JP3098704B2 - Image matching method and device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image collating method and apparatus for collating an image of a reference image registered in advance with an input image input from an image input device or the like. And when moving in parallel,
The present invention relates to an image matching method and apparatus for calculating a rotation angle and a translation amount of an input image and a reference image at high speed and comparing the two images.

[0002]

2. Description of the Related Art Conventionally, when comparing a reference image registered in advance with an input image input from an image input device or the like, a generalized Hough which can calculate a rotation angle and a translation amount between figures can be calculated. Transforms are often used.

For example, Japanese Patent Application Laid-Open No. Sho 62-77689 discloses a generalized Hough transform circuit in which a circuit capable of executing the generalized Hough transform at high speed is realized by hardware.

[0004] The generalized Hough transform is an image processing technique for detecting an arbitrary figure from an image, and a method of voting in a three-dimensional parameter space of a rotation angle θ and a translation amount (x, y). It is.

[0005]

However, the generalized Hough transform has a problem that a huge memory corresponding to the parameter space is required.

For example, when the resolution of the translation is 1 pixel and the resolution of the rotation angle is 1 degree, when collating an image composed of 256 pixels × 256 pixels, several tens of megabytes of parameter space are used. Memory capacity is required.

In the generalized Hough transform, the amount of rotation and the amount of translation are determined from all combinations between the edge points of the input image and the reference image. Therefore, the processing time increases as the number of edges increases. There's a problem.

In view of the above, the present invention solves the above-mentioned problems, and can quickly perform complicated image comparison while reducing the required memory capacity when comparing rotated and translated images. An object of the present invention is to provide an image matching method and apparatus.

[0009]

According to a first aspect of the present invention, there is provided an image collating method for collating an image between a reference image and an input image, the method comprising the steps of: Perform θ-ρ Hough transform and refer to θ
-P plane data is created, the reference θ-ρ plane data is Fourier transformed to create reference θ-f plane data, and the θ-ρ Hough transform is performed on the edge direction image created from the input image. To create input θ-ρ plane data, perform Fourier transform on the input θ-ρ plane data to create input θ-f plane data, and obtain the input θ-f plane data and the reference θ.
A rotation angle is calculated based on the position shift amount of the f-plane data, and a cross-correlation coefficient is calculated for each of the reference θ-ρ plane data and the input θ-ρ plane data corrected using the rotation angle, and It is characterized in that Hough transform is performed to calculate a parallel movement amount.

According to a second aspect of the present invention, there is provided an image matching apparatus for performing image matching between a reference image and an input image, wherein an edge direction image is created by differentiating the reference image and the input image, and the edge direction image is subjected to Hough transform. Then, the θ-ρ plane data is created, and the θ-ρ plane data is further Fourier-transformed to obtain θ
Data creating means for creating -f plane data; reference data storage means for storing reference θ-ρ plane data and reference θ-f plane data corresponding to the reference image; and input θ-f corresponding to the input image. Rotation angle calculation means for calculating a rotation angle based on a positional shift amount between the plane data and the reference θ-f plane data stored in the reference data storage means, and reference θ-ρ plane data stored in the reference data storage means Is corrected using the rotation angle calculated by the rotation angle calculation means, and the corrected reference θ-ρ plane data and the input θ−
A parallel movement amount calculating means for calculating a cross-correlation coefficient at each θ of the ρ plane data and performing an inverse Hough transform to calculate a parallel movement amount.

[0011]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a diagram showing a configuration of an image matching apparatus 10 used in the present embodiment.

As shown in FIG. 1, this image collating device 10
Rather than simply comparing the reference image prepared in advance and the input image using only the Hough transform, the θ-ρ plane data obtained by the Hough transform is further Fourier transformed to calculate the θ-f plane data. The rotation angle is obtained at the θ-f plane level, and the translation amount is obtained at the θ-ρ plane level using the obtained rotation angle.

More specifically, the image collating apparatus 10 includes an image input unit 11, a θ-f plane creating unit 12, an edge detecting unit 13, a Hough transform processing unit 14, a Fourier transform processing unit 15, Movement amount calculation unit 16 and reference data storage unit 17
And an inverse Hough transform processing unit 18.

Note that the θ-f plane creating unit 12
The edge detecting unit 13 is used to generate an edge direction image,
When creating the θ-ρ plane data, the Hough transform processing unit 14
Is used, and the Fourier transform processing unit 15 is used when creating the θ-f plane data.

The reference data storage unit 17 corresponds to the reference data storage unit of the second invention, and the movement amount calculation unit 16
Corresponds to the rotation angle calculation means and the parallel movement amount calculation means.

The image input unit 11 is a processing unit for inputting a reference density image R (x, y) and an input density image I (x, y) prepared in advance, and creates each of the input images in a θ-f plane. Output to the unit 12.

The θ-f plane creating unit 12 includes the edge detecting unit 1
3. An edge direction image, θ-ρ plane data, and θ-f plane data are sequentially created from the input image by using the Hough transformation processing unit 14 and the Fourier transformation processing unit 15, and the θ-
The ρ plane data and θ-f plane data are transferred to the movement amount calculation unit 16.
Output to

More specifically, an edge direction image is created from the input image using the edge detection unit 13, and then the edge direction image is Hough transformed using the Hough transformation processing unit 14. Further, the θ-ρ plane data obtained by the Hough transform is Fourier-transformed using the Fourier transform processing unit 15 to create θ-f plane data.

That is, when the input image is R (x, y), the θ-f plane creating unit 12 creates reference θ-f plane data corresponding to the R (x, y). If the input image is I (x, y), input θ-f plane data corresponding to the I (x, y) is created.

For example, when the input image is R (x, y), first, a reference edge direction image Re (x, y) is created by using the edge detection unit 13, and then, Re (x, y) is subjected to Hough transform by the Hough transform processing unit 14 to create reference θ-ρ plane data R (θ, ρ). Further, this R (θ, ρ) is Fourier-transformed using the Fourier-transform processing unit 15, and the reference θ-f plane data R (θ,
Create f).

Note that R (θ, ρ) and R (θ, f)
Is received, the movement amount calculator 16 calculates R (θ, ρ) and R
(Θ, f) is stored in the reference data storage unit 17 as reference data.

The edge detector 13 applies a Gaussian Laplacian filter to the density image, and a point at which the sign of the Gaussian Laplacian filter changes in the x-axis direction or the y-axis direction (hereinafter referred to as a “zero cross point”). ) Is calculated, and the edge strength Em and the edge direction Eθ are calculated while removing noise by applying a Sobel operator to the position of the zero-cross point. Is a processing unit that stores the edge direction Eθ in the edge direction image on condition that the edge direction is equal to or larger than the threshold value.

Specifically, according to the present invention, Japanese Patent Laid-Open No.
Using the edge detection method based on the zero-cross point disclosed in Japanese Patent Publication No.
Ie (x, y) corresponding to (x, y) or I (x, y)
Create

The Hough transform processing unit 14 includes an edge detecting unit 13
Is a processing unit that creates the θ-ρ plane data by Hough transforming the created edge direction image, and specifically, Re (x,
Create R (θ, ρ) corresponding to y) or I (θ, ρ) corresponding to Ie (x, y).

The Fourier transform processing unit 15 is a processing unit that performs a Fourier transform on the θ-ρ plane data created by the Hough transform processing unit 14 to create θ-f plane data. Specifically, R (θ, R (θ, f) or I (θ,
Create I (θ, f) corresponding to ρ).

The movement amount calculating section 16 includes a θ-f plane creating section 1
2 is a processing unit that calculates and outputs the rotation angle and the translation amount of the object included in the input density image I (x, y) and the object included in the reference density image R (x, y).

More specifically, the movement amount calculating section 16 calculates the parallel movement amount at the θ-ρ plane level on which the Hough transform is performed, and further calculates the θ-f plane level by further Fourier transforming the θ-ρ plane. Is the rotation angle. Note that θ-ρ
The reason why the plane is further Fourier-transformed is to quickly obtain the rotation angle while keeping the translation amount unchanged, and then calculate the translation amount using the calculated rotation angle.

When the movement amount calculating section 16 receives the reference data, ie, R (θ, ρ) and R (θ, f) from the θ-f plane creating section 12, the moving amount calculating section 16 converts the reference data into the reference data. The information is stored in the storage unit 17.

For this reason, although the detailed description is omitted,
When the θ-f plane creation unit 12 outputs reference data to the movement amount calculation unit 16, an identification flag or the like indicating that the output data is reference data is added to the data.

The reference data storage unit 17 stores the reference density image R
A storage unit that stores θ-ρ plane data R (θ, ρ) and θ-f plane data R (θ, f) corresponding to (x, y) as reference data, and the reference data is a movement amount calculation unit. 16 is accessed.

The inverse Hough transform processing section 18 includes a moving amount calculating section 1
6 is a processing unit used when calculating the amount of translation,
Specifically, the movement amount calculation unit 16 stores the correlation coefficient calculated between the reference θ-ρ plane and the input θ-ρ plane.
An inverse Hough transform is performed on the cross-correlation image C (θ, ρ).

As described above, when the reference density image R (x, y) and the input density image I (x, y) are compared with each other, the image matching apparatus 10 performs the Hough transform and then performs the Fourier transform. While obtaining the rotation angle at the θ-f plane level,
Based on the rotation angle, the translation amount is calculated at the θ-ρ plane level.

Next, the processing procedure of the image collating apparatus 10 will be described. Here, it is assumed that the reference data has already been set in the reference data storage unit 10.

FIG. 2 is a flowchart showing a processing procedure of the image collating apparatus 10 shown in FIG.

As shown in FIG. 2, the image collating device 10
Is, when the input density image I (x, y) is input, θ-f
The plane creation unit 12 uses the edge detection unit 13 to generate I (x,
An edge direction image Ie (x, y) is created from y) (step 201).

Then, the θ-f plane creating unit 12 subjects the Ie (x, y) to Hough transform using the Hough transform
After creating the ρ plane I (θ, ρ) (step 20
2) I (θ, ρ) is further Fourier-transformed using the Fourier transform unit 15 to create a θ-f plane I (θ, f) (step 203).

Then, the movement amount calculating section 16 calculates the I
(Θ, f) stored in the reference data storage unit 17.
The rotation angle is extracted by comparing with the f plane data R (θ, f) (step 204). Specifically, a normalized correlation coefficient is calculated while shifting R (θ, f) and I (θ, f) in the θ direction, and the angle θ at which the correlation coefficient becomes the maximum is the rotation angle. .

Next, the movement amount calculating section 16 calculates the reference θ-ρ plane data R obtained by correcting the rotation angle based on the rotation angle.
(Θ, ρ) and each θ of the input θ-ρ plane data I (θ, ρ)
, The normalized correlation coefficient is calculated while shifting in the ρ-axis direction, the position of the maximum value of the correlation coefficient is detected, and a ρ cross-correlation image C (θ, ρ) storing the correlation coefficient is created. (Step 205).

Then, the moving amount calculating section 16 performs an inverse Hough transform on the ρ cross-correlation image C (θ, ρ) using an inverse Hough transform processing section 18 to create an inverse Hough transformed image Inv (x, y). ,
The position of the maximum value is obtained to calculate the amount of parallel movement (step 206).

By performing the above series of processing, θ-f
The rotation angle can be obtained at the plane level, and the translation amount can be calculated at the θ-ρ plane level based on the rotation angle. In the above processing procedure, R (θ,
ρ) and R (θ, f) are not described, but can be generated by executing steps 201 to 203 as in the case of the input density image I (x, y).

Next, the procedure for creating an edge direction image shown in step 201 of FIG. 2 will be specifically described.

FIG. 3 is a flowchart showing the procedure for creating the edge direction image in step 201 of FIG.

As shown in FIG. 3, the edge detecting unit 13 first applies a Gaussian Laplacian filter expressed by the following equation to the input density image I (x, y) in order to obtain a zero-cross point. Image Ig (x,
y) is created (step 301).

[0045] Then, it is checked whether or not the pixel of interest of Ig (x, y) is negative and the pixel value of at least one of the four neighboring pixels is positive (step 302), and this condition is satisfied. In this case, a differential operator of Sobel is applied to I (x, y) to obtain an edge strength Em.
Is calculated (step 303).

If the edge strength Em is equal to or greater than a predetermined threshold value, the edge direction Eθ is calculated and stored in the edge direction image Ie (x, y). The process proceeds to pixel processing (step 305).

That is, since the zero cross point has the property that the number of noise edge points increases when the value of σ is reduced to detect a detailed edge of the image, the position of the zero cross point is Noise is removed by applying an operator.

The processing of steps 302 to 305 is repeated while shifting the pixel of interest (step 3).
06) When the processing for all the pixels is completed, the processing is completed.

As described above, the edge detecting unit 13 uses the same method as the edge detecting method disclosed in Japanese Patent Laid-Open No. 5-151352 to detect the edge direction image Ie (x, y).
Has been created.

The Sobel differential operator is:
As shown in FIG. 4, the mask operator 40 in the x direction and y
If the output from the mask operator 40 in the x direction is Mx and the output from the mask operator 41 in the y direction is My, the edge strength Em and the edge direction Eθ are calculated by the following equations.

[0051] Although the case where the Sobel differential operator is used has been described here, various differential operators such as Roberts and Robinson can be applied.

Next, θ-ρ shown in step 202 of FIG.
The procedure for creating plane data will be specifically described.

FIG. 5 is a flowchart showing the procedure for creating θ-ρ plane data in step 202 in FIG.

As shown in FIG. 5, the Hough transform processing unit 14
When the edge direction image Ie (x, y) is input (step 501), it is determined whether or not this Ie (x, y) is an edge point (step 502). Sets the angle variable θ to θ = Eθ−Δθ (step 503). Note that this Δθ is set to an appropriate value based on experiments.

Next, ρ = Xicos θ + Yisin θ is calculated, and I (θ, ρ) is calculated based on the calculated value.
After the plane is incremented (step 504), the angle variable θ is incremented (step 505). Then, as long as the angle variable θ is in the range from Eθ−Δθ to Eθ + Δθ, the processing of the above steps 504 and 505 is repeated (step 506).

Then, such processing is performed by the edge direction image Ie.
Repeat for all pixels of (x, y) (step 50)
7) When all pixels have been processed, the processing ends.

That is, when Ie (x, y) is an edge point, ρ is calculated while displacing the angle variable θ from Eθ−Δθ to Eθ + Δθ, and I (θ corresponding to the combination of θ and ρ is calculated. , Ρ) is incremented to create I (θ, ρ).

Here, the input density image I (x, y)
Has been described, but R (θ, ρ) corresponding to the reference density image R (x, y) has been described.
ρ) can be similarly created.

Next, θ-f shown in step 203 of FIG.
The procedure for creating plane data will be specifically described.

FIG. 6 is a flowchart showing the procedure for generating θ-f plane data in step 203 of FIG.

As shown in FIG. 6, the Fourier transform processing unit 1
5, when the input θ-ρ plane data I (θ, ρ) created by the Hough transform processing unit 14 is input (step 601),
After the angle variable θ is initialized to zero (step 60)
2), FFT, that is, I (θ,
ρ), and its power is calculated as I (θ, f)
(Step 603).

After incrementing the angle variable θ (step 604), it is checked whether or not the angle θ is smaller than θmax (step 605). If the angle variable θ is smaller than θmax, the processing of steps 603 and 604 is repeated. , Θma
The processing ends when x becomes equal to or greater than x.

Here, the input density image I (x, y)
Has been described, but R (θ, f) corresponding to the reference density image R (x, y) has been described.
f) can be similarly created.

Next, the procedure for calculating the rotation angle shown in step 204 of FIG. 2 will be specifically described.

FIG. 7 is a flowchart showing the procedure for calculating the rotation angle in step 204 in FIG.

As shown in FIG. 7, the movement amount calculating section 16
First, when the input θ-f plane data I (θ, f) is input, the reference θ-f plane data R (θ, f) is extracted from the reference data storage unit 17 (step 70).
1) Initialize the angle variable θi to zero (step 7)
02).

Then, R (θ + θi, f) and I (θ,
f) calculate the normalized cross-correlation coefficient Ci (step 70)
3) After the angle variable θi is incremented (step 704), the calculated normalized cross-correlation coefficient Ci is compared with the maximum value Cmax of the normalized cross-correlation coefficient (step 7).
05).

As a result, if Ci is larger than Cmax, Ci is substituted for Cmax to update Cmax, and the angle variable θi at that time is substituted for θk (step 706).

Then, until the angle θi exceeds θmax,
The processes of steps 703 to 706 are repeated (step 707), and the process is terminated when the angle θi becomes equal to or larger than θmax.

That is, the angle when the normalized cross-correlation coefficient is the maximum is stored in the above θk, and this angle θk
The input density image I (x, y) and the reference density image R
The rotation angle is (x, y).

Here, the normalized cross-correlation coefficient Ci is It is calculated by However, θmax is 360 degrees,
fmax is the highest frequency in the image.

Next, the procedure for creating the ρ cross-correlation image C (θ, ρ) shown in step 205 in FIG. 2 will be specifically described.

FIG. 8 is a flowchart showing the procedure for creating the ρ cross-correlation image C (θ, ρ) in step 205 of FIG.

As shown in FIG. 8, the movement amount calculating section 16
If the rotation angle is obtained by the processing shown in FIG.
ρ plane data I (θ, ρ) and reference θ-ρ plane data R
After inputting (θ, ρ) (step 801), the angle variable θ is initialized to zero (step 802).

After substituting −ρmax × 2 into the shift amount Δρ (step 803), R (θ
+ Θk, ρ + Δρ) and the normalized cross-correlation coefficient Ci of I (θ, ρ) are calculated and stored in the ρ cross-correlation image C (θ, ρ) (step 804), and Δρ is incremented (step 805).

Thereafter, it is confirmed whether or not this Δρ is less than ρmax × 2 (step 806), and if it is less than ρmax × 2, the flow shifts to step 804 to repeat the processing of steps 804 and 805. .

On the other hand, if Δρ is equal to or larger than ρmax × 2, the angle variable θ is incremented (step 807), and if the angle variable θ is smaller than θmax, the flow shifts to step 803. (Step 80
8).

That is, the moving amount calculating section 16 calculates I
For each θ of (θ, ρ) and R (θ + θk, ρ), a one-dimensional normalized cross-correlation coefficient is calculated while shifting in the ρ direction to create a ρ cross-correlation image C (θ, ρ). .

The one-dimensional normalized cross-correlation coefficient when the shift amount is Δρ is calculated by the following equation.

[0080] Next, the inverse Hough transform image In shown in step 206 of FIG.
A procedure for creating v (x, y) will be described.

FIG. 9 is a flowchart showing the procedure for creating the inverse Hough transform image Inv (x, y) in step 206 in FIG.

As shown in FIG. 9, the movement amount calculating section 16
When the ρ cross-correlation image C (θ, ρ) is created (Step 901), the angle variable θ is initialized to zero (Step 902), and then ρ is set to −ρmax (Step 9).
03).

This C (θ, ρ) is the C (θ, ρ)
ρ) is determined to be larger than Cmax (step 904), and if larger than Cmax, C (θ, ρ) is substituted for Cmax to update Cmax and The time ρ is substituted for ρk (step 90)
5). Therefore, the value of ρ when Cmax is maximum is stored in ρk.

Next, ρ is incremented (step 906), and it is confirmed whether ρ is smaller than ρmax (step 907). If ρ is smaller than ρmax, the processing of steps 904 to 906 is repeated.

On the other hand, when ρ is equal to or larger than ρmax, C (θ, ρ) is inversely Hough transformed to obtain Inv (x,
y) The value of C (θ, ρk) is added to the straight line on the plane y = − (1 / tan θ) x + ρk / sinθ (step 90)
8).

Then, after incrementing the angle variable θ (step 909), it is confirmed whether or not the angle θ is smaller than θmax (step 910). When the angle variable θ is smaller than θmax, the process proceeds to step 903.

That is, the movement amount calculating section 16 calculates C
A position (θ, ρk) having a maximum value at each θ is detected from (θ, ρ), and an inverse Hough transform is performed at that position to create an inverse Hough transform image Inv (x, y).

Then, the inverse Hough transform image Inv (x,
The position (Xmax, Ymax) of the maximum value of y) is the translation amount.

The inverse Hough transform performed by the inverse Hough transform processing unit 18 means that a point (θ, ρk) on C (θ, ρ) is calculated by ρk = xcosθ + ysinθy = − (1 / tanθ) x + ρk / Sinθ, and converts the image into an inverse Hough transform image Inv
The value of C (θ, ρk) is added to the straight line of (x, y).

Next, a processing result when the image collating apparatus 10 according to the present invention is applied to seal verification will be described.

FIG. 10 is a photograph showing a halftone image displayed on a display when the image collating apparatus 10 shown in FIG. 1 is applied to seal verification.

FIG. 10A is a reference density image R (x, y) of a seal stamp.
y) Reference edge direction image Re obtained by extracting the edge direction from
(X, y) is as shown in FIG.

When the Hough transform is performed on this Re (x, y), reference θ-ρ plane data R (θ, ρ) having a band-like pattern as shown in FIG. When this R (θ, ρ) is Fourier-transformed, FIG.
The reference θ-f plane data R (θ, f) as shown in FIG.

On the other hand, FIG. 10C shows the input density image I (x, y) of the seal, and this input density image I (x, y)
Edge direction image Ie extracted from the edge direction
(X, y) is as shown in FIG.

That is, the input density image I (x, y)
Is different from the reference density image R (x, y) not only in the presence of the enclosing frame, but also in the position and angle of the stamp.

Then, when this Ie (x, y) is subjected to Hough transform, input θ-ρ plane data I (θ, ρ) as shown in FIG. ρ)
Is Fourier-transformed, input θ-f plane data I (θ, f) as shown in FIG.

Here, this I (θ, ρ) is R (θ, ρ)
The reason why the undulation of the band is seen as compared with the above is that the angle and the position where the seal is stamped are different, and the reason that there are many noise components is that there is a line segment such as an enclosing frame.

Next, FIG. 10 (i) shows that I (θ, f) and R
It is a figure which shows the normalized cross-correlation coefficient for every angle (theta) of ((theta), f), and the coefficient value becomes the maximum in (theta) k.

FIG. 10 (j) shows that I (θ, ρ) and R
FIG. 10K is a diagram showing a ρ cross-correlation image C (θ, ρ) calculated while shifting in the ρ direction for each θ of (θ + θk, ρ), and FIG. This is an inverse Hough transformed image Inv (x, y) obtained by detecting a position (θ, ρk) having the maximum value and performing an inverse Hough transform at that position.

Further, FIG. 10 (l) shows a reference edge direction image Re (x,
y), the reference edge direction image Re
FIG. 9 is a diagram illustrating an image in which (x, y) is superimposed on an input edge direction image Ie (x, y). As shown in the figure, since the stamped portion of Re (x, y) and the stamped portion of Ie (x, y) match, it can be seen that the obtained rotation angle and translation amount are accurate.

As described above, in the present embodiment,
The θ-ρ plane creation unit 12 creates I (θ, ρ) and I (θ, f) corresponding to the input image using the edge detection unit 13, the Hough transform processing unit 14, and the Fourier transform processing unit 15, and moves The quantity calculation unit 16 compares this I (θ, f) with R (θ, f) stored in the reference data storage unit 17 to determine the rotation angle at the θ-f plane level, and based on the rotation angle, −
Since the translation amount is calculated at the ρ plane level, the required processing time can be reduced even when the number of edges of the image increases, while suppressing the memory size required for matching.

[0102]

As described in detail above, the first aspect of the present invention sequentially performs edge direction detection, θ-ρ Hough transform and Fourier transform on a reference image to obtain reference θ-ρ plane data and reference θ- The f-plane data is created, and edge direction detection, θ-ρ Hough transform and Fourier transform are sequentially performed on the input image to create input θ-ρ plane data and input θ-f plane data. Then, a rotation angle is calculated based on a positional shift amount between the input θ-f plane data and the reference θ-f plane data, and the reference θ-ρ plane data and the input θ-ρ plane data corrected using the rotation angle are calculated. Since the cross-correlation coefficient is obtained at each θ and the amount of parallel movement is calculated by performing the inverse Hough transform, the following effects can be obtained.

1) It is possible to reduce the memory size required for collation.

2) Even if the number of edges in the image increases, the required processing time can be reduced.

Further, in the second invention, edge extraction, Hough transform and Fourier transform are performed on a reference image, and reference θ-ρ plane data and reference θ-f plane data corresponding to the reference image are previously stored in the reference data. When an input image is input, the input θ-ρ corresponding to the input image is stored in a storage unit.
The plane data and the input θ-f plane data are created, and the rotation angle is calculated based on the positional shift amount between the input θ-f plane data and the reference θ-f plane data stored in the reference data storage means. The reference θ-ρ plane data stored in the storage means is corrected using this rotation angle, and a cross-correlation coefficient is calculated for each of the corrected reference θ-ρ plane data and each θ of the input θ-ρ plane data, Since the translation amount is calculated by performing the inverse Hough transform, the reference image in which the reference data is stored in the reference data storage unit can be quickly collated with the input image.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of an image matching device used in the present embodiment.

FIG. 2 is a flowchart showing a processing procedure of the image matching apparatus shown in FIG.

FIG. 3 is a flowchart showing a procedure of creating an edge direction image performed by the edge detection unit shown in FIG. 1;

FIG. 4 is a diagram showing a Sobel differential operator.

FIG. 5 is a flowchart showing a procedure for generating θ-ρ plane data performed by the Hough transform processing unit shown in FIG. 1;

FIG. 6 is a flowchart showing a procedure for generating θ-f plane data performed by the Fourier transform processing unit shown in FIG. 1;

7 is a flowchart showing a procedure of calculating a rotation angle performed by a movement amount calculation unit shown in FIG. 1;

8 is a flowchart showing a procedure for generating a ρ cross-correlation image C (θ, ρ) performed by the movement amount calculation unit shown in FIG. 1;

FIG. 9 is a flowchart showing a procedure for creating an inverse Hough transform image Inv (x, y) performed by the inverse Hough transform processing unit shown in FIG. 1;

FIG. 10 is a photograph showing a halftone image displayed on a display when the image matching device shown in FIG. 1 is applied to seal verification.

[Explanation of symbols]

10 image collating device 11 image input unit 12 θ
-F plane creation unit, 13 ... edge detection unit, 14 ... Hough transformation processing unit, 15 ... Fourier transformation processing unit, 16 ... movement amount calculation unit, 17 ... reference data storage unit, 18 ... inverse Hough transformation processing unit, 40, 41 ... Sobel operator

Claims (2)

(57) [Claims] 1. An image matching method for matching an image between a reference image and an input image, comprising:
Perform a −ρ Hough transform to create reference θ-ρ plane data,
Fourier transform the reference θ-ρ plane data to obtain a reference θ-f
Create plane data, and set θ with respect to the edge direction image created from the input image.
-Ρ Hough transform is performed to create input θ-ρ plane data,
Fourier transform the input θ-ρ plane data and input θ-f
Creates plane data, calculates a rotation angle based on the amount of position shift between the input θ-f plane data and the reference θ-f plane data, and inputs reference θ-ρ plane data corrected using the rotation angle. Find the cross-correlation coefficient at each θ of the θ-ρ plane data,
An image matching method comprising calculating an amount of parallel movement by performing an inverse Hough transform. 2. An image matching apparatus for performing image matching between a reference image and an input image, wherein an edge direction image is created by differentiating the reference image and the input image, and the edge direction image is Hough-transformed to obtain a θ-ρ plane. After creating the data, data creation means for further Fourier transforming the θ-ρ plane data to create θ-f plane data; reference θ-ρ plane data and reference θ-f plane data corresponding to the reference image A reference data storage means for storing the input θ-f plane data corresponding to the input image and a rotation for calculating a rotation angle based on a positional shift amount between the reference θ-f plane data stored in the reference data storage means Angle calculation means, while correcting the reference θ-ρ plane data stored in the reference data storage means using the rotation angle calculated by the rotation angle calculation means, and the corrected reference θ-ρ plane data Image matching apparatus characterized by a cross-correlation coefficient calculated at each theta force theta-[rho plane data, and a translation amount calculating means for calculating an amount of translation by performing inverse Hough transform.