JPH06149980A - Fingerprint dictionary registration processing system - Google Patents

Abstract

PURPOSE:To generate a fingerprint dictionary low in similarity to a registered fingerprint dictionary. CONSTITUTION:A feature selection part 2 selects prescribed number of fingerprint feature points from plural fingerprint feature points according to a prescribed rule. A correlation decision part 3 finds correlation between a selected fingerprint feature point and registered dictionary data 5 by comparing. When the degree of correlation shows a value over a prescribed value as a result of comparison at the correlation decision part 3, the feature selection part 2 performs the selection again, and the correlation decision part 3 finds the correlation by performing the comparison of the fingerprint feature points, and a registration processing part 10 performs processing according to the degree of correlation found again.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fingerprint dictionary registration processing method, and more particularly, to a fingerprint dictionary registration processing method which is registered for fingerprint verification prior to verification.

With the introduction of computers into a wide range of social systems in recent years, there has been a growing interest in system security. Many questions have been raised from the perspective of ensuring security for the ID cards and passwords that have been used so far as means for verifying the identity when entering a computer room or using a terminal. On the other hand, fingerprints are considered to be the most effective means of personal identification because they have the two major characteristics of "universality" and "lifetime immutability", and many simple personal verification systems using fingerprints have been proposed. ing.

[0003]

2. Description of the Related Art Since a fingerprint collation system normally handles a fingerprint as an image, it is composed of an input device for converting the fingerprint into image data and a collation device for collating the registered image data with the input fingerprint image data. To be done. The collation device performs collation by observing the distribution of the characteristic points (branch points, end points) of the fingerprint and the matching of the shapes.

That is, human fingertips have fine irregularities.
A series of convex parts is called a ridge, and the ridge forms various patterns unique to an individual. If you follow the ridge, you will hit a point that divides into two (branch point) or a dead end point (end point). Since the distributions of branch points and end points differ from person to person, these points are called fingerprint feature points and are used as a powerful means for identifying an individual.

In the fingerprint collation, it is possible to realize high speed and to flexibly cope with the difference in how the force is applied at the time of fingerprint input and the partial expansion and contraction of the fingerprint image caused by moving the glass with the finger attached. It becomes an issue. On the other hand, the present inventors
We are proposing a fingerprint registration and matching method called the "moving window method".

In this method, first, a fingerprint dictionary for fingerprint matching is registered prior to matching. At the time of this registration processing, the binary image of the fingerprint image (the pattern of ridges or valleys of the fingerprint) is thinned to extract the characteristic points. Then, the binary image in the vicinity of each feature point is cut out as a "window". At this time, it is desirable to match the coordinates of the extracted feature points with the center coordinates of the “window”. For the dictionary information, 2 in "window"
Value image, relative positional relationship between "window" and "window",
Register the type of feature points in the "window", etc. At this time, among a plurality of “windows”, one stable “window” that is relatively close to the center of the fingerprint is selected as a “registration window”, and the rest is defined as a “verification window”. .

Subsequently, in the matching process, as shown in FIG. 9, pattern matching between the "positioning window" and the input image (binary image) is performed to correct the difference between the input positions at the time of registration and at the time of matching. To do. Next, the peripheral “matching window” is moved in parallel according to the displacement amount of the “positioning window” to perform pattern matching. By the way, since a human finger is very soft, the input fingerprint image is partially expanded or contracted due to a difference in how force is applied at the time of pressing. For this reason, each “matching window” does not show a good match with the input image at the position corresponding to it simply by parallel translation. Therefore, each "matching window" is independently and slightly moved two-dimensionally to determine whether the input image matches or does not match. Thus, more than a predetermined number (threshold value) of "collation windows"
If is passed, the person is identified.

[0008]

According to the above method,
The distribution of feature points of fingerprints and the conformity / non-coincidence of the shapes are determined for a plurality of “matching windows”, and whether or not the “matching windows” equal to or larger than a threshold value pass is used as a criterion for identifying a person.

However, since the number of windows that coincide with each other is allowed, the following problems occur.
That is, if a fingerprint dictionary having a high degree of similarity between different (other's) fingerprints is created at the time of registration, there is a possibility that a collation may occur in the user group. In particular, in the case of a fingerprint with a flow pattern, the similarity of the fingerprint patterns is large, so that erroneous matching is likely to occur. Mismatching is also likely to occur when fingerprint feature ridges are blurred or pseudo feature points such as bridges between ridges due to sweat are included.

The erroneous collation rate can be reduced by stricter (increasing) the pattern matching threshold value, but in that case, the tolerance for the deformation of the finger is lowered, so that the collation rate is reduced. Is also unrealistic.

It is an object of the present invention to provide a fingerprint dictionary registration processing method capable of creating a fingerprint dictionary having a low degree of similarity with a registered fingerprint dictionary.

[0012]

FIG. 1 is a block diagram of the principle of the present invention, showing a fingerprint collation processing device according to the present invention. The fingerprint collation processing device includes a registration processing unit 10 including a feature extraction unit 1, a feature selection unit 2, and a correlation determination unit 3. The feature extraction unit 1 extracts a plurality of fingerprint feature points from the input fingerprint image. The feature selection unit 2 selects a fingerprint feature point from a plurality of fingerprint feature points according to a predetermined rule. The correlation determination unit 3 compares the selected fingerprint feature point with the registered dictionary data 5 to obtain the correlation.

The fingerprint dictionary 20 includes one or more registered dictionary data 5. In this fingerprint collation processing device, when the result of the comparison in the correlation determination unit 3 indicates that the degree of correlation is smaller than a predetermined value, the registration processing unit 10 newly creates dictionary data created based on the selected fingerprint feature points. Register in the fingerprint dictionary 20 at. This dictionary data is the newly registered dictionary data 6.

On the other hand, as a result of the comparison in the correlation judging section 3,
If the degree of correlation is greater than or equal to a predetermined value, the feature selection unit 2
Performs the selection again, the correlation determination unit 3 again compares the selected fingerprint feature points to obtain the correlation, and the registration processing unit 10 performs the process according to the re-obtained correlation degree.

[0015]

When the correlation between the selected fingerprint feature point and the registered dictionary data 5 is a predetermined value or more, a predetermined number of fingerprint feature points are selected again from the extracted plurality of fingerprint feature points. cure. Then, similar processing is repeated for the fingerprint feature point selected again.

Therefore, the registered dictionary data 5 must be registered as the newly registered dictionary data 6 in the fingerprint dictionary 20.
Only the dictionary data based on the selection (combination) of fingerprint feature points having a correlation degree between and is smaller than a predetermined value. As a result, the fingerprint dictionary 20 in a certain user group can be configured with only those having a small degree of correlation between the dictionary data.

As a result, the probability of erroneous collation in fingerprint collation using this fingerprint dictionary 20 can be reduced.
In addition, it is possible to reduce the probability of erroneous matching while maintaining the tolerance for the influence of fingerprint pattern, sweat, finger deformation, and the like.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The fingerprint collation processing apparatus shown in FIG. 1 includes a registration processing unit 10 for creating a fingerprint dictionary 20 (registration of dictionary data in the fingerprint dictionary 20) prior to collation for fingerprint collation, and a user's use. A fingerprint dictionary 20 that stores dictionary data composed of fingerprint images corresponding to the person ID, and a matching processing unit 30 that performs fingerprint matching using the fingerprint dictionary 20 are provided.

This fingerprint collation processing device first operates in the registration mode. That is, after the grayscale image of the fingerprint to be registered read by the fingerprint sensor is binarized and thinned, the registration processing unit 10 extracts fingerprint feature points of the registered fingerprint from the binarized and thinned image. Then, the extracted fingerprint feature points are registered in the fingerprint dictionary 20 in association with the user ID.

Then, when the dictionary data based on the fingerprint feature points of the fingerprint is registered according to the registration mode, the operation is then performed in the matching mode. That is, after the grayscale image of the fingerprint to be collated read by the fingerprint sensor is binarized, the collation processing unit 30 is based on the binarized image and the fingerprint feature points read from the fingerprint dictionary 20 according to the user ID. Identity verification is performed by comparing with dictionary data.

Actually, the (grayscale) image of the fingerprint output as analog data from the fingerprint sensor is temporarily stored in the frame memory of the image input section and then binarized by the binarization circuit to be converted into binary data. Stored in digitized memory. This binarized data (image) is processed by the collation processing unit 3 in the collation mode.
0 to the registration processing unit 10 in the registration mode.

In the collation mode, the collation processing section 30 performs collation using the fingerprint dictionary 20 and sends a confirmation signal as described above. In response to this confirmation signal, for example, an unlocking signal is sent from the CPU (central processing unit) to the door control unit to open the door.

Prior to this collation, the fingerprint dictionary 20 is created. That is, in the registration mode, dictionary data about the fingerprint of the user is created together with the user ID, and the fingerprint dictionary 2
Registered as 0. The registration processing unit 10 performs this creation and registration.

Regarding the processing performed by the registration processing unit 10,
It will be described with reference to FIG. The fingerprint image for registration is read by the fingerprint sensor and input (step 1). That is, the fingerprint sensor outputs a (gray) image of the fingerprint. An example of this grayscale image is shown in FIG. This image is stored in the frame memory.

The fingerprint image in the frame memory is binarized by the binarization circuit (step 2) and stored in the binary memory. Since the mode is the registration mode, the processing in the registration processing unit 10 is executed.

That is, the feature extraction unit 1 thins the fingerprint image in the binary memory and stores it in the thinning memory (step 3). From the stored thinned image, the branch point, Fingerprint feature points such as end points are extracted (step 4).

An example of fingerprint feature points is shown in FIG. The fingerprint feature points shown in FIG. 3 (B) are extracted from the fingerprint image of FIG. 3 (A). The fingerprint feature point is treated as a “window (window image)” obtained by cutting out an image in the vicinity of a branch point or an end point into a square of a predetermined size. The coordinates of the branch point or the end point and the center coordinates of the "window" are matched.

In FIG. 3B, the window "0",
"1" is an end point, and windows "2" to "6" are branch points. A window "0" near the center of the entire fingerprint is a "registration window" and the other windows are "verification windows". Further, in actuality, the number of windows is made larger than that shown in the figure, for example, 20.

Next, the feature selection unit 2 selects a predetermined number of fingerprint feature points according to a predetermined rule, and cuts out an image around the fingerprint feature points (that is, the window image shown in FIG. 3B) (step 5). ).

The predetermined rule is set in advance and is selected, for example, in a spiral form from the center of the fingerprint image to the outside. It is set such that selection is made in order from the upper right of the image. The predetermined number is, for example, 15 and is a sufficient number for fingerprint collation, that is, a predetermined number for forming the fingerprint dictionary 20 (dictionary data 5 or 6), and the number of extracted fingerprint feature points. Is less than or equal to.

The feature selection unit 2 registers the cut out 15 window images in a predetermined memory area to create a temporary dictionary.
At this time, the positional relationship of each window image is also registered at the same time. In addition,
In step 4, only the fingerprint feature points are extracted, and only the fingerprint feature points selected in step 5 are cut out and registered. In this case, an image may be cut out from all the fingerprint feature points extracted in step 4 and registered in the predetermined memory area, and only the selection may be performed in step 5 to create a temporary dictionary.

In any case, the number of extractions> the number of selections, but the number of extractions = the number of selections may be satisfied. When the number of extractions = the number of selections, the extraction itself is performed and repeated according to the above-mentioned predetermined rule.

Next, the correlation determination unit 3 compares the temporary dictionary for the new fingerprint with the registered dictionary data 5 for the already registered fingerprint and takes the correlation (step 6).
With respect to the registered dictionary data 5, all the items registered in the fingerprint dictionary at that time are sequentially read, and the correlation is checked for each of them.

Specifically, as shown in FIG. 4, the entire fingerprint image of the registered fingerprints (the image of FIG. 3A),
The 15 window images (corresponding to the image in FIG. 3B) of the new fingerprint are pattern-matched to obtain the correlation. For this reason, each registered dictionary data 5 includes not only 15 window images but also the entire image of the fingerprint image.

This pattern matching is performed by deforming the registered dictionary data 5 by an amount that deforms each window image corresponding to the distortion of the finger at the time of fingerprint collation. As a result, it is possible to accurately obtain the degree of correlation in response to finger distortion and the like.

Next, as a result of the correlation, the correlation determining unit 3 checks whether or not the degree of correlation is greater than or equal to the threshold value (step 7).
The threshold value is predetermined and is set to, for example, "10".
Note that, in FIG. 4, only the window “0” is the same among the seven windows, and thus the degree of correlation is “1”. For example, if the threshold value is set to “10”, 10 out of 15 window images will be displayed.
When they are the same, the degree of correlation is "10", which is equal to the threshold value.

When the degree of correlation is not greater than or equal to the threshold value (“10”) (smaller than the threshold value (“10”)), the registration processing unit 10 sets the temporary dictionary as the newly registered dictionary data 6 to the fingerprint dictionary 20. Register (step 8). At this time, user ID
And the entire image of the fingerprint image is also registered. Therefore, the newly registered dictionary data 6 is made to correspond the fifteen window images, which were temporary dictionaries, and the entire image of the fingerprint image to the user ID.

The newly registered dictionary data 6 will be used as the registered dictionary data 5 when registering another fingerprint next time. The degree of correlation is a threshold value (“1
0 ") or more, the correlation determination unit 3 notifies the feature selection unit 2 of this fact, and the feature selection unit 2 requests the fingerprint feature points to be selected again (step 9).

The feature selection unit 2 that has received the request checks, prior to the selection again, whether or not the selection has been completed for all the combinations (for example, ₂₀ C ₁₅ ) of window image selections (step 1).
0).

When the processing is completed, the registration processing unit 10 does not register the fingerprint (temporary dictionary or the like) of the user (user ID) in the fingerprint dictionary 20 (step 11). That is, no matter which of the fingerprint feature points is selected, it is not registered because it becomes similar to the registered dictionary data 5 and causes erroneous collation.

If not completed, the feature selection unit 2 selects, for example, 15 fingerprint feature points according to a predetermined rule so as to form an unselected combination, and cuts out the window image as a temporary dictionary. . The correlation determination unit 3 takes a correlation between the reselected temporary dictionary and the registered dictionary data 5 and checks whether the degree of correlation is greater than or equal to a threshold value. That is, step 7 (specifically step 5) and the subsequent steps are repeated.

FIG. 5 shows another registration process executed by the registration processing unit 10. In this registration processing, the registration processing unit 10 divides the input fingerprint image into a predetermined number of blocks, and the feature extraction unit 1 extracts the number and types of fingerprint feature points existing in the block for each block. The feature selecting unit 2 extracts the type of the fingerprint feature point selected and the block in which it exists, and the correlation determining unit 3 determines the correlation between the selected fingerprint feature point and the registered dictionary data 5 as the type of the fingerprint feature point. The block and are used for the calculation.

That is, after the execution of steps 1 to 4, the registration processing unit 10 causes the fingerprint image (shown in FIG. 3 (A)).
Is divided into a predetermined number, for example, 9 blocks (step 12).

Next, the feature extraction unit 1
The fingerprint feature points existing in the block are extracted, the number of fingerprint feature points and the type of each fingerprint feature point are extracted and stored in a predetermined memory area (step 13). The types of fingerprint feature points are different, such as branch points and end points.

Next, the feature selection unit 2 selects fingerprint feature points (for example, 15 points) according to a predetermined rule, and extracts (reads) the type of the selected fingerprint feature points and the existing blocks to determine them. Stored in the memory area (step 1
4).

Next, the correlation judgment unit 3 calculates the correlation between the registered dictionary data 5 and the fingerprint feature points (step 15). Specifically, for one selected fingerprint feature point, the existing block (position thereof) and its type are known (by step 14), and the corresponding block position in the entire image of the fingerprint image of the registered dictionary data 5 is detected. Check whether or not there is a feature point of the type corresponding to. In addition, when it exists, the degree of correlation is set to "1". The above process is repeated for each fingerprint feature point.

Next, step 16 similar to step 7 is executed, and when the correlation is not greater than or equal to the threshold value, the registration processing section 10 cuts out an image around the fingerprint feature point selected at that time ( In step 17), these are registered in the fingerprint dictionary 20 as newly registered dictionary data 6 in the same manner as in step 8 (step 18).

The types of fingerprint feature points and existing blocks are both coded in the feature extraction unit 1 according to a predetermined rule and held in the form of a table for each fingerprint feature point.

Correspondingly, in order to facilitate correlation, the registered dictionary data 5 of the fingerprint dictionary 20 is also provided with a table in which the type and the existing block are coded for each fingerprint feature point. To be That is, in step 18, the table created and used for the registration of the newly registered fingerprint is registered as it is as a part of the newly registered dictionary data 6. By comparing and correlating with codes in this way, the processing can be made extremely simple and the processing time can be shortened as compared with the case of comparing the images as shown in FIG.

Furthermore, the size of each block may be set so that the vertical and horizontal directions of the fingerprint image are equally divided into n parts, but the size of each block is adjusted so as to cope with the distortion when the finger is pressed against the fingerprint sensor. It is desirable to make the matching window match the range to be scanned during pattern matching.

As a result, the distortion can be dealt with, and the number of fingerprint feature points in each block can be set to either 0 or 1, and the processing time can be shortened. FIG. 6 shows another registration process executed by the registration processing unit 10.

In this registration processing, the registration processing unit 10 classifies the patterns of the input fingerprint image, and the correlation determination unit 3
The correlation is calculated only for the registered dictionary data 5 having the same pattern as the pattern.

That is, after the execution of steps 1 and 2, the registration processing unit 10 causes the fingerprint image (shown in FIG. 3A).
The patterns are classified (step 22). This classification is performed, for example, so that the fingerprint image is classified into eight types of right flow, left flow, vortex, twin, right bag, left bag, columnar, and irregular. Further, it may be classified into five types of right flow, left flow, vortex, columnar and irregular.

Next, steps 3 and 4 are executed, and then steps 12 to 14 are executed. After that, step 23 similar to step 15 is executed, but the registered dictionary data 5 to be correlated is limited to those having the same pattern as the pattern classified in step 22. That is, since it is mainly between the fingerprints of the same pattern that the false collation occurs, the correlation is limited to this and the false collation is prevented in advance. As a result, the number of registered dictionary data 5 to be compared can be reduced and the processing time can be shortened.

The fingerprint pattern is coded in the registration processing unit 10 and held in correspondence with each fingerprint image.
Correspondingly, the code of the fingerprint pattern is also registered in the registered dictionary data 5 of the fingerprint dictionary 20 to facilitate correlation. That is, in step 26 described later, the pattern of the newly registered fingerprint is registered as a part of the newly registered dictionary data 6.

After step 23, steps 24 to 29 are executed, and these are respectively performed in step 1
The steps are the same as steps 6 to 21. The comparison using this pattern is not limited to the case shown in FIG.
It can also be applied to the case shown in (when dividing images into blocks and comparing images to obtain correlation). In this case, the processing time can be shortened and the correlation can be accurately obtained.

FIG. 7 shows another registration process executed by the registration processing unit 10. In this registration processing, the registration processing unit 10 extracts the extending direction of the ridge for each fingerprint feature point selected by the feature selection unit 2, and the correlation determination unit 3 extracts the selected fingerprint feature points and the registered dictionary data. The correlation with 5 is obtained using the type of the fingerprint feature point, its existing block, and its ridge direction.

That is, after executing Step 1 and Step 2, Step 22, Step 3 and Step 4, Step 12 to Step 14 in this order, the registration processing unit 10
However, the feature selecting unit 2 extracts the direction in which the ridge extends for each fingerprint feature point selected in step 14 (step 30). As the direction, for example, eight types of up / down / left / right, diagonally upward and downward to the right, diagonally upward and downward to the left, and downward are used.

The ridge direction is extracted using, for example, the eight types of masks. That is, the direction of the mask that coincides with (or is close to) a certain fingerprint feature point by sequentially overlapping the masks or the direction of the ridge.

After this, step 3 similar to step 23
1, but the direction of the ridge is also used in the correlation. As a result, the correlation can be accurately obtained while shortening the processing time.

The direction of the ridge of the fingerprint feature point is coded in the registration processing unit 10 and held in the form of a table for each fingerprint feature point. Correspondingly, in order to facilitate correlation, the registered dictionary data 5 of the fingerprint dictionary 20 is also provided with a table in which the ridge direction is encoded for each fingerprint feature point. That is, in step 34 described later, the table created and used for registration of the newly registered fingerprint is registered as it is as a part of the newly registered dictionary data 6.

Furthermore, instead of extracting the direction of the ridge, the direction of the valley line may be extracted. Further, when the correlation is obtained by using the direction of the ridge, it may be executed separately from the pattern distinction or the block division.

After step 31, steps 32 to 37 are executed, and these are respectively performed in step 1
The steps are the same as steps 6 to 21. FIG. 8 shows another registration process executed by the registration processing unit 10.

In this registration processing, the registration processing unit 10 extracts the direction of the branch point or the end point of each of the fingerprint feature points selected by the feature selection unit 2, and the correlation determination unit 3 extracts the orientations of the selected fingerprint feature points. The correlation with the registered dictionary data 5 is obtained using the type of the fingerprint feature point, the block in which the fingerprint feature point exists, and the direction of the branch point or end point thereof.

That is, after performing Step 1 and Step 2, Step 22, Step 3 and Step 4, Step 12 to Step 14 in this order, the registration processing unit 10
However, for each of the fingerprint feature points selected in step 14 by the feature selection unit 2, the directions of the branch points and end points are extracted (step 38). As the direction, for example, eight types of up / down / left / right, diagonally upward and downward to the right, diagonally upward and downward to the left, and downward are used.

The extraction of these directions is performed by the same processing as the extraction of the ridge direction. After this, the same step 39 as step 23 is executed, but the orientations of the branch points and the end points are also used when taking the correlation. As a result, the correlation can be accurately obtained while shortening the processing time.

Note that the orientations of the branch points and end points of the fingerprint feature points are coded in the registration processing unit 10 and held in the form of a table for each fingerprint feature point. Correspondingly, in order to facilitate correlation, the registered dictionary data 5 of the fingerprint dictionary 20 is also provided with a table in which the ridge direction is encoded for each fingerprint feature point. That is, in step 42 to be described later, the table created and used for registration of the newly registered fingerprint is registered as it is as a part of the newly registered dictionary data 6.

Further, when the correlation is obtained by using the directions of the branch points and the end points, it may be executed separately from them without being used together with the distinction by the pattern or the block division. After step 39, steps 40 to 45 are executed.
These steps are the same as steps 16 to 21, respectively.

[0069]

As described above, according to the present invention, in the fingerprint dictionary registration processing, by correlating the dictionary data already registered in the fingerprint dictionary with the data of the fingerprint to be registered, Since the data having a small correlation with the already registered dictionary data can be used as the dictionary data of the fingerprint to be newly registered, the probability of erroneous matching can be reduced, and the reliability of the fingerprint matching processing device can be improved. Security can be secured.

[Brief description of drawings]

FIG. 1 is a principle configuration diagram of the present invention.

FIG. 2 is a registration processing flow.

FIG. 3 is an explanatory diagram of a fingerprint image.

FIG. 4 is an explanatory diagram of fingerprint correlation.

FIG. 5 is another registration processing flow.

FIG. 6 is another registration processing flow.

FIG. 7 is another registration processing flow.

FIG. 8 is another registration processing flow.

FIG. 9 is a diagram illustrating a conventional technique.

[Explanation of symbols]

 1 Feature Extraction Section 2 Feature Selection Section 3 Correlation Judgment Section 5 Registered Dictionary Data 6 Newly Registered Dictionary Data 10 Registration Processing Section 20 Fingerprint Dictionary 30 Collation Processing Section

Claims (4)

[Claims] 1. One or more registered dictionary data (5)
A fingerprint dictionary (20) including a fingerprint feature point (1) for extracting a plurality of fingerprint feature points from the input fingerprint image, and a fingerprint feature point is selected from the plurality of fingerprint feature points according to a predetermined rule. The registration processing unit (10) includes a feature selection unit (2) and a correlation determination unit (3) that compares the selected fingerprint feature point with the registered dictionary data (5) to obtain the correlation. If the degree of correlation is smaller than a predetermined value as a result of the comparison in the correlation determination unit (3), the registration processing unit (10) newly creates dictionary data created based on the selected fingerprint feature point. The fingerprint dictionary (2
0), and when the correlation degree is greater than or equal to a predetermined value as a result of the comparison in the correlation determination unit (3), the feature selection unit (2) makes the selection again, and the correlation determination unit (3) 3) again compares the reselected fingerprint feature points to obtain the correlation, and the registration processing unit (10)
The fingerprint dictionary registration processing method, wherein the processing is performed according to the re-obtained correlation degree. 2. The registration processing unit (10) divides the input fingerprint image into a predetermined number of blocks, and the feature extraction unit (1) includes the number and types of the fingerprint feature points existing in the block. For each block, the feature selection unit (2) extracts the type of the fingerprint feature point selected and the block in which it exists, and the correlation determination unit (3) compares the selected fingerprint feature point with the selected fingerprint feature point. The fingerprint dictionary registration processing method according to claim 1, wherein the correlation with the registered dictionary data (5) is obtained by using the type and block of the fingerprint feature point. 3. The registration processing unit (10) classifies patterns of the input fingerprint image, and the correlation determination unit (3)
The fingerprint dictionary registration processing method according to claim 1 or 2, wherein the correlation is obtained only for the registered dictionary data (5) having the same pattern as the pattern. 4. The fingerprint dictionary (2
0) The fingerprint dictionary registration processing method according to claim 1, wherein more than a predetermined number is extracted to form 0).