JP3366059B2 - Video signal processing device - Google Patents

Description

Detailed Description of the Invention

[0001]

FIELD OF THE INVENTION This invention relates to video signal processing.

[0002]

Motion compensated video signal processing is used in video processing applications such as television standard conversion and video to film conversion.

A previously proposed motion-compensated video processing device is described in British Published Patent Application GB-A-2231749, in which temporally adjacent pictures (fields or frames) of an input digital video signal are described. The pairs are processed to generate a corresponding set of motion vectors.

Since this process is performed on discontinuous blocks of the image, each motion vector represents the inter-image motion of the contents of the associated block.

After further processing, the set of motion vectors is provided to a motion compensation interpolator, which interpolates the output image from the pair of input images, taking into account the motion between the input images.

In the above device, the motion vector is generated as follows. The first stage compares each search block in the input image that temporally precedes the requested output image with a plurality of such blocks in the search area of the input image that temporally follows the output image. That is.

For each such comparison, a correlation test is performed which involves calculating the absolute intensity difference between pixels at corresponding positions of the respective blocks in the input image which precede or follow in time.

These absolute luminance differences are summed together to produce a correlation value representing the correlation between the two blocks being compared.

The block in the search area of the temporally succeeding image having the lowest correlation value is retrieved and correlated with the search block of the temporally preceding image, and the image from that search block to that block in the search area is extracted. A motion vector is generated that represents the motion of the.

[0010]

FIG. 1 of the accompanying drawings shows that an output image (eg an output field) is closer in time to a subsequent input image (input frame) than to a preceding input image. Indicates possible problems with the.

In FIG. 1, the output field 1 is a frame n of the input video signal (frame preceding in time).
Is closer in time to frame n + 1 of the input video signal (the one that follows in time).

In order to generate the motion vector associated with the search block 2 of frame n, the search block 2 has a plurality of blocks 5, 5 in the search area of frame n + 1.
A correlation test is performed which is compared with 6, 7 ,.

Since it can be seen that block 5 of frame n + 1 has the best correlation with search block 2 of frame n, a motion vector 9 representing the motion of the image from search block 2 to block 5 is generated.

This motion vector 9 is used in the search block 2
Associated with search block 2 used in the interpolation of block 3 of the output field co-located with.

However, the motion of the image, represented by motion vector 9, passes from search block 2 of frame n, through its position in block 4 of the output image, to frame n +.
It is associated with part of the movement of the image to block 5 of 1.

Therefore, the vector 9 is more related to the output field block 4 than to the output field block 3.

One previously proposed solution to this problem (where the generated motion vector is more related to block 4 than to block 3 to be interpolated) is a large number of solutions according to a predefined order of priority. Is to select a motion vector to be used for interpolation of one block of the output image from among the motion vector candidates of.

The motion vector candidates are: (a) a motion vector representing the motion of a block at a position corresponding to an interpolation block in an input frame preceding in time; (b) an interpolation block in an input frame preceding in time. (C) a motion vector generated from a peripheral block at a position corresponding to
It includes three types of common motion vectors generated from blocks of an input frame that are temporally preceding.

This process has been referred to as motion vector selection. The process then allows "borrowing" a motion vector more suitable for a block of the output image from another block.

[0020]

[Means and Actions for Solving the Problems] However,
The improved system will increase the probability of the most appropriate motion vector being generated directly for each block of the output image.

The present invention provides a video signal processing apparatus for interpolating an output image of an output video signal from a corresponding pair of temporally adjacent input images of an input video signal, the video signal processing apparatus comprising: Means for detecting which input image in the pair of input images adjacent to each other is temporally closer to the output image; A pair of temporally adjacent inputs by comparing a search block within one of them with respective search areas that each include a plurality of blocks within another of a pair of temporally adjacent input images. A motion vector generator for generating a set of motion vectors representing motion between images; motion compensation for interpolating an output image from a pair of temporally adjacent input images using the set of motion vectors Within And a vessel.

In the present invention, the search block used to derive the motion vector is taken from the input image,
It is taken from an input image that is closer in time to the required output image, rather than from a predetermined one of the input images.

By using the search blocks in the closer input image to generate motion vectors for interpolation of the corresponding blocks in the output image, the resulting motion vector is the image of those blocks in the output image. Is considered to be more related to the movement of.

The pair of temporally adjacent input images from which the output image is interpolated includes an input image that temporally precedes the output image and an input image that temporally follows the output image. Is preferred.

In the preferred embodiment, the apparatus provides a first clock signal including a clock pulse providing the output period of each input image of the input video signal and a clock pulse providing the output period of each output image of the output video signal. Second including
A means for receiving a clock signal, the detecting means comprising:
A counter that operates to start counting at a predetermined rate when the clock pulse of the first clock signal is received and to stop counting when the clock pulse of the second clock signal is received; It includes means for comparing the count value of the counter with a predetermined threshold value to determine whether or not the output image is close to the input image temporally preceding or temporally following. In an improved and simple embodiment, the device comprises means for receiving the input image of the input video signal in synchronism with the clock pulses of the first clock signal, the counter including a video line of the received input image of the input video signal. Can be counted.

In the preferred embodiment, the device comprises:
In order to control the reading of pairs of input images that are temporally adjacent to each other from the plurality of image storages for storing the input image of the input video signal and the plurality of image storages used for interpolation of the output image, Means for generating a read selection control signal.

The motion vector generator compares a search block within the input image supplied to its first input with a respective search area, which in the input image supplied to its second input, each includes a plurality of blocks. In an embodiment with means for this, the device preferably comprises a switch for selectively transforming the input image supplied to the first and second inputs of the motion vector generator.

In this way, the motion vector generator itself does not have to change to change the direction of motion vector generation. Only the input image supplied to the motion vector generator needs to be transformed.

Switches can be used to transform the actual image data supplied from the image store to the motion vector generator.

However, a switch (eg, a mechanical or electronic switch) that switches with a relatively low frequency control signal rather than a relatively high frequency data signal.
It is convenient to make and use.

Thus, in the preferred embodiment, the motion compensated interpolator is responsive to the temporal separation of the output image and one of the pair of temporally adjacent input images that are closer in time to the output image ( (Because the same input of the interpolator is always supplied with the closer input image), the switch causes either input image in the temporally adjacent input image pair to become more temporally related to the output image. In response to detecting the closeness, the read selection control signal is operated to be exchanged.

[0032]

[0033]

[0034]

The interpolator could be arranged to behave differently depending on whether its set of motion vectors was generated in the forward direction or in the opposite direction in time,
It would be advantageous to eliminate the need to change the behavior of the interpolator in response to changes in the order of the input images used to generate the set of motion vectors.

Therefore, in another preferred embodiment in which the input images supplied to the motion vector generator are exchanged,
This apparatus provides means for selectively inverting the motion vector of the set in response to detecting which one of the temporally adjacent input images is closer to the output image in time. Including.

In other words, the interpolator is set up to receive the motion vector generated by comparing the search area of the temporally succeeding input image with the search block of the temporally preceding input image. Then any motion vector generated in the opposite direction (ie generated by comparing the search block of the temporally subsequent input image with the search area of the temporally preceding input image) is inverted. .

Although the present invention is suitable for use with various input and output video standards, in the preferred embodiment the input video signal comprises a progressively scanned 60 Hz video signal and the output video signal is interlaced. 50 Hz video signal.

Viewed from a second aspect, the present invention provides a method for generating a motion vector used to interpolate an output image of an output video signal from a corresponding pair of temporally adjacent input images of an input video signal. And a step of detecting which one of a pair of input images temporally adjacent to each other is temporally closer to the output image, and a step temporally closer to or closer to the output image. Comparing a search block within one of the pair of input images with a respective search area including a plurality of blocks within another of the pair of temporally adjacent input images. Provide a vector generation method.

[0040]

First, referring to FIG. 2, one frame 1125
A previously proposed motion-compensated television for converting an input interlaced high definition video signal (HDVS) having 60 fields per line per second into an output interlaced video signal 40 having 625 lines per frame and 50 fields per second. The standard conversion device will be described.

The input video signal 10 is first supplied to a progressive scan converter 11, where the input video field is converted into a video frame at the same rate (60 Hz) as the input video field.

These frames correspond to the down converter 1
The number of lines in each frame is reduced to 625
The number of lines in the frame of the output video signal 40 is set.

The down-converted input frame is sent to the time base modifier 14, and in the same manner, a 60 Hz clock signal 32 locked to the field frequency of the input video signal 10 as an input signal and an output video signal 40.
50Hz locked to the required field frequency of
The lock signal 34 is received.

The time base modifier 14 determines the temporal position of each field of the output video signal and uses two of the downconverted frames of the input video signal 10 to interpolate that field of the output video signal. Choose one.

This down-converted input frame selected by the time base changer has two outputs 16 (for frames temporally preceding the required output field) and one temporally (for the required output field). On output 17 (for subsequent frames).

The third output 18 of the time base converter carries control information including the temporal position of the output field with respect to the two selected input frames.

2 selected by the time base changer 14
Video signals corresponding to the two down-converted input frames are provided to respective inputs of motion processor 20.

The motion processor comprises two substantially identical processing sections. One for the even output field,
The other one is for odd output fields.

The two down-converted input frames selected by the time base converter 14 are 50H
Under control of a signal (not shown) derived from the z clock signal 34, it is directed to the even field or odd field processing section as appropriate by the means of the changeover switch 21.

In the proper section (even or odd) of the motion processor 20, the down-converted input frame is first fed directly to the block matcher 22 and the correlation representing the spatial correlation between the blocks of the two frames. Calculate the face.

These correlation surfaces are the motion vector estimator 24.
To generate a set of motion vectors, which are supplied to the motion vector reducer 26.

The motion vector estimator performs a confirmation test on each generated motion vector to set if the motion vector is significantly above a typical noise level and represents the result of the confirmation test. Associate a confirmation flag with each motion vector.

This verification test itself includes a "threshold test" and a "Rings test", together with other features of the device shown in FIG. 2 and British published patent application GB-A-22317.
No. 49 describes in detail.

The motion vector reducer 26 operates to reduce the possible motion vector selection for each pixel in each block before the motion vector is provided to the motion vector selector 28.

As part of its function, the motion vector reducer 26 does not consider the position of the blocks of the input image used to obtain those motion vectors, but passes the "good" motion vectors (ie, the validation test). The appearance frequency of each motion vector) is counted.

The good motion vectors are ranked in order of decreasing frequency of occurrence. The three most common "good" motion vectors that differ significantly from each other are classified as "global motion vectors".

The three motion vectors that pass the verification test are selected for each block and together with the zero motion vector, the motion vector selector 28 for further processing.
Is supplied to.

These three motion vectors are selected from the motion vector generated from the block, the vector generated from the peripheral blocks, and the global motion vector in accordance with a predetermined priority order.

The motion vector selector 28 also receives as its input the two down-converted input frames selected by the time base modifier 14. These input frames are the same as those used by the motion vector estimator 24 to calculate the motion vector and have been delayed by the system delay compensator 30. The motion vector selector 28 operates using the two input frames input and assigning one motion vector per pixel in the output field.

The assigned motion vector is selected from among the four motion vectors supplied from the motion vector reducer 26 for the block (three motion vectors passing the confirmation test and zero motion vector).

The irregularity in selecting a motion vector by the motion vector selector 28 is eliminated by the motion vector post-processor 36. The motion vector selector 28 and motion vector post-processor 36 are described in detail in British Published Patent Application GB? A? 2231749.

The processed motion vector is fed to an interpolator 38, which interpolator 38 selects the proper even or odd number selected by the time base modifier (again properly delayed by the system delay compensator 30). A pair of down-converted input frames is also received.

Under control of the motion vector, the interpolator 38
Interpolates the output fields from the two downconverted input frames selected by the time base modifier, taking into account any image motion between the two frames.

The output 40 of the interpolator 38 is one frame 62.
It is an interlaced video signal having 5 lines and 50 fields per second.

FIG. 3 illustrates how a pair of temporally adjacent downconverted frames of the input video signal 10 are selected by the time base modifier 14 for use in the corresponding odd or even fields of the output video signal 40. Is represented.

The upper row of FIG. 3 represents a downconverted frame of the input video signal 10. As mentioned above, the input video signal has a field rate of 60 Hz and the output of the progressive scan converter 11 is a series of video frames of the same frequency as the field frequency of the input video signal.

To clarify FIG. 3, and thus to identify the down-converted input frame used to interpolate each output field, each down-converted input frame is associated with a corresponding number of input fields. Represented by.

For example, the down-converted input frame 42 is represented by the number "2+", which means that it originates from the second frame odd field of the input video signal 10 (before progressive scan conversion).

Similarly, another down-converted input frame 43 is represented by the number "4", indicating that it originates from the even field of the fourth frame of the input video signal.

The output video signal has a field frequency of 50H.
Having z, the temporal separation of adjacent fields of the output video signal is greater than the temporal separation of adjacent down-converted input frames of the input video signal.

The time base converter 14 operates to select the two downconverted input frames that are closest to the temporal position of the output field used in interpolating the output field.

For example, the (even) output field 44 is
A motion vector representing the motion between the two down-converted frames (represented by numbers 1+ and 2) closest in time to the output field is used to represent these two frames (represented by numbers 1+ and 2). Interpolated from the

The numerical value on the bottom line of FIG. 3 indicates the interpolation position for each output field. This interpolation position is
The ratio of the length of one input frame to the time separation between the start position of the output field (represented by the left-hand end of each block in the central section of FIG. 3) and the start position of the input frame preceding in time. It is expressed as.

For example, the start of the output field 44 is 6/1 of one input frame period after the start of the input frame 1 ^+.
0 happens. The output field 44 therefore has an interpolation position of 0.6.

In the conventional motion-compensated television standard conversion apparatus described above with reference to FIG. 2, motion vector estimation is required to be a search area containing blocks of the input frame that temporally follow the output field. This is done by comparing the search blocks of the input frame that precede the output field in time.

In this embodiment, the required output field is inverted when it is closer in time to the temporally subsequent input frame, so that the search block of the temporally subsequent input frame is temporally earlier. The search area is compared with a plurality of blocks of the input frame.

This can be accomplished by adding three features to the device of FIG. (A) means for detecting whether or not the output field is closer in time to the preceding or succeeding input frame, (b) means for changing the order of the input frames used for motion vector estimation, (c) used for interpolation of the output field A means of inverting the calculated motion vector "in the opposite direction" before being exposed.

Means for detecting whether the output field is closer in time to the preceding or following input frame will be described with reference to FIG.

FIG. 4 is a block diagram of the time reference modifier 14 ', which can be used in place of the time reference modifier 14 of FIG.

The time base modifier 14 'includes seven field accumulators (FS) 50 in which 60 Hz downconverted under control of the accumulator write select signal.
A video frame is written, from which two selection frames used for interpolation of the output field are read out under the control of the respective storage read selection signals.

The accumulator write selection signal is modulo 7
Generated by the frame counter 52, which is clocked by the 60 Hz clock signal 32.

Each time, the modulo 7 frame counter 52 receives the clock pulse of the 60 Hz clock signal 32 and its output is the sequence 0, 1, 2, 3, 4,
Progress in 5, 6, 0, 1, 2, ....

The output of the modulo 7 frame counter 52 is used as a store write select signal so that continuously received 60 Hz video frames are written to the corresponding store of field store 50 according to a periodic sequence.

The accumulator write selection signal is sampled by the 50 Hz sampling circuit 54. The sampling circuit is clocked with a 50 Hz output field clock signal 34.

50 Hz output field clock signal 34
When each clock pulse in the is received by the 50 Hz sampling 54, the current value of the store write select signal is sampled and sent to the two field delay unit 56.

The sampled output of modulo 7 frame counter 52 represents the most recent input frame at the time of the sampling.

The two field delay unit 56 outputs the output video signal 40 for two field periods (ie, two clock pulses of the 50 Hz output field clock signal 34).
Delays the output of the 50 Hz sampling circuit 54.

Since a two field delay is used, the same field store 50 will not be accessed simultaneously for reading and writing.

The output of the 2-field delay unit 56 is
Identify the input frame that temporally precedes the requested output field. One is added to the output (in a modulo 7 operation) by adder 58 to generate an identification signal for the input that follows in time.

The output of the 2-field delay unit 56 and the output of the adder 58 together form the accumulator read select signal.

To generate the interpolated position, the modulo 6
The 25 line counter 60 counts the video lines of the received down-converted 60 Hz video frame.

The output of the modulo 625 line counter 60 is sampled by the 50 Hz sampling circuit 62 in synchronization with the sampling of the storage device write selection signal by the 50 Hz sampling circuit 54.

The sampled output from the 50 Hz sampling circuit 62 is a 2-field delay unit 64.
And then provided to a mapping programmable read only memory (PROM) 66.

The output of the mapping PROM 66 provides an interpolated position signal, normalized to fit within the 0 to 1 range.

An inverted search signal is generated from the interpolated position signal by a comparator 68 which tests whether the interpolated position indicated by the interpolated position signal is equal in time to the half way between the preceding and succeeding input frames. Is generated.

If the interpolated position is greater than or equal to the halfway between the two input frames (ie, interpolated position> = 0.5), then the comparator 68 moves for that input frame pair. Generate an output that indicates that the direction of the vector search should be reversed.

There are several possible ways in which the inverted search signal generated by the comparator 68 of FIG. 4 can be used to reorder the input frames used for motion vector estimation.

The first technique is to use a changeover switch, such as the switch 70 shown in FIG.

The switch 70 may be an electronic or mechanical switch and has a pair of inputs 72 and a pair of outputs 74.

Under the control of the inverted search signal from the comparator 68, the switch 70 switches between two positions, in which one position the output is inverted with respect to the input.

Such a switch can be used to invert the connections 16 and 17 of FIG. 2 which supply the video signals representing the two selected input frames to be used for motion vector generation.

However, it is more straightforward to make a switch that operates based on a control signal rather than a video signal. Therefore, as an advanced alternative,
A switch 70 can be used to invert the store read select signal provided to the field store 50 from the two field delay unit 56 and the adder 58.

In this case (when the inverted search signal is set), the input frame received by all the parts forming the standard converter is inverted. That is,
It is not necessary to invert the motion vector, but it is necessary to supply the complement of the interpolation position to the motion vector selector 28 and the interpolator 38.

This arrangement is such that the interpolator always receives frames that are closer together in time on the same input terminal, and is between 0 and 0.5.
It means to always receive the interpolated position of the range.

In this way, the interpolator 38 causes the interpolator 38, rather than the temporal position of the output field with respect to the previous input frame, to
It is responsive to one of the pairs of input frames that are closer in time to the output field and the temporal separation of the output field.

The complement of the interpolation position maps the interpolation position to P
Generated by feeding into a ROM (not shown) or by using a subtractor (not shown) that subtracts the value of the interpolated position from 1 (normalized to be in the range 0-1) You can

Another arrangement in which only the frames fed to the direct block matcher are exchanged is shown in FIG.

FIG. 6 shows a changeover switch 21 'which replaces the even / odd switch 21 of FIG.

This changeover switch 21 'can be switched between the odd and even processing sections of the motion processor 20 under the control of the signal derived from the 50 Hz clock signal 34 in the same manner as the changeover switch 21. Connections 16 and 17 have an additional set of connections that can be inverted under the control of the inverted search signal generated by comparator 68.

FIG. 7 is a diagram showing generation of a motion vector by reverse search. The output field 78 is located closer in time to the temporally subsequent input frame n + 1 than to the temporally preceding input frame n.

In this respect, the situation is the same as that shown in FIG.

The interpolation position of the output field 78 is 0.75.
, Which indicates that the output field is located 3/4 between frame n and frame n + 1.

The comparator 68 detects that the interpolated position of the output field is larger than 0.5, and as described above,
It has the effect of reversing the order of the search for motion vector generation, and thus the temporally subsequent input frame (frame n
Energizing the inversion search signal that causes the +1) search block to be compared with a search area containing a plurality of blocks of the input frame (frame n) each preceding in time.

In the example shown in FIG. 7, frame n +
One search block 80 is compared with a plurality of blocks 83, 84, 85 ... Of frame n.

The best correlation is found between block 80 and the search block 85 in the search area of frame n, where the motion vector 90 is generated.

The motion vector 90 is stored in the block 94 (FIG. 1).
(Equivalent to block 3 of output field 1) of output field 78).

The motion vector 90 has the same localization as the motion vector 9 of FIG. That is, this vector points in the opposite direction.

Assuming that only the frames supplied to the direct block matcher have been exchanged in response to the inverted search signal, the direction of the motion vector is the motion vector estimated from the inverted search signal generated by the comparator 68. Bowl 24
, The motion vector reducer 26, and the motion vector selector 28 of FIG. 2, so that when the comparator 68 detects an interpolation position greater than 0.5, The resulting motion vector (represented numerically during the process) is inverted.

The inversion of the numerical value representing the motion vector is the logical exclusive O of the sign bit associated with the numerical value of the motion vector.
This can be done by applying to one input of the R-gate together with an inverted search signal forming the other input.

The value "1" for the inverted search signal indicates that the interpolated position is greater than or equal to 0.5, and thus the direction of the search for motion vector generation is inverted, and the exclusive logic The output of the OR gate is as follows.

[0121]

[Table 1]

The above table shows that by applying the inverted search signal "1" to one input of the exclusive OR gate, the exclusive OR gate is applied.
It represents that the value applied to the other input of the gate is inverted.

This process is performed to invert the sign bit associated with the numerical value of each motion vector.

[0124]

According to the present invention, since the motion vector is created by selecting the image closest to the output image from the input images, the accuracy of interpolation is good.

[Brief description of drawings]

FIG. 1 is a chart showing the generation of motion vectors.

FIG. 2 is a block diagram of a conventional motion compensation television standard conversion apparatus.

FIG. 3 is a chart showing the operation of the apparatus of FIG.

FIG. 4 is a block diagram of a time base converter.

FIG. 5 is a schematic view of a changeover switch.

FIG. 6 is a schematic view of a changeover switch.

FIG. 7 is a chart showing generation of a motion vector by reverse search.

[Explanation of symbols]

78 output fields 80 search blocks 83,84,85 Blocks to be compared 90 motion vector 92,94 Output field block

Front Page Continuation (72) Inventor Steven Mark Keating UK RG6 3AB, Berkshire, Reading, Lower Early, Huntington Close 28 (58) Fields investigated (Int.Cl. ⁷ , DB name) H04N 7 / 00-7/015

Claims (9)

(57) [Claims] 1. A video signal processing apparatus for interpolating an output image of an output video signal from a corresponding pair of temporally adjacent input images of an input video signal, the pair of temporally adjacent inputs. Means for detecting which of the input images in the image is closer in time to the output image, and in one of a pair of temporally adjacent input images that are closer in time to the output image. Representing motion between a pair of temporally adjacent input images by comparing the search block with respective search areas that each include a plurality of blocks within another of the pair of temporally adjacent input images. Video including a motion vector generator for generating a set of motion vectors and a motion compensation interpolator for interpolating an output image from a pair of temporally adjacent input images using the set of motion vectors Signal processing equipment Place 2. A video signal processing apparatus as claimed in claim 1, wherein a pair of temporally adjacent input images comprises an output image temporally preceding and an input image temporally following the output image. 3. The apparatus according to claim 2, wherein the first clock signal includes a clock pulse which gives an output cycle of each input image of the input video signal, and the clock pulse which gives an output cycle of each output image of the output video signal. Means for receiving a second clock signal including, and the detecting means starts counting at a predetermined rate when a clock pulse of the first clock signal is received,
A counter operative to stop counting when a clock pulse of the second clock signal is received, each output image of an input image temporally preceding and subsequent to that output image. A video signal processing device including means for comparing a count value of a counter with a predetermined threshold value to determine which one is closer. 4. The apparatus of claim 3, including means for receiving the input image of the input video signal in synchronization with the clock pulses of the first clock signal, the counter including a video line of the received input image of the input video signal. Video signal processing device capable of counting. 5. The apparatus according to claim 1, wherein a plurality of image accumulators for accumulating an input image of an input video signal and a plurality of image accumulators for use in interpolation of an output image. Means for generating two read selection control signals for controlling the reading of pairs of input images that are temporally adjacent to each other from the image storage device. 6. The apparatus according to claim 5, wherein the motion vector generator has a plurality of search blocks inside the input image supplied to its first input, each search block in the input image supplied to its second input. A video signal processing device comprising means for comparing with respective search areas including blocks of the blocks and switches for selectively exchanging input images supplied to the first and second inputs of the motion vector generator. 7. The apparatus of claim 6, wherein a motion compensated interpolator is provided for temporal separation of the output image and one of the temporally adjacent input image pairs that is closer in time to the output image. A video signal processing device responsive to which the switch operates to exchange read selection control signals in response to which input image in the temporally adjacent input image pair is detected. 8. The apparatus according to claim 6, wherein a set of motions is detected according to a detection result of which input image in the temporally adjacent input image pair is temporally closer to the output image. A video signal processing apparatus including means for selectively inverting a vector. 9. A method of generating a motion vector used to interpolate an output image of an output video signal from a corresponding pair of temporally adjacent input images of an input video signal, the pair being temporally adjacent. Detecting which of the input images of the input images is temporally closer to the output image, and within one of a pair of temporally adjacent input images that are temporally closer to the output image. And a search block in each of which is included in the other of a pair of temporally adjacent input images, each search area including a plurality of blocks.