A, and obtains a normalized matching value NC by the following equation.
NC = C / A (2)
The normalized matching value NC is input to the MV determiner 8 for each temporary MV, the matching values for each temporary MV in the block are compared, and the temporary MV that has the smallest value is finally determined as MV. .

  Next, frame interpolation using motion estimation for moving image interpolation according to the first embodiment of the present invention will be described. An example of the configuration is shown in FIG. Here, the frame interpolation is to convert a progressively scanned image signal of 30 frames per second into a progressively scanned image signal of 60 frames per second.

  The state of this frame interpolation is shown in FIG. In FIG. 4, one circle is one pixel, and the vertical continuation indicates a part of the scanning line or a pixel row in the vertical direction. Also, the pixels indicated by the solid line are incoming moving image signals at 30 frames per second, and the pixels indicated by the broken line are interpolated image signals. The arrow is motion compensation interpolation.

  In FIG. 3, a moving image signal of 30 frames per second (30 fps) progressive scanning coming from the image input terminal 1 is directly supplied to the motion estimator 31 and the motion compensator 32, while the frame delay unit 3 is approximately 1 After being delayed by the number of frames, it is supplied to the motion estimator 31 and the motion compensator 33. The motion estimator 31 estimates the motion vector MV with the configuration shown in FIG.

  The motion compensator 32 spatially moves the input image signal according to the MV from the motion estimator 31. The motion compensator 33 spatially moves the 1-frame delayed image signal according to the MV from the motion estimator 31. The image signal motion-compensated by the motion compensator 32 and the image signal motion-compensated in the reverse direction by the motion compensator 33 are added by an adder 34, and then divided by 2 to become an interpolated image signal. Given to.

  The frame buffer 36 stores the image signal for one frame output from the adder 34, and the stored image signal is read out at a double speed. Similarly, the incoming moving image signal is also stored in the frame buffer 35 and read out at a double speed. In the time of one frame of the incoming moving image signal, both the incoming moving image signal held in the frame buffer 35 and the interpolated image signal held in the frame buffer 36 are alternately read out and doubled. It is output as a moving image signal having an image rate of 60 fps.

  FIG. 5 shows a state of frame interpolation in the embodiment. FIG. 5 is similar to FIG. 4 except that the shade of each circle is a pixel value, and the image moves gently with time. 5A is suitable for motion in the case of frame interpolation using a correct motion vector, but FIG. 5B is not suitable for motion in the case of frame interpolation using an incorrect motion vector. In the conventional method, since the determination is made simply by matching, even if the movement shown in FIG. 5 (a) is correct, a slight error occurs due to a slight shift of the moving amount and aliasing distortion component.

  On the other hand, since FIG. 5B is a completely flat portion, the matching error is small. In such a case, FIG. 5B is selected as the final MV in the conventional method of determining based on the matching error. On the other hand, the spatial activity described above is much larger in FIG. 5A, and the matching value normalized by the spatial activity as in the present embodiment is that in FIG. 5A. According to the present embodiment, the correct MV in FIG. 5A is selected.

  In the above embodiment, since the conversion is from 30 fps to 60 fps, the interpolation image time is the center of the incoming image, and the added image is obtained by equal addition. However, for example, in the conversion from 50 fps to 60 fps, the interpolated image is not necessarily centered in time, so the addition may be linear interpolation or the like according to the time distance. In this case, the relationship between the movement amounts of the motion vectors also differs depending on the time relationship.

  Next, motion estimation for moving image interpolation according to the second embodiment of the present invention will be described. FIG. 6 shows a block diagram of a second embodiment of a motion estimation apparatus for moving image interpolation according to the present invention. In the figure, the same components as those in FIG. In the second embodiment of FIG. 6, activity memories 23 and 24 are added to the first embodiment of FIG. 1, and the input images and operation speeds of the activity detectors 21 and 22 are the same. Different.

  The second embodiment of FIG. 6 is similar to the first embodiment of FIG. 1 in that a progressive scan image signal of 30 frames per second (30 fps) is converted into a progressive scan image signal of 60 frames per second (60 fps). It is a case to convert to. In the first embodiment, the degree of activity is obtained for each provisional MV with respect to the motion compensated block. However, in the second embodiment, the degree of activity is previously set for the incoming image and the one-frame delayed image. The degree of activity corresponding to the motion compensation block is read from the degree-of-activity memories 23 and 24 according to the provisional MV.

  In FIG. 6, the 30 fps progressive scanning signal coming from the image input terminal 1 is supplied to the motion compensator 2 and the activity detector 21 and simultaneously delayed by about one frame by the frame delay unit 3. It is supplied to the motion compensator 4 and the activity detector 22, respectively. The operations of the motion compensators 2 and 4, the subtractor 6, and the matching detector 7 are the same as in FIG. 1, and a matching value is obtained for each temporary MV.

  On the other hand, the operations of the activity detectors 21 and 22 are basically the same as those of the activity detectors 11 and 12 in FIG. 1, but the activity is not calculated for each provisional MV, The activity is obtained for all the pixels while moving the block in units of one pixel. The obtained activity is given to the activity memories 23 and 24 as information for each pixel.

  The activity memories 23 and 24 hold the supplied activity information for each pixel, read the activity of the corresponding motion compensation block from the processing block position and the temporary MV from the temporary MV setting unit 5, and perform matching normalization. To the generator 13. The operations of the matching normalizer 13 and the MV determiner 8 are the same as those in the first embodiment shown in FIG.

  In the case of the second embodiment of FIG. 6, processing is required for the number of pixels of one frame in the interpolation of one frame, but when it is obtained for each temporary MV as in the first embodiment, ) × (temporary MV number) processing is required, so if the number of pixels in one block is less than the temporary MV number, the processing amount of the second embodiment is smaller. For example, if the block is 8 × 8 pixels, the provisional MV is 15 pixels × 15 pixels (horizontal ± 7 pixels × vertical ± 7 pixels), and the accuracy is 1 pixel, the ratio of the processing amount is (second embodiment) / The first embodiment is 64/225. That is, in this case, the processing amount of the second embodiment is smaller than that of the first embodiment.

  Next, motion estimation for moving image interpolation according to the third embodiment of the present invention will be described. The configuration diagram is the same as that of the first embodiment shown in FIG. This embodiment is different from the first embodiment in the target image signal and the specific processing contents of each part. In the third embodiment, an interlaced scanning image signal of 60 fields per second is received every second. It is for conversion into a 60 frame progressive (progressive) scanned image signal.

  In this case, the interpolation target is a scanning line that is missing in the interlaced scanning with respect to the sequential scanning, and is temporally interpolated from the preceding and succeeding fields at the same position as the existing field. Therefore, in the third embodiment, the incoming image signal is an interlaced scanning image signal, which is processed by the motion compensators 2 and 4, the activity detectors 11 and 12, the subtractor 6, and the matching detector 7 in FIG. The signal to be transmitted is in one field of interlace scanning.

  On the other hand, since the frame delay unit 3 needs to delay by two fields, it performs the frame delay as it is. The operations of the matching normalizer 13 and the MV determiner 8 are the same as those in the first embodiment shown in FIG.

  Next, scanning line interpolation (conversion to sequential scanning) using motion estimation for moving image interpolation according to a third embodiment will be described. An example of the configuration is shown in FIG. The interlaced scanning image signal of 60 fields per second is converted into a progressive scanning image signal of 60 frames per second, and the state of this scanning line interpolation is shown in FIG. In FIG. 8, one circle is a scanning line, and a vertical continuation indicates a field. Also, the pixels indicated by solid lines are incoming interlaced scanning image signals of 60 fields per second, and the pixels indicated by broken lines are interpolated scanning lines. The arrow is motion compensation interpolation.

  In FIG. 7, an interlaced scanned image signal of 60 fields per second coming from the image input terminal 41 is supplied to a motion estimator 42 and a motion compensator 43, respectively, while being shortened by cascaded field delay units 44 and 45. The signals are delayed by two fields and supplied to the motion estimator 42 and the motion compensator 46, respectively. The interlaced scanned image signal of 60 fields per second delayed by approximately one field by the field delay unit 44 is supplied to the intra-field interpolator 47 and the line buffer 48, respectively.

  The motion estimator 12 supplies the motion compensators 43 and 46 with the motion vector (MV) obtained by the motion estimation for moving image interpolation according to the third embodiment. The motion compensator 43 spatially moves the input image signal according to the MV. On the other hand, the motion compensator 46 spatially moves the 1-frame (2-field) delayed image signal according to the MV.

  The MV supplied to both motion compensators 43 and 46 is common, but the direction in which the image signal is moved is reversed between the motion compensator 43 and the motion compensator 46. The image signal motion-compensated by the motion compensator 43 and the image signal motion-compensated in the reverse direction by the motion compensator 46 are added by an adder 49 and divided by 2 to become an interpolated image signal. 50.

  On the other hand, the intra-field interpolator 47 forms an interpolated image signal from the scanning lines located above and below the interpolated scanning line on the screen of the same field, and supplies the interpolated signal to the multiplier 51. The first interpolated image signal interpolated from the preceding and following fields of the interpolated scanning line multiplied by the first multiplication coefficient by the multiplier 50 and the interpolated scanning line multiplied by the second multiplication coefficient by the multiplier 51. The second interpolated image signal interpolated from the upper and lower scanning lines is supplied to the adder 52 and added to be supplied to the line buffer 53 as the third interpolated image signal of the final interpolated scanning line. The sum of the first multiplication coefficient and the second multiplication coefficient is set to be 1.

  The line buffer 53 stores the third interpolated image signal for one line, and the stored third interpolated image signal is read out at a double speed. Similarly, the moving image signal delayed by approximately one field, which is the output of the field delay device 44, is also stored in the line buffer 48 for one line, and the stored moving image signal is read out at a double speed.

  The incoming moving image signal held in the line buffer 48 and the interpolated image signal held in the line buffer 53 are alternately read out by the switch 54 in one line time of the moving image signal delayed by approximately one field. By alternately selecting, an image signal of a frame having twice the density of scanning lines is formed. This image signal is output from the image output terminal 55 as a progressively scanned image signal of 60 frames per second.

  Note that the functions of the above-described apparatus may be realized by a computer by a program. This program may be read from a recording medium and loaded into a computer, or may be transmitted via a communication network and loaded into a computer.

It is a block diagram of 1st, 3rd embodiment of the motion estimation apparatus for image interpolation of this invention. It is a figure which shows the mode of the spatial activity detection of embodiment of this invention. It is a block diagram of the frame interpolation apparatus using the motion estimation for the moving image interpolation which is the 1st Embodiment of this invention. It is a figure which shows the mode of the motion compensation frame interpolation of embodiment of this invention. It is a figure which shows the mode of the motion compensation frame interpolation of embodiment of this invention. It is a block diagram of 2nd Embodiment of the motion estimation apparatus for image interpolation of this invention. It is a block diagram of the scanning line interpolation apparatus using the motion estimation for the moving image interpolation which is the 3rd Embodiment of this invention. It is a figure which shows the mode of the motion compensation scanning line interpolation of the 3rd Embodiment of this invention. It is a block diagram of an example of the conventional motion estimation apparatus for image interpolation.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1, 41 Image input terminal 2, 4, 32, 33, 43, 46 Motion compensator 3 Frame delay device 5 Temporary MV setting device 6 Subtractor 7 Matching detector 8 MV determination device 9 MV output terminal 11, 12, 21, 22 Activity detector 13 Matching normalizer 23, 24 Activity memory 31, 42 Motion estimator 38, 55 Image output terminal 34, 49, 52 Adder 35, 36 Frame buffer 44, 45 Field delay 47 Field interpolation 48, 53 line buffer 50, 51 multiplier 54 switch

Claims (2)

In the moving image interpolation estimation method for generating a motion vector for forming a time-axis interpolated image by performing motion compensation for each predetermined region for an incoming moving image,
A first step of setting a plurality of temporary motion vectors in a search range in each predetermined region;
A second step of calculating a matching value between predetermined regions between a plurality of reference images used for motion compensation interpolation in each of the plurality of temporary motion vectors;
A third step of calculating a spatial activity of a predetermined region for performing motion compensation in each of the reference images used for motion compensation interpolation in each of the plurality of temporary motion vectors;
A fourth step of normalizing the matching value based on the spatial activity of the predetermined region to obtain a normalized matching value;
A fifth step of comparing the normalized matching values for each of the temporary motion vectors in the predetermined region and setting a temporary motion vector that gives a minimum value as the motion vector in the predetermined region. An estimation method for moving image interpolation. In the moving image interpolation estimating device for generating a motion vector for forming a time-axis interpolated image by performing motion compensation for each predetermined region with respect to an incoming moving image,
Provisional motion vector setting means for setting a plurality of provisional motion vectors in the search range in each predetermined region;
In each of the plurality of provisional motion vectors, a matching value calculation unit that calculates a matching value between predetermined regions between a plurality of reference images used for motion compensation interpolation;
In each of the plurality of provisional motion vectors, spatial activity calculation means for calculating the spatial activity of a predetermined region for performing motion compensation in each reference image used for motion compensation interpolation;
Normalized matching value calculation means for normalizing the matching value based on the spatial activity of the predetermined region to obtain a normalized matching value;
Motion vector determination means for comparing the normalized matching value for each temporary motion vector in the predetermined region and outputting a temporary motion vector that gives a minimum value as the motion vector in the predetermined region. A feature estimation apparatus for video interpolation.

Images