JPH0624040B2 - Image matching method and device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an image collation method which is necessary when a computer identifies characters and figures.

(Prior Art) As an example of a conventional method for collating images, a document [Makoto Nagao, Image Recognition Theory, Corona Publishing Co., Ltd., February 15, 1983, first edition, p. 130] will be described below.

For example, with the pixel 10 illustrated in FIG. 8 (A) as the origin, FIG.
A standard pattern image 20 representing a square in a pixel group having a value of 1 illustrated in (B) and a square 40 represented by a pixel group having a value of 1 with the pixel 25 illustrated in FIG. When collating the input image 50 including the image, in the conventional method, the X-axis direction illustrated in FIG. 9 is used as the main scanning direction while scanning from pixel 30 illustrated in FIG. , Y coordinates (M, N) (1) as a center and the similarity at positions M and N between the partial pattern image 20 and the standard pattern image 20 illustrated in FIG. 8B. After calculating the distance image θ (M, N) = cos ⁻¹ R (M, N) [degree] (6) from the distance image, the pixel value Pm of the pixel Pm having the smallest value among the pixel values of the distance image is calculated. Is compared with an appropriately determined threshold value T, and if, for example, P <T (7) holds, it is determined that a square appears at the position corresponding to the pixel Pm of the input image 50, and if the above equation
If (7) is not established, it is determined that the input image 50 does not include a square.

[Problems to be Solved by the Invention] However, in the partial area 70 indicated by diagonal lines in the image 60 shown in FIG. Of the pixels forming the range image 80 illustrated in B), the position of the pixel having the smallest value is uncertain in the region of 9 ≦ X ≦ 11, 9 ≦ Y ≦ 11 (8), and thus the input It was difficult to determine the position of the partial area including the square 40 having a size different from the square represented by the standard pattern image 20 in the partial area of the image 50.

SUMMARY OF THE INVENTION It is an object of the present invention to provide an image collation method and apparatus capable of receiving a deformed pattern and removing the uncertainty with respect to the position of the deformed pattern.

(Means for Solving the Problem) In the image matching method of the present invention, the expansion coefficient group is calculated by expanding the input signal with the wave function group of the harmonic oscillator to obtain the characteristic quantity of the input signal, and the characteristic quantity is standardized. It is characterized by matching with a pattern.

A first image collating apparatus of the present invention is an input image expanding means for calculating expansion coefficients of respective partial areas by expanding a partial area group forming an input image with a wave function group of a harmonic oscillator as a base, and the input. Input image description group storage means for storing the expansion coefficient for each partial area calculated by the image expansion means, and standard pattern description storage for storing the expansion coefficient calculated by expanding the standard pattern with the wave function group of the harmonic oscillator. Means, and the expansion coefficient of the standard pattern stored in the standard pattern description storage means and the expansion coefficient for each partial area stored in the input image description group storage means, in the subspace spanned by the harmonic oscillator wave function group. A distance image synthesizing means for calculating and storing the distance between the standard pattern and each partial area,
Among the partial areas forming the input image, the position of the partial area closest to the standard pattern is found.

A second image collation device of the present invention is a standard pattern description storage unit that stores a development coefficient of a standard pattern, a local position storage unit that stores a center position of a partial area forming an input image,
Scanning means for moving the central position of the partial area stored by the local position storage means according to a predetermined rule, and partial area corresponding to the position stored by the local position storage means among the partial areas forming the input image. An input image local expansion means for calculating expansion coefficient by expanding the wave function group of the harmonic oscillator as a base; an input image local description storage means for storing expansion coefficient calculated by the input image local expansion means; Using the expansion coefficient of the standard pattern stored in the standard pattern description storage means and the expansion coefficient stored in the input image local description storage means, the portion of the standard pattern and the input image in the subspace where the wave function of the harmonic oscillator extends A distance calculating means for calculating a distance between regions, and finding a position of a partial region closest to the standard pattern among partial regions forming the input image. It is characterized in.

A third image collating apparatus of the present invention comprises a standard pattern description storage means for storing expansion coefficients of standard patterns, a one-dimensional function group storage means for storing a wave function group of a one-dimensional harmonic oscillator, and an input image. Input image XY separation expansion means for calculating expansion coefficient of each partial area by expanding the partial function group to be stored with the wave function group of the harmonic oscillator stored in the one-dimensional function group storage means, and the input image XY. Input image description group storage means for storing the expansion coefficient for each partial area calculated by the separation and expansion means, expansion coefficient of the standard pattern stored by the standard pattern description storage means, and partial area stored by the input image description group storage means The input image is configured by including distance image synthesizing means for calculating and storing the distance between the standard pattern and each partial area in the partial space spanned by the wave function of the harmonic oscillator from each expansion coefficient. Characterized in that to find the position of the portion nearest regions in the standard pattern of that subregion.

A fourth image collating apparatus of the present invention comprises a standard pattern description storage means for storing expansion coefficients of standard patterns, a one-dimensional function group storage means for storing a wave function group of a one-dimensional harmonic oscillator, and an input image. The local position storage means for storing the position of the partial area to be stored, the scanning means for moving the central position of the partial area stored by the stored local position input means according to a predetermined rule, and the partial area forming the input image Input image local XY separation expansion for calculating expansion coefficient by expanding the partial region corresponding to the position stored in the local position storage means using the wave function group of the harmonic oscillator stored in the one-dimensional function group storage means as a basis. Means and the input image local XY
Input image local description storage means for storing the expansion coefficient calculated by the separation expansion means, harmony using the expansion coefficient of the standard pattern stored by the standard pattern description storage means and the expansion coefficient stored by the input image local description storage means In a subspace spanned by the wave function of the oscillator, a distance calculating means for calculating the distance between the standard pattern and the partial area of the input image is provided, and the most part of the standard pattern among the partial areas constituting the input image is provided. It is characterized by finding the position of a close partial area.

(Operation) The first principle of the present invention is to map an image onto a subspace spanned by a wave function group of a standardized harmonic oscillator, and then calculate the distance between the standard pattern image and the input image within the subspace. It is a thing.

That is, in the image matching method of the present invention, first, in the first step, for example, a two-dimensional standard pattern image f (x, y) (9) is standardized and the wave function group of the harmonic oscillator is standardized. However, Expansion coefficient group calculated by expanding with In the second step, for example, the two-dimensional input image g (x, y) (12) at the position (M, N) (13) is normalized by the harmonic oscillator illustrated in the equation (10). Expansion coefficient group calculated by expanding with the wave function group of In the third step, the standard pattern image f (x, y) whose pixel value is expressed by the above equation (9) and the position (M,
Distance of the input image g (x, y) shown in (12) in (N) However, a distance image Θ (M, N) for which | K | ≧ 0, L ≧ 0 is generated, and the pixel Pm having the smallest value among the pixel values of the distance image shown in the above equation (15) as the fourth step. comparing the pixel value P and appropriately threshold T defined, for example, p <T (19) is the input image g if satisfied (x, y) wherein said standard pattern image at a position corresponding to the pixel Pm of It is determined that the same image as f (x, y) appears, and if the above equation (19) is not satisfied, the input image g (x, y) has the standard pattern image f (x, y) and It is determined that the same image is not shown.

The second principle of the present invention is to use the expansion coefficient group calculated by the equation (11) as the first step, for example, And the expansion coefficient group calculated by the equation (14) as the second step In the third step, the judgment is made in the same manner as the procedure described in the first principle.

First Embodiment A first embodiment of the present invention will be described in detail below with reference to the drawings.

FIG. 1 shows a first embodiment for outputting the position of a square 40 shown in the same figure among the partial area groups constituting the input image 50 shown in FIG. 9 using the first principle of the present invention. It is a block diagram shown.

In FIG. 1, 1000 is the pixel 1 illustrated in FIG. 8 (A).
A standard pattern storage means for storing a standard pattern image representing a square in a pixel group whose value is 1 illustrated in FIG. 8 (B) with 0 as the origin, and 1090 as the origin with the pixel 25 illustrated in FIG. 9 as the origin. The input image storage means stores an input image 50 including a square 40 represented by pixels having a value of 1.

The operation starts when the control means 1100 composed of, for example, a microprocessor and a memory activates the standard pattern developing means 1010.

The activated standard pattern developing means 1010 is represented by the pixel values fij (-5 ≦ i ≦ 5, −5 ≦ j ≦ 5) (22) of the image stored in the standard pattern storage means 1000 (22) and the equation (10). Using the wave function group of the harmonic oscillator, the expansion coefficient group shown in the equation (11) is set to σ = 1, for example. After the above calculation, the expansion coefficient group is stored in the standard pattern description storage means.
Store in 1020 and end the process.

When the standard pattern developing means 1010 finishes the processing, the control means 1100 activates the input image developing means 1080. The activated input image expansion means 1080 has the pixel value gij (0≤i≤20, 0≤j≤20) stored by the input image storage means 1090 as the first step (24) and the equation (23-1). Wave function group Ψkli of the harmonic oscillator shown in
Using j as the main scanning direction in the X-axis direction illustrated in FIG. 9, for example, starting from the pixel 30 illustrated in FIG. 9 and scanning the pixel, the position of the pixel illustrated in the formula (13)
Let the expansion coefficient group shown in the equation (14) of the partial image determined centering on (M, N) be σ = 1, for example. However, k = 0,1,2,3,1 = 0,1,2,3 is calculated, and in the second step, the expansion coefficient group is a matrix whose pixel values are, for example, the matrix illustrated in FIG. 2 (A). Substituting at the position (M, N) of the description image 100 illustrated in FIG. 2 (B), as the third step,
The above process is repeated for the partial area 70 illustrated in FIG. 10 (A), and the description image 100 is stored in the input image description group storage means 1070 as the fourth step, and the process is terminated.

When the input image developing means 1080 ends the processing, the control means 1100 activates the distance image synthesizing means 1030. The distance image synthesizing means 1030 which has been activated is, for example, the first scanning step with the X-axis direction of the description image 100 stored in the input image description group storage means 1070 illustrated in FIG. While scanning pixels starting from the pixel 110 illustrated in FIG. 2B, the expansion coefficient group β _kl (M, N) shown in the equation (25) in the partial area centered on the position (M, N) of the pixel is scanned. ) And the expansion coefficient group represented by the equation (23) stored in the standard pattern description storage means 1020, the distance Θ (M, N) defined by the equation (15).
Is calculated, and in the second step, the distance is stored in the distance image storage means 1040 as a pixel value at the position (M, N),
As a step, the above processing is repeated for the partial area 70 illustrated in FIG. 10 (A), and the distance image 120 illustrated in FIG. 3 corresponding to the partial area 70 is acquired as the distance image storage means 104.
The value is stored in 0, and the processing is completed.

When the distance image synthesizing means 1030 ends the processing, the control means 1100 activates the position detecting means 1050. The activated position detecting means 1050 has a pixel value P of, for example, P <10 in the pixel group of the distance image 120 illustrated in FIG. 3 among the distance images stored in the distance image storing means 1040 in the first step. The number N of pixels that satisfy (26) and the positions of the pixels are checked to obtain N = 5 (27) and the pixel group 130 illustrated by the diagonal lines in FIG. 4 is obtained. Number of pixels N satisfying equation (26)
After confirming that N satisfies N ≧ 1 (28), the position (10,10) (29) of the pixel 140 shown in FIG. 4 having the smallest pixel value is displayed in the pixel group 130 shown in FIG. The data is output to the means 1060 and the processing is terminated.

The display means 1060 displays the coordinate values output by the position detection means 1050 on, for example, a display, so that the position (10, 10) of the image 50 stored in the input image storage means 1090 illustrated in FIG. 9 is stored. The user is informed that a square is shown in the partial area centered at, and the above ends all processing.

In the above description, the standard pattern storage means 1000, the standard pattern description storage means 1020, and the input image storage means
1090, the input image description group storage means 1070 and the distance image storage means 1040 can be realized by a memory, and the display means 1060 can be realized by a current display technology.

Second Embodiment Next, a second embodiment will be described in detail with reference to the drawings.

FIG. 5 shows a second embodiment for outputting the position of the square 40 shown in the same figure among the partial area groups forming the input image 50 shown in FIG. 9 using the first principle of the present invention. It is a block diagram showing.

In FIG. 5, 2030 indicates coordinate values (0,
0) and the distance 360, the minimum value storage means 1000, for example, the pixel 10 illustrated in FIG.
The standard pattern storage means for storing a standard pattern image representing a square in a pixel group having a value of 1 illustrated in FIG. 10 is represented by a pixel having a value of 1 with the pixel 25 illustrated in FIG. 9 as an origin. An input image storage means for storing an input image 50 including a square 40, and a local position storage means 2060 for storing the coordinate value (5, 5) of the pixel 30 illustrated in FIG. 9 as an initial value.

The operation starts when the control means 1100 composed of, for example, a microprocessor and a memory activates the standard pattern developing means 1010.

The activated standard pattern expansion means 1010 stores the expansion coefficient group shown in the equation (23) in the standard pattern description storage means 1020 in the same manner as the procedure shown in the first embodiment, and ends the processing. .

When the standard pattern developing means 1010 completes the processing, the control means 1100 activates the input image local expanding means 2040.
The input image local expansion means 2040 that has been activated changes the coordinate values stored in the local position storage means 2060 as the first step.
As (M, N), the expansion coefficient group β _kl (M, N) shown in the equation (25) is calculated in the same manner as the procedure shown in the first embodiment, and the expansion coefficient group is calculated as the second step. β _kl (M, N) is stored in the input image local description storage means 2050, and the process ends.

When the input image local expansion means 2040 completes the processing, the control means 1100 activates the distance calculation means 2000. The distance calculating means 2000 which is started up is the distance Θ (M, N) defined by the equation (15) in the same manner as the procedure shown in the first embodiment.
Is calculated to obtain the value 81, the distance 81 is stored in the distance storage means 2010, and the process ends.

When the distance calculation means 2000 ends the processing, the control means 1100 activates the distance comparison means 2020. The activated distance comparison means 2020 compares the distance P = 360 (30) stored in the minimum value storage means 2030 with the distance Θ = 81 (31) stored in the distance storage means 2010, and Θ ≦ p ( 32) holds, the distance Θ and the local position storage means 20
The position (5, 5) stored in 60 is stored in the minimum value storage means 2030, and the process is ended.

When the distance comparison means 2020 completes the processing, the control means
1100 activates the scanning means 2070. The activated scanning means 2070 has a position stored in the local position storage means 2060.
Starting from (5, 5), 1 is added to the X coordinate value of the position stored by the local position storage means 2060 so that the pixel group forming the partial area 70 shown in FIG. 10 is scanned in raster order (6, Change to 5) and finish the process.

When the scanning means 2070 completes the processing, the control means 11
00 again activates the input image local expansion means 2040, and the processing is continued in the same manner as the above-mentioned procedure, for example, the scanning means 2070 updates the contents of the local position storage means 2060 to (11, 10). It is assumed that the processing is completed. At this time, the minimum value storage means 2030 stores the distance 7 and the position (10, 10).

When the scanning means 2070 completes the processing, the control means 1100
After activating the input image local expansion means 2040, activates the distance calculation means 2000 through a process similar to the procedure shown above.

The activated distance calculating means 2000 calculates the distance in the same manner as the above-mentioned procedure to obtain a value of 9, for example, and then stores the distance 9 in the distance storing means 2010 and ends the processing.

When the distance calculation means 2000 finishes the processing, the control means
1100 activates the distance comparison means 2020. The activated distance comparison means 2020 compares the distance P = 7 (33) stored in the minimum value storage means 2030 with the distance Θ = 9 (34) stored in the distance storage means 2010, and the calculated expression (32) is obtained. Since the condition shown is not satisfied, the process ends.

When the distance comparison means 2020 completes the processing, the control means
The 1100 activates the scanning means 2070, and the processing is continued in the same manner as described above until the scanning means 2070 updates the position (15, 15) stored in the local position storage means 2060, for example.

When the scanning means 2070 updates the position (15, 15) stored in the local position storage means 2060, the control means 1100 activates the determination means 2080. The activated determination means 2080 confirms that the distance P = 7 (35) stored by the minimum value storage means 2030 satisfies the condition of the equation (26), and then the minimum value storage means 2030 stores it. The position (1) of the image 50 illustrated in FIG. 9 stored in the input image storage means 1090 is displayed by displaying the position (10, 10) on the display, for example.
The user is informed that a square is imaged in the partial area centered on (0, 10), and all processing is completed.

In the above description, the standard pattern storage means 1000, the standard pattern description storage means 1020, and the input image storage means
1090, the input image local description storage means 2050, the local position storage means 2060, the distance storage means 2010, and the minimum value storage means 2030 can be constituted by memories.

Third Embodiment Next, a third embodiment will be described in detail with reference to the drawings.

FIG. 6 shows a third embodiment for outputting the position of the square 40 shown in the same figure among the partial area groups forming the input image 50 shown in FIG. 9 using the second principle of the present invention. It is a block diagram shown.

In FIG. 6, reference numeral 1000 is a standard for storing a standard pattern image representing a square with a pixel group having a value of 1 illustrated in FIG. 8 (B) with the pixel 10 illustrated in FIG. 8 (A) as an origin. The pattern storage means 1090 stores, for example, an input image storage means 30 for storing an input image 50 including a square 40 represented by a pixel having a value of 1 with the pixel 25 illustrated in FIG. 9 as an origin.
20 is the wave function group of the harmonic oscillator shown in the above equation (10), for example, with σ = 1. For example It is a one-dimensional function group storage means for discretizing and storing in.

The operation starts when the control means 1100 composed of a microprocessor and a memory activates the standard pattern XY separation / expansion means 3000.

The activated standard pattern XY separation and expansion means 3000 is the harmonic oscillator wave function group stored in the one-dimensional function group storage means 2020 and the pixel value fij shown in the equation (22) stored in the standard pattern storage means 1000. And the expansion coefficient group shown in the equation (20) is set to, for example, σ = 1. After the above calculation, the expansion coefficient group is stored in the standard pattern description storage means.
Store in 1020 and end the process.

When the standard pattern XY separation / expansion means 3000 finishes the processing, the control means 1100 activates the input image XY separation / expansion means 3010. The input image XY separation and expansion means 30 that has been started
In the first step, reference numeral 10 denotes the pixel value gij shown in the equation (24) stored in the input image storage means 1090 and the wave function of the harmonic oscillator shown in the equation (36) stored in the one-dimensional function group storage means 3020. group Using the X-axis direction illustrated in FIG. 9 as the main scanning direction, for example, starting from the pixel 30 illustrated in FIG. 9 and scanning the pixel, the position of the pixel illustrated in the formula (13) is used.
Let the expansion coefficient group shown in the equation (21) of the partial area determined around (M, N) be σ = 1, for example. And the first step from the expansion coefficient group as the second step.
The description image 100 illustrated in FIG. 2 (B) is generated in the same manner as the procedure described in the above embodiment, and as the third step, the description image 100 is stored in the input image description group storage means 1070 and processed. finish.

When the input image XY separation / expansion means 3010 completes the processing, the control means 1100 activates the distance image synthesis means 1030.
The activated distance image synthesizing means 1030 synthesizes the distance image 120 illustrated in FIG. 3 in the same manner as the procedure described in the first embodiment, and then the distance image 120 is stored in the distance image storage means 1040. Store and finish the process.

When the distance image synthesizing means 1030 ends the processing, the control means 1100 activates the position detecting means 1050. The activated position detecting means 1050 obtains a pixel group 130 shown by hatching in FIG. 4 in the same manner as the procedure described in the first embodiment,
The number N of the pixel group 130 shown in the equation (27) is equal to the condition (2
After confirming that 8) is satisfied, the pixel group shown in FIG.
The position (10, 10) of the pixel 140 shown in FIG. 13 having the smallest pixel value out of 130 is output to the display means 1060, and the process ends.

The display means 1060 displays the coordinate value output by the position detection means 1050 on, for example, a display, and the position (10, 10) of the image 50 stored in the input image storage means 1090 illustrated in FIG. 9 is shown. The user is informed that a square is shown in the partial area centered at, and the above ends all processing.

In the above description, the standard pattern storage means 1000, the standard pattern description storage means 1020, and the input image storage means
1090, the input image description group storage means 1070, the distance image storage means 1040, and the one-dimensional function group storage means can be constituted by a memory, and the display means 1060 can be realized by the current display technology.

Fourth Embodiment Next, a fourth embodiment will be described in detail with reference to the drawings.

FIG. 7 shows a fourth embodiment for outputting the position of a square 40 shown in the same figure among the partial areas constituting the input image 50 shown in FIG. 9 using the second principle of the present invention. It is a block diagram shown.

In FIG. 7, 2030 indicates coordinate values (0,
0) and the distance 360, a minimum value storage means, 2060 is a local position storage means for storing the coordinate value (5, 5) of the pixel 30 illustrated in FIG. 9 as an initial value, and 1000 is the eighth position, for example.
A standard pattern storage means for storing a standard pattern image representing a square in a pixel group having a value of 1 illustrated in FIG. 8 (B) with the pixel 10 illustrated in FIG. An input image storage means for storing an input image 50 including a square 40 represented by a pixel having a value of 1 with the pixel 25 exemplified in the figure as an origin, 3020 is a harmonic shown in the formula (10), for example, σ = 1. It is a one-dimensional function group storage means for discretizing and storing the wave function group of the oscillator by, for example, the above formula (36).

The operation starts when the control means 1100 composed of a microprocessor and a memory activates the standard pattern XY separation / expansion means 3000.

The activated standard pattern XY separation and expansion means 3000 calculates the expansion coefficient group α _kl shown in (37) in the same manner as the procedure described in the third embodiment, and then describes the expansion coefficient group in the standard pattern description. It is stored in the storage means 1020 and the process is terminated.

When the standard pattern XY separation / expansion means 3000 finishes the processing, the control means 1100 causes the input image local XY separation / expansion means 40.
Start 00. The activated input image local XY separation / expansion means 4000 is the local position storage means 2060 as the first step.
With the coordinate value stored by (M, N) as the expansion coefficient group β shown in the equation (38) in the same manner as the procedure shown in the third embodiment.
kl (M, N) is calculated, and as a second step, the expansion coefficient group βk
l (M, N) is stored in the input image local description storage means 2050, and the process ends.

When the input image local XY separation / expansion means 4000 finishes the processing, the control means 1100 activates the distance calculation means 2000. The distance calculating means 2000 which is activated calculates the distance Θ (M, N) defined by the equation (15) in the same manner as the procedure described in the second embodiment, and after obtaining the value 81, for example. , Above distance 8
The value 1 is stored in the distance storage means 2010, and the process ends.

When the distance calculation means 2000 ends the processing, the control means 1100 activates the distance comparison means 2020. The activated distance comparison means 2020 compares the distance P shown in the (30) stored in the minimum value storage means 2030 with the distance Θ shown in the equation (31) stored in the distance storage means 2010, Since the condition shown in (32) is satisfied, the distance Θ and the position (5, 5) stored in the local position storage means 2060 are set to the minimum value storage means 2030.
And the process ends.

When the distance comparison means 2020 completes the processing, the control means
1100 activates the scanning means 2070. The activated scanning means 2070 changes the position stored in the local position storage means 2060 to (6, 5) in the same manner as the procedure described in the second embodiment, and ends the processing.

When the scanning means 2070 completes the processing, the control means 1100
Activates the input image local XY separation / expansion means 4000, and the processing is continued in the same manner as described above. For example, the scanning means 2070 causes the contents of the local position storage means 2060 to be (11, 10).
It is assumed that the process is finished by updating to. At this time, the minimum value storage means 2030 stores the distance 7 and the position (10, 10).

When the scanning means 2070 completes the processing, the control means 1100
After activating the input image local XY expansion means 2040, the distance calculation means 20 is subjected to a process similar to the procedure described above.
Start 00.

The activated distance calculation means 2000 calculates the distance in the same manner as described above to obtain a value of 9, for example, and then stores the distance 9 in the distance storage means 2010 and ends the process.

When the distance calculation means 2000 finishes the processing, the control means
1100 activates the distance comparison means 2020. The activated distance comparison means 2020 is the equation stored in the minimum value storage means 2030.
The distance P shown in (33) is compared with the distance Θ shown in the equation (34) stored in the distance storage means 2010. Since the condition shown in the equation (32) is not satisfied, the process ends.

When the distance comparison means 2020 finishes the processing, the control means 1100 activates the scanning means 2070, and in the same manner as the procedure described above, the processing is performed, for example, by the scanning means 2070 at the position stored in the local position storage means 2060. Continue until you update (15,15).

When the scanning means 2070 updates the position (15, 15) stored in the local position storage means 2060, the control means 1100 activates the determination means 2080. The determination means 2080 that has been started up has the distance P shown in the equation (35) stored in the minimum value storage means 2030.
Is confirmed to satisfy the condition of the equation (26), the position (10, 10) stored in the minimum value storage means 2030 and the input image storage means 1090 are stored by displaying on the display, for example. The position of the image 50 illustrated in FIG. 9 (1
The user is informed that a square is imaged in the partial area centered on (0, 10), and all processing is completed.

In the above description, the standard pattern storage means 1000, the standard pattern description storage means 1020, and the stored image input means
1090, the input image local description storage means 2050, the local position storage means 2060, the distance storage means 2010, the minimum value storage means 2030, and the one-dimensional function group storage means 3020 can be constituted by memories.

(Effects of the Invention) The first invention described above has the effect of being able to determine the position of a partial area including the square 40 having a size different from that of the square represented by the standard pattern image 20, and the second invention is the above-mentioned input. Image description group storage means 1070 and the distance image storage means 1040
Since it can be omitted, there is an effect that it can be configured with less memory than the first invention, and the third invention is that the expansion coefficient calculated by the equation (25) in the first invention has a smaller number of calculations. By calculating in (38), there is an effect that processing can be executed faster than in the first invention, and the fourth invention is better than the third invention for the same reason as explained in the effect of the second invention. There is an effect that can be configured with less memory.

In the above description, for example, the two-dimensional image in the equation (11) is targeted, However, it generally holds in any dimension and any value.

[Brief description of drawings]

FIG. 1 is a block diagram showing the first embodiment of the present invention, and FIG.
(A), (B), FIG. 3, and FIG. 4 are diagrams for explaining the operation of the first embodiment, and FIG. 5 is a block diagram showing the second embodiment of the present invention. FIG. 7 is a block diagram showing a third embodiment of the present invention, FIG. 7 is a block diagram showing the fourth embodiment of the present invention, FIG. 8 (A), (B), FIG. 9 and FIG. (A), (B) is a figure for explaining a prior art. In the figure, 1020 is a standard pattern description storage means, 1030 is a distance image composition means, 1080 is an input image expansion means, 1090 is an input image storage means, 2000 is a distance calculation means, 2050 is an input image local description storage means, and 2040 is an input. Image local expansion means, 2060 local position storage means, 2070 scanning means, 3010 input image XY separation expansion means, 3020 one-dimensional function group storage means, 4000 input image local XY separation expansion means, 20 is an embodiment The standard pattern used,
Reference numeral 50 is an input image used in the examples.

Claims (5)

[Claims] 1. An image collation, wherein an expansion coefficient group is calculated by expanding an input signal with a wave function group of a harmonic oscillator to obtain a characteristic amount of the input signal, and the characteristic amount is collated with a standard pattern. Method. 2. An input image expanding means for calculating a expansion coefficient of each partial area by expanding a partial area group forming an input image with a wave function group of a harmonic oscillator as a base, and the input image expanding means for calculating. Input image description group storage means for storing the expansion coefficient for each partial area, standard pattern description storage means for storing the expansion coefficient calculated by expanding the standard pattern with the wave function group of the harmonic oscillator, and the standard The standard pattern in the subspace spanned by the wave function group of the harmonic oscillator, which is obtained from the expansion coefficient of the standard pattern stored in the pattern description storage means and the expansion coefficient of each partial area stored in the input image description group storage means, A distance image synthesizing unit for calculating and storing the distance of each partial region is provided, and the position of the partial region closest to the standard pattern among the partial regions forming the input image is found. That the image matching system. 3. A standard pattern description storing means for storing a standard pattern expansion coefficient, a local position storing means for storing a center position of a partial area forming an input image, and the partial area stored by the local position storing means. The scanning means for moving the center position of the object according to a predetermined rule, and the partial area corresponding to the position stored by the local position storage means among the partial areas forming the input image, based on the wave function group of the harmonic oscillator as a base. The input image local expansion means for calculating the expansion coefficient by expanding, the input image local description storage means for storing the expansion coefficient calculated by the input image local expansion means, and the standard pattern stored by the standard pattern description storage means The standard pattern and the input image in the subspace spanned by the wave function of the harmonic oscillator using the expansion coefficient and the expansion coefficient stored in the input image local description storage means. Comprising a distance calculating means for calculating the distance between the partial regions, the image matching system to find the location of the nearest partial region in the standard pattern of the partial areas constituting the input image. 4. A standard pattern description storage means for storing expansion coefficients of a standard pattern, a one-dimensional function group storage means for storing a wave function group of a one-dimensional harmonic oscillator, and a partial area group constituting an input image. The input image XY separation and expansion means for calculating the expansion coefficient of each partial region by expanding the wave function group of the harmonic oscillator stored in the one-dimensional function group storage means as a base, and the input image XY separation and expansion means are calculated. Input image description group storage means for storing the expansion coefficient for each partial area
Based on the expansion coefficient of the standard pattern stored in the standard pattern description storage means and the expansion coefficient of each partial area stored in the input image description group storage means, the standard pattern and each of the standard patterns in the subspace in which the wave function of the harmonic oscillator extends. An image collating apparatus comprising a distance image synthesizing means for calculating and storing a distance of a partial area, and finding the position of the partial area closest to the standard pattern among the partial areas forming the input image. 5. A standard pattern description storage means for storing expansion coefficients of a standard pattern, a one-dimensional function group storage means for storing a wave function group of a one-dimensional harmonic oscillator, and positions of partial areas forming an input image. Local position storing means for storing the local position storing means, scanning means for moving the central position of the partial area stored by the local position storing means according to a predetermined rule, and the local position storing means for the partial area forming the input image. Input image local XY separation expansion means for calculating expansion coefficients by expanding the partial region corresponding to the position stored by the above with the wave function group of the harmonic oscillator stored in the one-dimensional function group storage means as a base; The input image local description storage means for storing the expansion coefficient calculated by the image local XY separation expansion means, the expansion coefficient of the standard pattern stored in the standard pattern description storage means and the input The image local description storage means is provided with distance calculation means for calculating a distance between a standard pattern and a partial area of an input image in a subspace spanned by a wave function of a harmonic oscillator, using the expansion coefficient, An image collation device for finding the position of a partial area closest to the standard pattern among partial areas forming an input image.