WO2002003314A1

WO2002003314A1 - Apparatus and method for collating image

Info

Publication number: WO2002003314A1
Application number: PCT/JP2000/004424
Authority: WO
Inventors: Tsuneharu Takeda; Katsuhiko Satoh
Original assignee: Casio Computer Co., Ltd.
Priority date: 2000-07-04
Filing date: 2000-07-04
Publication date: 2002-01-10
Also published as: CN1198232C; KR100443142B1; CN1372674A; KR20020033462A

Abstract

An image collation apparatus has stored therein first image (1a) (registered fingerprint) beforehand. Regarding second image (1b) (fingerprint to be collated) that is input when collation is performed, the apparatus thereby determines the identicalness between this second image and the first image. At this time, the apparatus (3) defines with respect to the first image a plurality of regions each having a predetermined positional relationship with respect thereto. The apparatus (4) searches from the second image an image region having a maximum correlation with the image of each of the respective defined regions. The identicalness (5) between the first image and the second image is determined according to the difference between the positional relationship of each of the respective image regions defined on the first image and the positional relationship of the maximum correlation region that comprises been searched out on the second image.

Description

D E S C R I P T I O N

APPARATUS AND METHOD FOR COLLATING IMAGE

Technical Field

The present invention relates to a technique of collating image data, and more particularly to a technique of collating images, which is suitable for certifying the person himself and which collation is performed using finger print image data.

Background Art Conventionally, in order to prevent illegal access to be had to secret information, the collation of the person himself is carried out through the input of a password, fingerprint, etc.. Regarding a password, there is the likelihood of its being used by another's stealth, whereas a fingerprint is the characterizing feature of the body and so it is less likely that it is used by another's stealth. In view of this, the collation made by fingerprint has been being expected to have a high level of safety.

Conventional collation of the fingerprint was carried out in accordance with the following procedures (1) obtain a fingerprint to be collated by an image reader as multi-gradation image data. (2) binarize the multi-gradation image data. (3) extract the characterizing points such as diverging points or ending points of the wrinkles constituting the fingerprint, from the binarized data. These characterizing points are called "characterizing figures" or "maneusha". (4) compare the positional relationship between the characterizing figures that have been extracted and the characterizing figures of the person's own fingerprint registered beforehand, and thereby determine the identicalness between the both. However, in a case where the valley of the wrinkles of the subject's finger-skin is shallow, or where the skin is too soft, when at the time of input the subject applies his finger onto the glass surface of the image reader, it happens, that the wrinkles are crushed. As a result, there is the possibility that unclear images will only be obtained and so the subject will not be able to be authenticated. Although the proportion of the persons having such finger-skin wrinkles is on the order of several percentages, this value is the one that is as great as the values that cannot be ignored when considering the actual circumstances that 0.1 % is required as the percentage of a failure to collate a fingerprint such as rejecting the right person, erroneously admitting another as being the right person, etc..

Also, it happens that the person has his finger injured or cut. Even, when the finger has been slightly injured, in a case where the place is near the characterizing features, it results that the person himself can not be recognized as being the right person. In the above-described conventional fingerprint collation method, unless the fingerprint image data that is input has its fingerprint clearly imaged out, this data couldn't be accurately collated with the finger's original. Disclosure of Invention

Accordingly, it is an object of the present invention to provide an apparatus for and a method of collating image data, which make it possible to perform fingerprint collation, etc. without being affected by the characteristic of the finger skin, the presence of injuries, etc ..

According to the present invention, there is provided an image collation apparatus for calculating identicalness between first image that is registered beforehand and second image that is input at a time of collation, comprising: a define section which defines a plurality of regions in the first image; a search section which searches out from the second image a region having a maximum correlation with each of the plurality of regions defined by the define section, as a maximum correlation region; and a calculate section which calculates the identicalness between the first image and the second image according to a difference between a positional relationship of each of the plurality of regions defined by the define section and a positional relationship of the maximum correlation region.

According to the present invention, there is provided a computer readable recording medium storing a computer program which calculates identicalness between first image that is registered beforehand and second image that is input at a time of collation, the computer program comprising: a define program which defines a plurality of regions in the first image; a search program which searches out from the second image a region having a maximum correlation with each of the plurality of regions defined by the define program, as a maximum correlation region; and a calculation program which calculates the identicalness between the first image and the second image according to a difference between a positional relationship of each of the plurality of regions defined by the define program and a positional relationship of the maximum correlation region.

According to the present invention, there is provided an image collation method of calculating identicalness between first image that is registered beforehand and second image that is input at a time of collation, comprising: a definition step which defines a plurality of regions in the first image; a search step which searches out from the second image a region having a maximum correlation with each of the plurality of regions defined by the definition step, as a maximum correlation region; and a calculation step which calculates the identicalness between the first image and the second image according to a difference between a positional relationship of each of the plurality of regions defined by the definition step and a positional relationship of the maximum correlation region. Brief Description of Drawings FIG. 1 is a block diagram illustrating the principle of the present invention; FIG. 2 is a block diagram illustrating the construction of an apparatus for collating a fingerprint according to a first embodiment of the present invention;

FIG. 3 is an illustration of recording media that stores therein a fingerprint collation processing program and that is readable by a computer; FIG. 4 is a flow chart illustrating the fingerprint collation processing according to the first embodiment of the present invention;

FIGS. 5A and 5B illustrate fingerprint images that are used to explain the fingerprint collation processing;

FIGS . 6A and 6B show a memory map of a RAM relating to the fingerprint collation processing;

FIG. 7 is a flow chart showing the fingerprint collation processing according to a second embodiment of the present invention;

FIGS. 8A and 8B illustrate fingerprint images that are used to explain the fingerprint collation processing;

FIGS. 9A and 9B illustrate fingerprint images that are used to explain the fingerprint collation processing;

FIGS. 10A and 10B illustrate fingerprint images that are used to explain the fingerprint collation processing; FIGS. 11A and 11B illustrate fingerprint images that are used to explain the fingerprint collation processing;

FIG. 12 is a view that is used to explain a procedure of setting an effective region; and FIG. 13 is a flow chart illustrating the resetting processing of a sub-template. Best Mode for Carrying Out the Invention A preferred embodiment of an apparatus for and a method of collating image according to the present invention will now be described with reference to the accompanying drawings . First Embodiment

FIG. 1 is a block diagram illustrating the principle of the present invention.

A first image data la is image data that is registered in a database beforehand. In a finger collation system, a registered fingerprint corresponds to this first image data.

A second image data lb is image data that has been obtained using an image reader, etc.. In a finger collation system, a fingerprint to be collated corresponds to this second image data.

A defining section 3 defines a plurality of regions over the first image data la. This defining of each region is performed by locating a template having a predetermined shape at a predetermined position on the image data. Each template may be of the same shape and size, or be of a different shape and size. Further, each template may be positioned in a state of its being partly overlapped upon another on the image . An extracting section 4 serves to set an image region, with respect to the second image data, that has the same size and shape as those of a corresponding one of the respective templates that have been positioned on the first image by the defining section 3. It serves to move the image region that has been thus set . It thereby serves to find out on the second image a region having a maximum correlation with the first image data on which the corresponding template is positioned. The region that has been thus found out is called "a maximum correlation region".

A determining section 5 examines the positional relationship of the templates defined by the defining section 3 and the positional relationship of the maximum correlation regions found out by the extracting section . According to the extent by which the both positional relationships differ from each other, the determining section 5 determines the degree of identicalness, or the degree of similarity, between the first image and the second image data.

It is to be noted, regarding the image data, that even binarized data enables determination and that, however, multi-gradation image data is more preferable in order to make a perfect determination of the correlation. Further, it is also possible to set the first image as the fingerprint to be collated and set the second image as the registered fingerprint by interchanging the first image and the second image data with each other.

The above-described functions of the defining section 3, extracting section 4, and determining section 5 can be performed by a computer. An explanation will now be given of an embodiment in the form of a fingerprint collation apparatus that uses a computer. In this case, the fingerprint collation processing as later described is prepared as a program that is to be executed by the computer. Then this program is stored in a storage device of the computer and is executed by the central operation-processing unit.

FIG. 2 is a block diagram illustrating the construction of the apparatus for collating a fingerprint according to the first embodiment of the present invention. This apparatus is constructed of a CPU 11, a storage device 12, a RAM 13, an image reader 14, a display section 15, and an input section 16, which are connected to one another via a bus line 17.

The CPU 11 is a central operation-processing unit, which, in accordance with a control program stored in the storage device 12, controls the entire operation of the fingerprint collation apparatus (hereinafter referred to merely as "the present apparatus") while using the RAM 13 as a work area. The CPU 11 performs the fingerprint collation processing as later described by executing the control program.

The storage device 12 is composed of a ROM, a hard disk, etc.. The storage device 12 previously stores therein the above-described control program that is read out from the CPU 11 immediately after turning on a power source of the present apparatus. In addition, the storage device 12 also previously stores therein subject's registered fingerprint that corresponds to the first image data la illustrated in FIG. 1.

The RAM 13 is a work memory that the CPU 11 uses when executing the above-described control program. The image reader 14 is a device for obtaining a subject' fingerprint to be collated that corresponds to the second image data lb illustrated in FIG. 1. The image reader 14 is, for example, an image scanner, or an image sensor such as a CCD (charge coupled device) . The display section 15 is a display device such as a CRT or a liquid crystal display, which makes a display of the determination result of the collation, which determination result corresponds to a determination result 2 illustrated in FIG. 1.

The input section 16 is an input device such as a keyboard device, by means of which a user of the present apparatus instructs the CPU 11 to obtain a fingerprint to be collated, or to start the fingerprint collation processing as later described.

Incidentally, in the present apparatus, the fingerprint collation program was stored in the storage device 12 beforehand as mentioned above. It can be also arranged that, from a storage medium having a fingerprint collation processing program stored therein, this fingerprint collation processing program be made to be read by a reading device equipped to a computer that corresponds to the storage medium. For example, this fingerprint collation-processing program is temporarily stored in a main memory, and the program is executed by the central processing unit.

In FIG. 3 illustration is made of an example of a computer-readable storage medium that stores therein the fingerprint collation-processing program. As illustrated in this figure, as the storage media, there can be used a storage device 22 such as a ROM or a hard disk device, which is equipped to a computer 21 as an accessory in a built-in form or in a form externally equipped thereto, portable storage media 23 such as a floppy disk, MO (magneto-optical disk) , CD-ROM, or DVD-ROM, a storage device 26 that is an accessory to a program server that is a computer connected to the computer 21 via a network 24, etc.. Next, the control operation that is performed by the CPU 11 will be explained.

FIG. 4 is a flow chart that, of the control process of the present apparatus as a whole realized by the execution of the above-described control program by the CPU 11, illustrates the contents to be processed of a fingerprint collation processing having particular relevance to the present invention. In this figure, the contents of the pieces of processing performed in step Sll and step S14 correspond to the pieces of processing that are performed in the defining section 3 of FIG. 1. Also, the contents of the pieces of processing performed in steps S12, S13 and step S16 correspond to the pieces of processing that are performed in the extracting section 4. Further, the contents of the pieces of processing performed in steps S18 and step S19 correspond to the pieces of processing that are per ormed in the determining section 5.

FIGS . 5A and 5B illustrate examples of the images of fingerprints used to explain the fingerprint collation processing. FIG. 5A illustrates a subject's registered fingerprint that is stored in the storage device 12 beforehand. FIG. 5B illustrates a subject's fingerprint to be collated, which has been sampled by the image reader 14. Also, it is assumed that each of the both images of these figures be a multi-gradation image . FIGS. 6A and 6B show a memory map regarding the fingerprint collation processing, illustrating the state of use of the RAM 13, that prevail when the CPU 11 performs the fingerprint collation processing. The fingerprint collation processing that is performed by the CPU 11 will hereinafter be explained along FIG. 4 while referring to FIGS. 5A, 5B, 6A and 6B, When an instruction to start the fingerprint collation processing is made to the input section 16, first, a main template "mt" is positioned on a registered fingerprint image "A" as illustrated in FIG. 5A, in order to define a rectangular region over the image of the registered fingerprint (step Sll) .

The size of this main template "mt" may be arbitrary so long as the required collation precision of a collation between fingerprints is obtained.

Next, a maximum correlation region MT that corresponds to the region indicated by the main template "mt" is searched from the fingerprint image "B" of the fingerprint to be collated. This searching operation is performed as follows . Setting is performed of a rectangular region having the same shape and size as those of the main template "mt" with respect to the image "B" of the collation fingerprint. Then, this rectangular region is two-dimensionally scanned in units of a pixel. Then, the correlation coefficient between the rectangular region and the region, which is indicated by the main template "mt", is calculated each time the movement is made. The rectangular region that is positioned where the correlation coefficient is maximum is set as being a maximum correlation region MT (step S12). The calculation of the correlation coefficient will be described later.

Here, determination is performed of the coordinates that serve as references for respectively particularizing the positions on the both images that respectively correspond to the main template "mt" and the rectangular region MT, e.g. the coordinates of the respective apexes at the left/upper corners of the both rectangles. These coordinates are set to be mt(xø, yg) and MT(X_Q, O)_/ respectively. They are stored respectively in predetermined areas of the RAM 13 illustrated in FIGS. 6A and 6B (step S13). Next, as illustrated in FIG. 5A, on the image "A" of the registered fingerprint are positioned sub-templates stl to st4 so that the respective apexes of the square main template "mt" may be positioned at their centers, respectively, to thereby define corresponding rectangular regions (step S14). It is noted here that the position of the sub-templates can be arbitrarily made. Also, the number of the sub- templates is not limited to four, and this number may be arbitrary so long as the collation precision of a collation between the fingerprints is obtained as required. In addition, the size of the rectangular region that is defined here as mentioned above may be similarly arbitrary.

In the same way as in the case of extracting the rectangular region MT, regarding each (hereinafter referred to representatively as st-^ (the i = 1, 2, 3, and 4)) of the sub-templates stl to st4, a rectangular region having the same size as that of the sub-template st is set on the image "B" of the collation fingerprint, as illustrated in FIG. 5B. Then, this rectangular region is two-dimensionally scanned in units of a pixel. Then, the correlation coefficient between this rectangular region and the region designated by the sub-template st_^ is calculated each time the movement is made. As a result, the rectangular region that is positioned where the correlation coefficient is maximum is set to be ST-L. Then, determination is performed of the respective coordinates that serve as references for respectively particularizing the positions on the images of the sub- template st^ and the rectangular region ST^, e.g. the respective coordinates of the apexes at the left/upper corners of the both rectangles . These coordinates are set to be stj x^, γ ) and ST-j^X-^, Yj_) . They are stored respectively in predetermined areas of the RAM 13 illustrated in FIGS. 6A and 6B (steps S15 to S17). It is to be noted that when searching a maximum correlation region corresponding to the sub-template stj_, using the position of the already particularized MT as a reference, prediction is done of the position where it is thought that the S ^ will exist. Searching is thereby done of the position and its neighboring area. This sufficiently serves the purpose.

Thereafter, regarding the relative distance between the "mt" and the st-^ and the relative distance between the MT and the ST^, the difference Δ_ between the both relative distances is calculated every value of the i (the i = 1, 2, 3, and 4) in accordance with the following equation (step S19). Δi = |((_Xi - x₀) - (X_± - X₀), (Yi - Y0) - (Yi - Yθ))l

(i)

It is determined whether every calculated value of Δi falls under a predetermined value. If all the values fall under the predetermined value, it is determined that the registered fingerprint and the collation fingerprint coincide with each other. On the other hand, if otherwise, it is determined that the registered fingerprint and the collation fingerprint are out of coincidence. Whereby, the determination result is displayed on the display section 15 (steps S19 and S20). It is to be noted that the predetermined value, which is used here, is the one that is obtained, as a value enable to obtain a desired collation precision, by actually calculating the Δ-^ values from the items of fingerprint image data that have been sampled from a plurality of persons, and in accordance with the distribution of these calculation results . The foregoing processing is the fingerprint collation processing. It is to be noted that, although in the above-explained fingerprint collation processing the shape of the main template "mt" and sub-template sti has been made rectangular, the shape of these templates is not limited to a rectangular one. Namely, each of these templates may be of any given shape. Also, it is preferable that the size and shape of the "mt" and those of the MT as well as the size and shape of the st ^and ^"those of the ST be the same to each other. However, even if there is a small difference between the two, such difference can be permitted so long as a required level of collation precision of one fingerprint with another is obtained.

Also, in the collation determination, in addition to the method of evaluating the Δ _j_ in the above- described processing, it is also possible to adopt various methods such as a determination that is performed based on the difference in shape or area between a figure, that is formed using, for example, the st ^{as n a}P^e^ and a figure, that is formed using the ST-L as an apex. Next, an explanation will now be given of the calculation of the coefficients of correlation that are used in the step S12 or step S16 of the fingerprint collation processing illustrated in FIG. 4. It is to be noted that the calculation of the coefficients of correlation between the rectangular region RA and the rectangular RB is explained here.

First, it is assumed that RA(i, j) and RB(m, n) respectively represent the pixels that are contained respectively in the rectangular region RA and the rectangular region RB, provided. It is also assumed that the sum totals of the pixels contained respec- tively in the rectangular region RA and the rectangular region RB be equal. Also, it is assumed that Xi_j and Y_mn respectively represent signal intensities, each of which is the multi-gradation value indicating the thick and thin of a corresponding pixel. When having generalized these signal intensities and expressed them as Z_pg, the following equation is defined.

In the above equation, N represents the sum total of the pixels contained in the corresponding rectangular region. Also, in the above equation, Σ represents the sum total of the signal intensities about all the pixels contained in the corresponding rectangular region. Namely, the above equation shows the average value of the signal intensities about the pixels contained in the corresponding region.

Next, the following equation is further defined.

The above equation shows the value of root mean square of the signal intensities about the pixels contained in the corresponding rectangular region.

Here, the correlation coefficient C^g between the rectangular region RA and the rectangular region RB can be calculated using the following equation that is expressed using the above-defined equations.

^CAB = < XY > - < X >< Y > /

((< X² > - < X >²)(< Y² > - < Y >²))¹/² (4) where < XY > = (1 / N) (Σ XijY_ran) Calculation is performed of the correlation coefficient between the regions by the use of the above equation.

It is to be noted that in the calculation of the correlation coefficient using the above equation, instead of calculating it using the signal intensities of all pixels within the rectangular region, for example, the following calculation may be performed. Namely, calculation that is performed using only the pixels arrayed on the line constituting any one row within the rectangular region, calculation that is performed using only the pixels contained in part of the rectangular region, or calculation that is performed using the pixels that have been selected by arbitrarily thinning from within the rectangular region may be performed. Even when these calculations are performed, there are no problems if a required level of collation precision of one fingerprint with another is obtained. Using these procedures of calculation, the number of pixels used as an object with respect to which the correlation coefficient is calculated is reduced. As a result of this, the amount of calculation is reduced. Therefore, such procedures of calculation are useful. Also, other methods of calculation of the correlation coefficients may be adopted for the above-described fingerprint collation processing.

Other embodiments of an apparatus for and a method of collating image according to the present invention will be described. The same portions as those of the first embodiment will be indicated in the same reference numerals and their detailed description will be omitted. Second Embodiment In the fingerprint collation, a subject is made to put his finger on a fingerprint reader of the fingerprint collation apparatus and to have his fingerprint to be collated read out by this fingerprint reader. At this time, when the position at which the subject has put his finger gets out of the proper position, this inconveniently causes an erroneous determination result to be produced.

On this account, a method of collation, which, even when at the time of sampling a collation fingerprint the subject's finger somewhat gets out of the proper position, enables the correct collation to be made, will now be explained as a second embodiment. In this embodiment, it is assumed that the main template and the sub-template be both rectangular in shape and the same in size, and that the sub-templates each be positioned so as to contact with their corresponding apex of the main template.

Since the construction of the system is the same as that in the first embodiment, an explanation thereof is omitted. The processing operation will hereinafter be explained using FIGS. 7 to 13. When an instruction to start the fingerprint collation processing is made to the input section 16, first, in step S101, the main template "mt" representing a rectangular region is positioned on the registered fingerprint image "A" , as illustrated in FIG. 8A. The size and the shape of this main template "mt" may be arbitrary so long as a required level of collation precision of one fingerprint with another is obtained. It is assumed now that ( xQ _f yn ) represents the coordinates for particularizing the position on the registered image "A" of the main template "mt" that has been defined here as mentioned above, here the coordinates of the center of gravity of the main template "mt" that is rectangular.

Subsequently, in step S102, as illustrated in FIG. 8A, on the registered image "A" are positioned the sub-templates sti (i ≥ 1) that define a plurality of rectangular regions that are different from that which is defined by the main template "mt". In an example illustrated in FIG. 8A, it is set that i = 4. The sub- templates sti to st that are four rectangles are defined in the positional relationship illustrated in FIG. 8A in such a way as to surround the main template "mt". The size and the shape of these sub-template st-^ also may be arbitrary so long as a required level of collation precision of one fingerprint with another is obtained. It is assumed now that ( i, yi) represents the coordinates for particularizing the position on the registered image "A" of the sub-template st^ that has been defined here as mentioned above, here the coordinates of the center of gravity of the sub- template s that is rectangular. Next, in step S103, a rectangular region that has the same shape and size as those of the main template "mt" is set on the fingerprint image "B" to be collated. While this rectangular region is being scanned over this image "B", calculation is performed of the correlation coefficient between this rectangular region and the region on the image "A" designated by the main template "mt " , one after another .

In step S104, as a result of the successive calculation of the correlation coefficient, determination is made of the above-described rectangular region, regarding which the correlation coefficient between itself and the main template "mt" is maximum. This rectangular region is set as a region MT. The coordinates for particularizing the position on the image "B" of this region MT, which are the coordinates of the region MT corresponding to the coordinates (XQ, yg) of the main template "mt", are represented as ( XQ , Y(j ) •

Subsequently, in step S105, it is determined whether the region MT belongs to an effective region E that is set on the image "B". Concretely, it is determined whether the coordinates (X()_r o) f°^r particularizing the position of the region MT are included within an range E_{x v} that represents the effective region E. If this determination result is Yes, the flow proceeds to step S107 while if it is No the flow proceeds to step S106.

Here, the "effective region E" represents the range on the image "B", wherein the region MT can exist, and within which it is possible to obtain a maximum correlation region corresponding to the sub-template ^sti' necessary for collation. In this embodiment, the range E_{x v} that is expressed by the following equation is set on the image "B".

^Ex,y = {(x, Y) I x _x ≤ K ≤ β _{κ r γ} ≤ x ≤ β _γ } (5) where α_x = MAX[(XQ - 2) + W2 2 + X2SEARCH, (χ_Q - χ₃) + W3/2 + X3SEARCH] ]3_X = WIDTH - MAX[(xi - xo) + W /2 + xxSEARCH, (χ - x₀) + W /2 + x SEARCH] _γ = MAX[(y₀ - yi) + Hχ/2 + _yιSEARCH, (y₀ - y₂ ) + ^H2^{/2 +} y2^SEARCH] β _γ = HEIGHT - MAX[(y₃ - y₀) + H3/2 + y₃SEARCH,

(Y4 " Yθ) + ^H4/² + Y4SEARCH]

FIG. 8B illustrates a state where, in a case where having defined the main template "mt" and the sub- template st_j^ on the registered image "A" as illustrated in FIG. 8A, the range E_{x γ} of the effective region E is set on the image "B". It is to be noted that in the above equation the and the H represent the width (the length as viewed in the x direction) and the height (the length as viewed in the y direction) of the sub-template sti- The WIDTH and the HEIGHT respectively represent the width (the length as viewed in the x direction) and the height (the length as viewed in the y direction) of each of the image "A" and the image "B" having the same size. Also, the x SEARCH and the y SEARCH represent the ranges for searching on the image "B" the maximum correlation region corresponding to the sub-template sti positioned on the image "A". For each of them, for example, a value that corresponds to four to six pixels or so as taken in units of a pixel of the image data is used. Also, the MAX[p, <ΪJ represents the function for obtaining a larger one of the p and the q value (that in a case where the both are equal is either one of them) . A further explanation is given of the above equation. FIG. 12 is a view illustrating the procedure of setting the effective region E. FIG. 12 is given for the purpose of explaining the _x and the β _x of the above equation (4), and illustrates only the upper side portion of the image "A" . It is to be noted that for brevity of explanation it is assumed that the sub- templates st₃ and st4 be left out of consideration. If referring to the above equation (4) and FIG. 12, it will be easily understood that _x = L^, and j3_x = WIDTH - L2. By setting the effective region E as viewed in the x direction by using the values of the a _x and the β _x that are determined from the above equation, it is possible to say as follows. In a case where the sub-templates st^ and st2 each were defined in such a way as to be positioned being widened in the x direction, the range of the effective region E, i.e. the range within which the region MT can exist in the case where the above-described positional relationship is maintained as is, is narrowed. On the other hand, in a case where the sub-templates st^ and st₂ each were defined in such a way as to be positioned with a small width in the x direction, the range of the effective region E is widened.

Turning back to the explanation of FIG. 7, reference will now be made to the above-described step S105. If in this step S105 it has been determined that the region MT falls out of the effective region E, namely if as in the relationship between FIG. 9A and FIG. 9B the region MT on the image "B" corresponding to the main template "mt" on the registered image "A" falls outside the range E_{x v} of the effective region E, the following procedure is taken. Namely, in step S106, the sub-template st is positioned on the registered image "A" again after correcting the position thereof. The re-positioning of the sub-template st^ is performed in accordance with the procedure illustrated in FIG. 13.

FIG. 13 is a flow chart illustrating the contents to be processed of a, re-positioning processing for repositioning the sub-template. This re-positioning processing is the processing for re-locating the sub-template st^ whose position is particularized by the coordinates (xi, yi) to a position that is particularized by the coordinates (xr , yr ) .

An explanation will now be given of this procedure. First, regarding the x direction, it is determined whether the region MT falls within the range of the effective region E (step S201 and step S203).

As a result of the determination, if the x coordinate Xo for particularizing the position of the region MT is smaller than the α_x that represents the range as viewed in the x direction of the effective region E, since the fingerprint "B" to be collated is excessively displaced to the left, the x coordinate for particularizing the position of the sub-template st is set to be xr = x-^ + ( _x - XQ ) and is moved rightwards (step S202). If the x coordinate XQ is larger than the _x that represents the range as viewed in the x direction of the effective region E, since the fingerprint "B" to be collated is excessively displaced to the right, the x coordinate for particularizing the position of the sub-template st is set to be xr = xi - (Xo - j3_x) and is moved leftwards (step S204) .

Here, in a case where the result of determination is outside the both cases, namely if the value of Xo is larger than the _x and smaller than the β _x, it is set that r_j_ = x-^, namely the position of the sub-template stj_ is maintained to be as is at the original value . Then, the flow proceeds to step S206.

Subsequently, regarding the y direction, it is similarly determined whether the region MT falls within the range of the effective region E (step S206 and step S208).

The processing performed according to this result of determination is the same as the above-described processing that was performed regarding the x direction, If the y coordinate Yo for particularizing the position of the region MT is smaller than the a _γ that represents the range as viewed in the y direction of the effective region E, the y coordinate for particularizing the position of the sub-template st that i-^s after repositioning thereof is set to be yr = yi + ( a _γ - YQ) (step S207). And if the y coordinate Yo is larger than the βy that represents the range as viewed in the y direction of the effective region E, it is set that yr = y - (Y_Q - β _γ) (step S209). In a case where the result of determination is outside the both cases, namely if the value of Y_Q is larger than the _γ and smaller than the β _γ, it is set that yr = y , namely the position of the sub-template sti i-^s maintained to be as is at the original value (step S210).

In accordance with the above-described procedure, re-positioning is performed of all the sub-templates st ^tnat were previously defined on the registered image "A" . FIG. 10A illustrates the result of re-positioning that has been performed in a case where the region MT on the image "B" fell out of the range E_{x y} of the effective region E to the right as illustrated in FIG. 9B. Namely, FIG. 10A illustrates the result wherein, in the case, the sub-templates st^ to st previously defined as illustrated in FIG. 8A have been moved leftwards and thus re-positioned. Also, FIG. 10B illustrates the region MT corresponding to the main template "mt" in FIG. 10A. Since in this case the position of the main template "mt" is not moved, the position of the region MT is not moved from the position illustrated in FIG. 9B. Turn is made back to the explanation of FIG. 7. When the result of the determination processing in step S105 is Yes, namely when the re-positioning of the sub-template st ^1S unnecessary, or after the re- positioning of the sub-template st-^ has been performed through the processing in step S106, the following successive calculations are performed in step S107. Namely, in step S107, a rectangular region that is the same in shape and in size as the sub-template st ^1S set on the image "B". While scanning this rectangular region over this image "B", the correlation coefficient between this region and the sub-template sti ^1S calculated one after another. Although the method of calculation of the correlation coefficient uses here the one that was used in the previously stated step S103, other methods of calculation thereof may be adopted .

In step S108, as a result of the successive calculations of the correlation coefficients, the rectangular region, regarding which the correlation coefficient between itself and the sub-template st ^^as become maximum, is determined. This rectangular region is set to be a region ST .

Subsequently, in step S109, it is determined whether the region ST^ has finished being detected with regard to all the sub-templates stj_. If the determination result is Yes, the flow proceeds to step S110. On the other hand, if the determination result is No, the flow returns to step S107, in that the above-described processing for determining the region STi is repeatedly performed. FIGS. 11A and 11B illustrate, regarding all of the re-positioned sub- templates sti ^{to s}^ _f the results of the corresponding regions ST^ to ST4 that have been obtained, together with the main template "mt" and its corresponding region MT. In step S110, the distribution of the positions of the main template "mt" and the sub-template st i-^{n tne} image data "A" is compared with the distribution of the positions of the region MT and the region S ^ in the image data "B". By this comparison is evaluated the identicalness between the image "A" (the registered fingerprint) and the image "B" (the fingerprint to be collated) . The evaluation on the identicalness that is executed in this step is substantially the same as the processing that is performed in step S19 of FIG. 4 and that was explained as the first embodiment. It is to be noted that, when evaluating the identicalness, it is possible to exclude the main template "mt" and the region MT from the objects to be evaluated. It is also possible, with this exclusion, to perform evaluation through only the comparison of the distribution of the positions of the sub-template st and the distribution of the positions of the region ST - The fingerprint collation processing that has up to here been explained is executed through the operation of the CPU 11, and the collation of the fingerprints is thereby performed in the present apparatus .

Additionally, although in the above-described embodiments of the present invention each of the main template "mt", sub-template st , region MT, and region ST^ has been made to be of a rectangular shape, the shape of these templates and regions is not limited to being rectangular. The shape may be any desired one. Also, it is preferable that the size and shape of the main template "mt" and those of the region MT be the same and the size and shape of the sub-template st and those of the region ST be the same. However, even when some difference more or less exists between the both, such difference is permitted so long as a required level of collation precision of one fingerprint with another is obtained. Also, the relationship of position of the main template "mt" relative to the sub-template st ^{or v}i^ce versa, the both templates being defined beforehand, has been determined, in each of the above-described embodiments of the present invention, as follows. Namely, the both templates have been positioned in such a positional relationship wherein the main template "mt" and the sub-template st contact with each other as illustrated in FIG. 8A. However, the both templates may also be positioned and thereby defined in such a positional relationship wherein the two partially overlap one upon the other, or in such a positional relationship wherein the two are completely separated from each other.

Further, in the sub-template re-positioning processing, which is executed in a case where the region MT falls out of the effective region E and whose contents to be processed are illustrated in FIG. 13, each of the sub-templates st has been re-positioned by being moved by the same distance. However, considering in which direction of the upper, lower, leftward, and rightward directions on the image "B" within the effective region E the region MT has gotten out of this effective region E, it may be also arranged to make different the movement distance of one sub-template stj from the original position thereof to the repositioning thereof from the movement distance of another in correspondence with that direction. A concrete explanation will now be given of a case where determining the coordinates (x -_L, yr-_j for particularizing the position, after re-positioning, of every one of the sub-templates st^ to st that were defined as illustrated in FIG. 8A, by way of example. First, regarding the x coordinates, the r is determined from the relationship of the x coordinate XQ for particularizing the position of the region MT with the above-described a _x and the jS _x that represent the range in the x direction of the effective region E, as follows . When XQ < _x ri = x_i + (α_x - X₀) (i = 2, 3) xri = x^ + (α_x - XQ) X U (i = 1, 4) where u represents a constant that is in the range of 0 ≤ u < 1. When _x ≤ X₀ ≤ β _x xri = X-_L

ri = xi - (X₀ - β _x) (i = 1, 4) xri = x_± - (X₀ - _x) X u (1 - 2, 3) where u represents a constant that is in the range of 0 ≤ u < 1.

Next, regarding the y coordinates, the yr is determined from the relationship of the x coordinate Yo for particularizing the position of the region MT with the above-described a _γ and the β _γ that represent the range in the y direction of the effective region E, as follows .

When YQ < a _γ yri = y_± + ( _γ - Y₀) (1 = 1, 2) yri = y_± + ( _γ - Y₀) X v (i = 3, 4) where v represents a constant that is in the range of 0 ≤ v < 1.

Claims

When a γ ≤ Y0 ≤ β γ yr-L = y± When β γ < Y0Yri = Yi - (Yθ - 0y) (i = 3, 4) y i = y± - (Y0 - β γ) X v (i = 1, 2) where v represents a constant that is in the range of 0 ≤≥ v < 1.Additionally, in a case where performing the repositioning of the sub-template in accordance with the above equations, particularly if the re-positioning is performed under the conditions of u = v = 0, the following merit will be obtained. Namely, if so, even when the region MT is outside the effective region E, it becomes possible, according to the positional relationship of the both regions , to perform the fingerprint collation without re-locating part of the sub-templates. Namely, it becomes possible to do so with those sub-templates being kept as were when they were initially defined. In such a case, the period of time that is required for the re-positioning processing of the sub-template is shortened, with the result that it is possible to shorten the processing time period needed for the fingerprint collation processing. C L A I M S

1. An image collation apparatus for calculating identicalness between first image that is registered beforehand and second image that is input at a time of collation, comprising: a define section configured to define a plurality of regions in said first image; a search section configured to search out from said second image maximum correlation regions each having a maximum correlation with each of the plurality of regions defined by the define section; and a calculate section configured to calculate the identicalness between said first image and said second image according to a difference between a positional relationship of each of the plurality of regions defined by the define section and a positional relationship of each of the plurality of the maximum correlation regions searched by said search section.

2. The image collation apparatus according to claim 1, wherein said define section defines one reference region and a plurality of sub-regions each having a predetermined positional relationship with the reference region; said search section initially searches from said second image a first maximum correlation region having a maximum correlation with the reference region, and then searches a second maximum correlation region with regards to each of the sub-regions , according to a predetermined positional relationship with reference to the first maximum correlation region.

3. The image collation apparatus according to claim 2, further comprising: a detector configured to detect whether the first maximum correlation region is in a predetermined relationship to said second image; and a re-define section configured to define again the plurality of regions that have been defined by said define section when said detector detects that the first maximum correlation region is in the predetermined relationship to said second image .

4. The image collation apparatus according to claim 3, wherein: said detector detects that the first maximum correlation region has been displaced from said second image by an amount that is larger than a predetermined amount; and said re-define section moves the sub-regions based on a displacement amount of the first maximum correlation region and thereby re-defines the plurality of sub-regions when it is detected that the first maximum correlation region has been displaced from said second image by the amount that is larger than the predetermined amount .

5. The image collation apparatus according to claim 1, wherein said first image is a subject's fingerprint image and said second image is a fingerprint image to be collated.

6. The image collation apparatus according to claim 1, wherein said search section sets on said second image regions corresponding to the plurality of regions in said first image defined by said define section, calculates a correlation coefficient by using data of the plurality of regions defined by said define section and data of the plurality of the corresponding regions while moving the corresponding regions over said second image, and thereby searches for the maximum correlation region, the correlation coefficient of which becomes maximum.

7. A computer readable recording medium storing a computer program which calculates identicalness between first image that is registered beforehand and second image that is input at a time of collation, the computer program comprising: a define program which defines a plurality of regions in said first image; a search program which searches out from said second image, maximum correlation regions each having a maximum correlation with each of the plurality of regions defined by the define program; and a calculation program which calculates the identicalness between said first image and said second image according to a difference between a positional relationship of each of the plurality of the maximum correlation regions searched by said search program.

8. The recording medium according to claim 7, wherein said definition program defines one reference region and a plurality of sub-regions each having a predetermined positional relationship with the reference region; said search program searches from said second image a first maximum correlation region having a maximum correlation with the reference region, and a program code which searches a second maximum correlation region with regards to each of the sub- regions, according to a predetermined positional relationship with reference to the first maximum correlation region.

9. The recording medium according to claim 8, further comprising: a detection program which detects whether the first maximum correlation region is in a predetermined relationship to said second image; and a re-define program which defines again the plurality of regions that have been defined by said define program when said detector program detects that the first maximum correlation region is in the predetermined relationship to said second image.

10. The recording medium according to claim 9, wherein: said detection program detects that the first maximum correlation region has been displaced from said second image by an amount that is larger than a predetermined amount; and said re-define program moves the sub-regions based on a displacement amount of the first maximum correlation region and thereby re-defines the plurality of sub-regions when it is detected that the first maximum correlation region has been displaced from said second image by the amount that is larger than the predetermined amount .

11. The recording medium according to claim 7, wherein said first image is a subject's fingerprint image and said second image is a fingerprint image to be collated.

12. The recording medium according to claim 7, wherein said search program sets on said second image regions corresponding to the plurality of regions in the first image defined by said define program, calculates a correlation coefficient by using data of the plurality of regions defined by said define program and data of the plurality of the corresponding regions while moving the corresponding regions corresponding to regions over the second image and thereby searches for the maximum correlation region, the correlation coefficient of which becomes maximum.

13. An image collation method of calculating identicalness between first image that is registered beforehand and second image that is input at a time of collation, comprising: a definition step configured to define a plurality of regions in said first image; a search step configured to search out from said second image a region having a maximum correlation with each of the plurality of regions defined by the definition step, as a maximum correlation region; and a calculation step configured to calculate the identicalness between said first image and said second image according to a difference between a positional relationship of each of the plurality of regions defined by the definition step and a positional relationship of each of the plurality of the maximum correlation regions searched by said search step.

14. The image collation method according to claim 13, wherein said definition step defines one reference region and a plurality of sub-regions each having a predetermined positional relationship with the reference region; said search step initially searches from said second image a first maximum correlation region having a maximum correlation with the reference region, and then searches a second maximum correlation region with regards to each of the sub-regions , according to a predetermined positional relationship with reference to the first maximum correlation region.

15. The image collation method according to claim 14, further comprising: a detection step configured to detect whether the first maximum correlation region is in a predetermined relationship to said second image; and a re-definition step configured to define again the plurality of regions that have been defined by said definition step when said detection step detects that the first maximum correlation region is in the predetermined relationship to said second image.

16. The image collation method according to claim 15, wherein: said detection step detects that the first maximum correlation region has been displaced from said second image by an amount that is larger than a predetermined amount; and said re-definition step moves the sub-regions based on a displacement amount of the first maximum correlation region and thereby re-defines the plurality of sub-regions when it is detected that the first maximum correlation region has been displaced from said second image by the amount that is larger than the predetermined amount.

17. The image collation method according to claim 13, wherein said first image is a subject's fingerprint image and said second image is a fingerprint image to be collated.

18. The image collation method according to claim 13, wherein said search step sets on said second image regions corresponding to the plurality of regions in said first image defined by said definition step, calculates a correlation coefficient by using data of the plurality of regions defined by said definition step while moving the corresponding regions over the second image, and thereby searches the maximum correlation region, the correlation coefficient of which becomes maximum.