JP2006127419A - Image processor, image processing method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To propose an image processor for improving processing efficiency. <P>SOLUTION: This image processor for processing image data acquired by picking up the image of an authentication object is provided with a conversion means for converting image data into data expressing the luminance and locations of pixels being non-zero elements, a work memory for temporarily storing data and an arithmetic means for executing arithmetic processing corresponding to the data for the data stored in the work memory, and configured to reduce any data other than the non-zero elements in the image data, and to operate the arithmetic processing corresponding to the data acquired after the reduction of the data. Thus, it is possible to realize an image processor for accelerating the arithmetic processing, and for sharply reducing the use region of the work memory, and for improving the processing efficiency. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an image processing apparatus, an image processing method, and a program, and is suitable for application to an authentication apparatus that authenticates, for example, a blood vessel unique to a living body.

  Conventionally, in this type of authentication device, the registrant's ability to absorb light in the near infrared band is specifically absorbed by deoxygenated hemoglobin (venous blood) or oxygenated hemoglobin (arterial blood) inherent in blood vessels. A blood vessel image obtained by imaging a blood vessel (hereinafter referred to as a registered blood vessel image) is subjected to predetermined processing, and data based on the registered blood vessel image (hereinafter referred to as registered blood vessel image data). This is registered in a predetermined database as an authentication target.

Further, the authentication device images the blood vessel of the collator to be collated with the authentication target at an arbitrary timing, and performs a predetermined process on the blood vessel image obtained as a result of the imaging (hereinafter referred to as a collator blood vessel image). To generate data based on the collator blood vessel image (hereinafter referred to as collator blood vessel image data), and collate the collator blood vessel image data with the registered blood vessel image data registered in the database at this time. Thus, the presence / absence of the person (registered person) is determined (see, for example, Non-Patent Document 1).
Japanese Patent Application No. 2003-242492

  By the way, in such an authentication apparatus, for example, when registering a registered blood vessel image as an authentication target in a predetermined database, a predetermined process is performed on the registered blood vessel image to generate registered blood vessel image data. If the data volume of the registered blood vessel image can be reduced, the processing for the registered blood vessel image can be speeded up, so that the processing efficiency can be improved.

  On the other hand, in such an authentication device, in the case where the collator blood vessel image is collated with the registered blood vessel image, when collating the collator blood vessel image data and the registered blood vessel image data, the collator blood vessel image data and If the data amount of the registered blood vessel image data can be reduced, the collation process of the collator blood vessel image data and the registered blood vessel image data can be speeded up as in the case described above, so that the processing efficiency can be improved. It is considered possible.

  On the other hand, in such an authentication device, in the case where a collator blood vessel image is sequentially collated with a plurality of registered blood vessel images, as shown in FIG. 13, collator blood vessel image data CB and a database based on an imaging process are used. A plurality of registered blood vessel image data RB read from (not shown) is stored in the cache memory CM, and a plurality of registered blood vessel image data until the registered blood vessel image data RB corresponding to the collator blood vessel image data CB is detected. The RBs are sequentially verified.

  However, when the storage area of the cache memory CM is small, for example, the authentication apparatus caches the collator blood vessel image data CB based on the imaging process and the registered blood vessel image data RB1 read from the database (not shown) as shown in FIG. When the registered blood vessel image data RB1 does not correspond to the collator blood vessel image data CB in the memory CM, the registered blood vessel image data RB1 is deleted from the cache memory CM.

  Then, the authentication device collates the registered blood vessel image data RB2 newly read from the database (not shown) with the collator blood vessel image data CB in the cache memory CM, and the registered blood vessel image data RB2 is collated with the collator blood vessel image data CB. When the registered blood vessel image data RB2 is erased from the cache memory CM, the same processing as described above is repeated until the registered blood vessel image data RB corresponding to the collator blood vessel image data CB is detected.

  For this reason, the authentication apparatus obtains the collator blood vessel image data CB and the registered blood vessel image data RB by, for example, deleting the registered blood vessel image data RB and reading new registered blood vessel image data RB from a database (not shown). There is a problem that processing time for collation increases and processing efficiency decreases. In addition, when a large number of such registered blood vessel image data RB are registered, the delay of this collation processing increases, and the processing efficiency further decreases.

  The present invention has been made in consideration of the above points, and proposes an image processing apparatus, an image processing method, and a program capable of improving the processing efficiency.

  In order to solve such a problem, in the present invention, in an image processing apparatus that processes image data obtained by imaging an authentication target, the image data is converted into data representing the brightness and position of a pixel that is a non-zero element. Conversion means, a work memory for temporarily storing data, and a calculation means for executing calculation processing corresponding to the data for the data held in the work memory.

Therefore, since data other than non-zero elements in the image data is reduced and the arithmetic processing corresponding to the data obtained after the reduction is performed, the arithmetic processing can be speeded up, and the work memory The use area can be drastically reduced.

  In the present invention, in the image processing method for processing image data obtained by imaging an authentication target, the first step of converting the image data into data representing the luminance and position of a pixel that is a non-zero element. And a second step of temporarily storing the data in the work memory and a third step of executing a calculation process corresponding to the data for the data held in the work memory.

  Therefore, since data other than non-zero elements in the image data is reduced and the arithmetic processing corresponding to the data obtained after the reduction is performed, the arithmetic processing can be speeded up, and the work memory The use area can be drastically reduced.

  Furthermore, in the present invention, the image data is converted into data representing the luminance and position of the pixel that is a non-zero element for an image processing apparatus that processes image data obtained by imaging an authentication target by a program. A first step, a second step of temporarily holding data in the work memory, and a third step of executing an arithmetic process corresponding to the data for the data held in the work memory It was made to execute.

  Therefore, since data other than non-zero elements in the image data is reduced and the arithmetic processing corresponding to the data obtained after the reduction is performed, the arithmetic processing can be speeded up, and the work memory The use area can be drastically reduced.

  According to the present invention, in an image processing apparatus that processes image data obtained by imaging an authentication target, conversion means for converting image data into data representing luminance and position of a pixel that is a non-zero element, and data Is provided for the data stored in the work memory and the calculation means for executing the calculation processing corresponding to the data for the data stored in the work memory. In addition to reducing the data and performing the arithmetic processing corresponding to the data obtained after the reduction, in addition to being able to speed up the arithmetic processing, the use area in the work memory can be drastically reduced Thus, an image processing apparatus that can improve the processing efficiency can be realized.

  According to the present invention, in the image processing method for processing image data obtained by imaging an authentication target, the first image data is converted into data representing the luminance and position of a pixel that is a non-zero element. A step, a second step for temporarily storing the data in the work memory, and a third step for executing the arithmetic processing corresponding to the data for the data held in the work memory are provided. In addition to reducing data other than non-zero elements in the image data and performing arithmetic processing corresponding to the data obtained after the reduction, in addition to speeding up the arithmetic processing, work memory Thus, it is possible to realize an image processing method that can drastically reduce the use area in this way and thus improve the processing efficiency.

  Furthermore, according to the present invention, for an image processing apparatus that processes image data obtained by imaging an authentication target by a program, the image data is converted into data representing the luminance and position of pixels that are non-zero elements. A first step of converting, a second step of temporarily storing data in the work memory, and a third step of executing arithmetic processing corresponding to the data for the data held in the work memory; As a result of executing the above, it is possible to reduce the data other than non-zero elements in the image data and to perform the arithmetic processing corresponding to the data obtained after the reduction, thereby speeding up the arithmetic processing. In addition to being able to do this, it is possible to dramatically reduce the work memory usage area, thus realizing a program that can improve processing efficiency. It is possible.

  Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

(1) Overall Configuration of Authentication Device In FIG. 1, reference numeral 1 denotes an authentication device 1 according to the present embodiment as a whole, and the blood vessel imaging unit 2 and the signal processing unit 3 are connected to each other via a cable. Composed.

  The blood vessel imaging unit 2 has a curved guide groove 11 formed so as to imitate a finger FG at a predetermined position of the housing 1A of the authentication device 1, and an imaging opening 12 is formed on the bottom surface of the guide groove 11. Is arranged.

  As a result, the blood vessel imaging unit 2 guides the finger pad of the finger FG to be contacted so as to be applied to the guide groove 11 onto the imaging opening 12, and is brought into contact with the tip of the guide groove 11 in contact with the fingertip. The position of the imaging opening 12 with respect to the finger FG is determined according to the photographer.

  A colorless and transparent opening cover portion 13 made of a predetermined material is provided on the surface of the imaging opening portion 12, while a camera portion 14 is disposed immediately below the imaging opening portion 12 inside the housing 1A. ing.

  Further, on the side surface of the guide groove 11, a pair of near-infrared light sources 15 (15 </ b> A and 15 </ b> B) that irradiate near-infrared light specifically absorbed with respect to hemoglobin as imaging light for blood vessels are provided on the guide groove 11. Is arranged so as to sandwich the imaging opening 12 in parallel with the short direction of the finger, and can be irradiated with near-infrared light on the finger pad side portion of the finger FG in contact with the guide groove 11. .

  Therefore, in this blood vessel imaging unit 2, the ratio of near-infrared light reflected on the surface of the finger FG is remarkably suppressed as compared with the case where near-infrared light is irradiated on the finger pad base of the finger FG. Near-infrared light incident on the inside of the finger FG via the finger FG is absorbed by hemoglobin passing through the blood vessel and scattered in tissues other than the blood vessel through the inner side of the finger FG and from the finger FG. As near-infrared light for projecting blood vessels (hereinafter referred to as blood vessel projection light), the light enters the camera unit 14 via the imaging opening 12 and the opening cover 13 in sequence.

  In the camera unit 14, the macro lens 16 and near-red light that transmits only near-infrared light in a wavelength range (approximately 900 [nm] to 1000 [nm]) that depends on both oxygenation and deoxygenation hemoglobin. An external light transmission filter 17 and a CCD image sensor 18 are sequentially arranged, and blood vessel projection light incident from the opening cover 13 is sequentially passed through the macro lens 16 and the near-infrared light transmission filter 17 to the CCD image sensor 18. To the imaging surface. Thus, the camera unit 14 can faithfully image both venous and arterial capillaries mixed inside the adjacent finger FG.

  Under the control of the signal processing unit 3, the CCD image pickup device 18 picks up an image of a blood vessel formed on the image pickup surface, and the image pickup result is an image signal (hereinafter referred to as a blood vessel image signal) SA1, SA2, ... Are sequentially output to the signal processing unit 3 as SAn.

  On the other hand, the signal processing unit 3, as shown in FIG. 2, communicates with the control unit 20 the blood vessel imaging drive unit 21, the image processing unit 22, the authentication unit 23, the flash memory FM, and an interface (data transmission / reception). This is called an external interface (hereinafter referred to as an external interface).

  The control unit 20 includes a central processing unit (CPU) that controls the authentication apparatus 1 as a whole, a read only memory (ROM) that stores various programs, a random access memory (RAM) as a work memory of the CPU, and the like. A mode for registering a blood vessel of a registrant in the control unit 20 in accordance with an operation of an operation unit (not shown) provided at a predetermined position on the surface of the housing 1A of the authentication device 1 ( Hereinafter, an execution command COM1 of a blood vessel registration mode) or an execution command COM2 of a mode for determining the presence / absence of the registrant (hereinafter referred to as an authentication mode) is given.

  Then, when the blood vessel registration mode execution command COM1 is given from the operation unit (not shown), the control unit 20 changes the operation mode to the blood vessel registration mode based on the corresponding program stored in the ROM. The blood vessel imaging drive unit 21, the image processing unit 22, and the authentication unit 23 are each controlled.

  In this case, the blood vessel imaging drive unit 21 activates the blood vessel imaging unit 2 so as to drive the near-infrared light source 15 and the CCD imaging device 18 of the camera unit 14, respectively. As a result, in the blood vessel imaging unit 2, near infrared light is irradiated from the near infrared light source 15 onto the finger pad side of the imager's finger FG (FIG. 1) that is in contact with the guide groove 11 (FIG. 1) at this time. Then, the blood vessel projection light guided to the imaging surface of the CCD image sensor 18 via the finger FG (FIG. 1) is output from the CCD image sensor 18 as blood vessel image signals SA1, SA2,. The data are sequentially output to 22 A / D (Analog / Digital) converter 22A.

  The A / D converter 22A performs A / D conversion processing on the blood vessel image signals SA1, SA2,..., SAn, and blood vessel image data obtained as a result (hereinafter referred to as blood vessel image data) DA1. , DA2,..., DAn are sent to the filter unit 22B.

  The filter unit 22B performs various filtering processes corresponding to noise component removal and contour enhancement on the blood vessel image data DA1, DA2,..., DAn, and blood vessel image data DB1, DB2,. DBn is sent to the binarization unit 22C.

  The binarization unit 22C performs binarization processing on the blood vessel image data DB1, DB2,..., DBn, and obtains black and white blood vessel images (hereinafter referred to as binary blood vessel images) obtained as a result. (Hereinafter referred to as binary blood vessel image data) DC1, DC2,..., DCn are sent to the thinning unit 22D.

  The thinning unit 22D performs, for example, a morphological process on the binary blood vessel image data DC1, DC2,..., DCn, and the binary blood vessel image based on the binary blood vessel image data DC (DC1 to DCn). Linearize blood vessels.

  Then, the thinning unit 22D selects one binary blood vessel image from a plurality of binary blood vessel images formed of linearized blood vessels (hereinafter referred to as blood vessel lines), and selects the selected 2 The binary blood vessel image data DD corresponding to the value blood vessel image is sent to the authentication unit 23.

  The authentication unit 23 generates binary blood vessel image data DD as registration authentication information RC having a predetermined format, and sends this to the control unit 20.

  When the control unit 20 receives the registration authentication information RC from the authentication unit 23 by controlling the blood vessel imaging drive unit 21, the image processing unit 22, and the authentication unit 23 in this manner, the control unit 20 stores the registration authentication information RC in the flash memory FM. And the blood vessel imaging unit 2 is stopped so as to cancel the control of the blood vessel imaging drive unit 21, the image processing unit 22, and the authentication unit 23.

  In this way, the control unit 20 can execute the blood vessel registration mode.

  On the other hand, when the authentication mode execution command COM2 is given from the operation unit (not shown), the control unit 20 changes the operation mode to the authentication mode based on the corresponding program stored in the ROM. Then, the blood vessel imaging drive unit 21, the image processing unit 22, and the authentication unit 23 are controlled, and the registration authentication information RC registered in the flash memory FM is read out and sent to the authentication unit 23.

  In this case, the blood vessel imaging drive unit 21 activates the blood vessel imaging unit 2 as in the above-described blood vessel registration mode. Further, the image processing unit 22 performs various processes on the blood vessel image signals SA (SA1 to SAn) sequentially output from the blood vessel imaging unit 2 in the same manner as in the blood vessel registration mode, and 2 is obtained as a result. The value blood vessel image data DD is sent to the authentication unit 23.

  The authentication unit 23 displays the blood vessel line formation pattern in the binary blood vessel image based on the binary blood vessel image data DD and the binary blood vessel image based on the registration authentication information RC read from the flash memory FM by the control unit 20. Match.

  Then, the authentication unit 23 determines whether or not the photographer imaged by the blood vessel imaging unit 2 at this time is a registrant according to the matching degree, and the determination result is sent to the control unit 20 as determination data D1. Send it out.

  When the control unit 20 receives the determination data D1 from the authentication unit 23 by controlling the blood vessel imaging drive unit 21, the image processing unit 22, and the authentication unit 23 in this way, the control unit 20 transmits the determination data D1 via the external interface IF. Then, the blood vessel imaging unit 2 is stopped so as to cancel the control of the blood vessel imaging drive unit 21, the image processing unit 22, and the authentication unit 23.

  In this way, the control unit 20 can execute the authentication mode.

  As described above, the authentication device 1 performs biometric authentication for determining the existence of the person (registrant) with the blood vessel of the unique structure interposed inside the living body as an authentication target, thereby obtaining a fingerprint or the like on the surface of the living body. Compared to the case of the target, not only direct theft from a living body but also impersonation of a registrant by a third party can be prevented.

(2) Specific Processing Contents of Authentication Unit 23 Here, in the authentication unit 23, as shown in FIG. 3, a sparse matrix conversion unit 31, a blood vessel, and a cache memory 30 as a work memory are provided. The line processing unit 32 and the matching processing unit 33 are connected to each other.

  In this case, in the blood vessel registration mode, the authentication unit 23 performs sparse matrix conversion processing by the sparse matrix conversion unit 31 on the binary blood vessel image data DD sequentially supplied from the thinning unit 22D (FIG. 2) at this time, This conversion processing result is held by temporarily storing it in the cache memory 30. Then, the authentication unit 23 performs a feature point extraction process and a feature point connection process in the blood vessel line processing unit 32 on the conversion process result held in the cache memory 30, and the blood vessel formation pattern obtained as a result of this process is registered and registered in the authentication information RC. Is registered in the flash memory FM (FIG. 2).

  On the other hand, in the authentication mode, the authentication unit 23 holds the plurality of registration authentication information RC1, RC2,..., RCn registered in the flash memory FM at this time sequentially in the cache memory 30 for each predetermined unit, and the thinning is performed. The sparse matrix conversion unit 31 performs sparse matrix conversion processing on the binary blood vessel image data DD supplied from the unit 22D (FIG. 2). Then, the authentication unit 23 sequentially performs the blood vessel line reconstruction process in the blood vessel line processing unit 32 on the predetermined unit of registration authentication information RCi (i = 2, 3,..., Or n) held in the cache memory 30. Based on this processing result and the sparse matrix conversion processing result, the matching processing unit 33 executes the matching processing.

  Hereinafter, the sparse matrix conversion processing by the sparse matrix conversion unit 31, the feature point extraction processing by the blood vessel line processing unit 32, the feature point connection processing and the image reconstruction processing, and the verification processing by the verification processing unit 33 will be described in detail.

(2-1) Sparse Matrix Conversion Processing In the binary blood vessel image data DD supplied from the thinning unit 22D (FIG. 2), for example, as shown in FIG. ) Of binary blood vessel image IM1. Therefore, when the luminance value of the binary blood vessel image data DD is represented as a two-dimensional matrix, most matrix elements are represented as “0” in the two-dimensional matrix.

  Therefore, the sparse matrix conversion unit 31 performs a conversion process called a sparse matrix (hereinafter referred to as a sparse matrix conversion process) on the binary blood vessel image data DD expressed as a two-dimensional matrix, The binary blood vessel image IM1 based on the binary blood vessel image data DD is converted into data DE representing the blood vessel line IM2 (hereinafter referred to as sparse matrix conversion blood vessel line data) DE.

  In practice, the sparse matrix conversion unit 31 has a luminance value “0” in the binary blood vessel image data DD in the binary blood vessel image IM1 (FIG. 4A) based on the binary blood vessel image data DD. Pixels (Pixels) that are not (black) (in this case, “1”) (hereinafter referred to as non-zero elements) are sequentially detected.

  Then, the sparse matrix conversion unit 31 generates the luminance values and positions (coordinates) of these non-zero elements as sparse matrix conversion blood vessel line data DE, and sends this sparse matrix conversion blood vessel line data DEDE to the cache memory 30. As a result, the binary blood vessel image data DD is held in the cache memory 30 as sparse matrix data (sparse matrix converted blood vessel line data DE) representing only the blood vessel line IM2.

  In this way, the sparse matrix conversion unit 31 converts the binary blood vessel image data DD represented as a two-dimensional matrix as data representing the blood vessel line IM2 (the luminance value and position of the non-zero element), thereby obtaining the binary value. The use area in the cache memory 30 can be drastically reduced by the amount of data other than non-zero elements in the blood vessel image data DD.

  Here, according to the experiment, in the number of non-zero elements in the image data of 57600 (Pixel) blood vessel line which is 240 (Pixel) long and 240 (Pixel) wide, as shown in FIG. The value was 538 (Pixel) (0.93 (%)), the maximum value was 1265 (Pixel) (2.20 (%)), and the average value was 898 (Pixel) (1.56 (%)).

  At this time, when the sparse matrix conversion process is performed on the above-described binary blood vessel image data DD (57600 (Byte)) that holds 1-bit data per pixel (Pixel), the data amount is As shown in Fig. 5 (B), the minimum value is 3654 (Byte) (6.34 (%)), the maximum value is 7289 (Byte) (12.65 (%)), and the average value is 5542 (Byte) (9.47 (% It was confirmed that the amount of data can be reduced to about 1/8 to 1/8 of the case where the binary blood vessel image data DD is the data as it is.

  In addition, when sparse matrix conversion processing is performed on double-precision blood vessel image data (460800 (Byte)) that holds 8-bit data per pixel (Pixel), the data amount is as shown in FIG. ), The minimum value is 7420 (Byte) (1.61 (%)), the maximum value is 16144 (Byte) (3.50 (%)), and the average value is 11736 (Byte) (2.55 (%)). It was confirmed that the amount of data can be reduced to about 1/60 to 1/30, compared to the case of accurate blood vessel image data. Thus, it was confirmed that the amount of data can be reduced by performing the sparse matrix conversion process as the amount of data held per pixel (Pixel) increases.

  Furthermore, regarding the relationship between the ratio of non-zero elements and the data ratio with the blood vessel image data, as shown in FIG. 6A, the case of binary blood vessel image data DD or as shown in FIG. Thus, it was confirmed that even in the case of double-precision blood vessel image data, the linearity was such that the data ratio with the blood vessel image data increased as the proportion of non-zero elements increased.

(2-2) Feature Point Extraction Processing and Feature Point Linkage Processing The blood vessel line processing unit 32, as shown in FIG. 7, has branch points and bending points that are characteristic elements from the blood vessel line IM2 (FIG. 7A). (Hereinafter referred to as feature points) CP is extracted (FIG. 7B), and then, as shown in FIG. 8, a plurality of feature points CP (FIG. 8A) on blood vessel line IM2 are Connection is made so as to approximate the blood vessel line IM2 (FIG. 8B).

  In practice, the blood vessel line processing unit 32 extracts a plurality of feature points CP from the blood vessel line IM2 based on the sparse matrix conversion blood vessel line data DE supplied from the sparse matrix conversion unit 31 and held in the cache memory 30. Then, the blood vessel line processing unit 32 generates these feature points CP as data DF (hereinafter referred to as feature point data) DF represented by coordinate values.

  Further, as shown in FIG. 9, the blood vessel line processing unit 32 selects a feature point (hereinafter referred to as a search reference point) A as a search target from these feature points CP, and selects the selected search reference. As a search candidate for a feature point (hereinafter referred to as a blood vessel line corresponding line segment forming point) that forms a line segment corresponding to the point A and the blood vessel line IM2 (hereinafter referred to as a blood vessel line corresponding line segment), An arbitrary feature point (hereinafter referred to as a search candidate point) B other than the search reference point A is selected, and the search reference point A and the search candidate point B are connected to generate a line segment AB.

  In this state, the blood vessel line processing unit 32 recognizes all pixels (Pixels) on the line segment AB generated between the search reference point A and the search candidate point B, and based on this recognition result, the line segment AB. Are sequentially determined as to whether or not the pixel (Pixel) is at the same coordinate as the pixel (Pixel) of the blood vessel line IM2 based on the sparse matrix conversion blood vessel line data DE.

  Then, the blood vessel line processing unit 32 determines all pixels (Pixels) on the line segment AB generated between the search reference point A and the search candidate point B, and a cross-correlation value larger than a predetermined threshold is determined as the determination result. Is obtained, it is determined that the search candidate point B is a blood vessel line corresponding line segment forming point that forms a blood vessel line corresponding line segment with the search reference point A, and the search candidate point B and the search reference point A selected at this time are determined. Are replaced with the search candidate point B, the remaining feature points CP are selected as new search candidate points B, and the above-described search process is repeated.

  On the other hand, when the cross-correlation value equal to or less than the threshold value is obtained, the blood vessel line processing unit 32 determines whether the search candidate point B forms a blood vessel line corresponding line segment with the search reference point A. After the connection between the search candidate point B selected at this time and the search reference point A is discarded, the remaining feature point CP is selected as a new search candidate point B instead of the search candidate point B. The above search process is repeated.

  When the blood vessel line processing unit 32 finishes searching for all the feature points CP as new search candidate points B in this way, the feature point CP different from the feature point selected as the search reference point A so far is used. Is selected as a new search reference point A, and the same processing as described above is executed.

  In this way, the blood vessel line processing unit 32 performs data (hereinafter referred to as connection state data) DG (referred to as connection state data) DG representing the connection state between the feature points corresponding to the feature point data DF and the connection part (blood vessel line corresponding line segment). 8 (B)) and these are registered in the flash memory FM as registration authentication information RC.

  At this time, in the blood vessel line processing unit 32, since the binary blood vessel image data DD is held in the cache memory 30 as a sparse matrix (as a sparse matrix converted blood vessel line image data DE), for example, MATLAB (Registered by The MathWorks) The above-described feature point extraction process and feature point connection process are executed using a predetermined matrix calculator 34 that speeds up a sparse matrix calculation such as a trademark.

  As a result, the blood vessel line processing unit 32 extracts a feature point CP from the blood vessel line IM2 as in the above-described feature point extraction process, or local access frequently occurs as in the above-described feature point connection process. Even in this case, the feature point extraction process and the feature point connection process for the sparse matrix conversion blood vessel line data DE held in the cache memory 30 can be accelerated.

  Here, according to an experiment, as shown in FIG. 10, as compared with the case where the feature point extraction process and the feature point connection process are performed on the binary blood vessel image data DD composed of a general two-dimensional matrix, about 2 It was confirmed that the processing speed could be increased by about 5 times. Incidentally, the processing time in FIG. 10 is an average value obtained by performing the feature point extraction process and the feature point connection process as described above 10 times.

(2-3) Blood vessel line reconstruction processing The blood vessel line processing unit 32 generates the blood vessel line IM2 (FIG. 8B) based on the registration authentication information RC representing the plurality of feature points CP and the connection state data DG. Reconfigure.

  Actually, the blood vessel line processing unit 32, based on the connection state data DG of the registration authentication information RC held in the cache memory 30 at this time, each feature point CP represented by the feature point data DF of the registration authentication information RC. Recognize the connected state.

  In this state, the blood vessel line processing unit 32 performs, based on the recognition result, a pixel (Pixel) on a line segment between the feature points CP to be connected among the feature points CP by a line drawing process based on a predetermined algorithm. To connect. As a result, the blood vessel line corresponding to the original blood vessel line IM2 (FIG. 8B) is reconstructed.

  Then, the blood vessel line processing unit 32 uses the sparse matrix data (hereinafter referred to as reconstructed blood vessel line data) DH of the reconstructed blood vessel line (hereinafter referred to as reconstructed blood vessel line) IM3. Generate.

  In this way, the vascular line processing unit 32 can reconstruct the vascular line IM2 (FIG. 8B) based on the registration authentication information RC representing the plurality of feature points CP and the connection state data DG. It is made like that.

(2-4) Collation Processing As shown in FIG. 11, the collation processing unit 33 reconstructs the vascular line IM3 (FIG. 11A) and the vascular line IM2 obtained from the collator's imaging result (FIG. 11B). ))), The presence / absence of the registrant is determined.

  In practice, the verification processing unit 33 includes, for example, the reconstructed blood vessel line IM3 based on the reconstructed blood vessel line data DH1 among the reconstructed blood vessel line data DHi based on the predetermined unit of registration authentication information RCi held in the cache memory 30, and the sparse The cross-correlation with the blood vessel line IM2 based on the sparse matrix conversion blood vessel line data DE held in the cache memory 30 from the matrix conversion unit 31 is calculated, and the blood vessel lines are collated.

  When the cross-correlation value equal to or smaller than the predetermined threshold is obtained as the collation result, the collation processing unit 33 reconstructs the vascular line based on the remaining reconstructed vascular line data DH2 to DHi held in the cache memory 30 at this time. IM3 is sequentially compared with the vascular line IM2 in the same manner as described above until the reconstructed vascular line IM3 corresponding to the vascular line IM2 is detected.

  In this way, the verification processing unit 33 stores all the registration authentication information RC (RC1 to RCn) registered in the flash memory FM for each reconstructed blood vessel line data DHi based on the registration authentication information RCi in a predetermined unit. When the cross-correlation values are sequentially compared as one side and all of the cross-correlation values are less than or equal to the predetermined threshold value, it is determined that the imager who is imaged by the blood vessel imaging unit 2 is not a registrant. When the value is obtained, it is determined that the photographer is the registrant, and the determination result is transmitted as determination data D1 to the outside via the external interface IF under the control of the control unit 20. Yes.

  At this time, the matching processing unit 33 executes the above-described matching processing using the matrix calculator 34 in the same manner as the blood vessel line processing unit 32, so that the local access frequently occurs. The collation process can be speeded up.

(3) Operation and Effect In the configuration described above, the authentication device 1 uses the binary image data represented as a two-dimensional matrix as a luminance value of a non-zero element (pixel) in a binary image based on the binary image data. And sparse matrix data representing the position. Then, the authentication device 1 temporarily holds the sparse matrix data in the cache memory 30 and performs a corresponding arithmetic process on the sparse matrix data.

  Therefore, the authentication apparatus 1 reduces the data other than the non-zero elements in the binary image data and performs the arithmetic processing corresponding to the data obtained after the reduction. As apparent from 10), in addition to being able to speed up the arithmetic processing, the area used in the cache memory 30 can be drastically reduced. As a result, the processing efficiency of the arithmetic processing is improved. be able to.

  In this case, the authentication device 1 performs the above-described feature point extraction processing, feature point connection processing, and matching processing on the sparse matrix data held in the cache memory 30 using the matrix calculator 34.

  Therefore, in this authentication device 1, since the calculation processing can be speeded up, the processing efficiency during the calculation processing is remarkably increased by reducing the processing load for the feature point extraction processing, the feature point connection processing, and the matching processing. Can be improved.

  Also, in this case, the authentication device 1 stores a plurality of data composed of sparse matrices in the flash memory FM, and temporarily stores the stored data of each sparse matrix in the cache memory 30 for each predetermined unit during the arithmetic processing. Hold.

  Accordingly, in this authentication device 1, the use area in the cache memory 30 can be drastically reduced, so that the number of units of sparse matrix data held in the cache memory 30 at this time is increased as compared with binary image data. As a result, the processing efficiency during the arithmetic processing can be remarkably improved by the amount that the number of data transfers between the flash memory FM and the cache memory 30 is reduced. In this case, this is particularly effective in a situation where the storage area of the cache memory 30 is limited.

  According to the above configuration, the binary image data represented as a two-dimensional matrix is converted as sparse matrix data representing the luminance values and positions of non-zero elements (pixels) in the binary image based on the binary image data. Then, the data is temporarily stored in the cache memory 30 and the corresponding arithmetic processing is performed on the data of the sparse matrix, thereby reducing data other than non-zero elements in the binary image data and the reduction. Since arithmetic processing corresponding to data obtained later is performed, in addition to speeding up the arithmetic processing, the area used in the cache memory 30 can be drastically reduced. Processing efficiency can be improved.

(4) Other Embodiments In the above-described embodiment, the case where the blood vessel inside the finger FG of the living body is applied as an authentication target has been described. In addition, blood vessels in various other biological parts such as blood vessels of the arms or eyes may be applied, and various other authentication targets such as fingerprints on the surface of the living body and nerves in the living body may be applied. May be. Incidentally, when a blood vessel is applied, the amount of image data can be further reduced because there are fewer non-zero elements and the amount of image data based on the blood vessel is large compared to an authentication target such as a fingerprint. As a result, the processing efficiency can be improved. When a nerve is an authentication target, for example, if a marker specific to the nerve is injected and the marker is imaged, the nerve may be the authentication target in the same manner as in the above-described embodiment. it can. Furthermore, the present invention is not limited to this. For example, general image data obtained by imaging a background or a person as a subject can be set as an authentication target in the same manner as in the above-described embodiment.

  In the above-described embodiment, as the conversion means for converting the image data into data representing the brightness and position of the pixel that is a non-zero element, the non-zero element (“1”) from the binary blood vessel image data DD and the data Although the case where the coordinate position of the pixel (Pixel) determined to be a non-zero element (“1”) is extracted has been described, the present invention is not limited to this, and various images such as gray scale and multi-value image data, for example. Can be applied to data. For example, in the case of gray scale, the non-zero element may be an element having a numerical value equal to or larger than a predetermined threshold, and the numerical value of the non-zero element (for example, “1 to 256”) can be held as a matrix element.

  Further, in the above-described embodiment, the case where the sparse matrix conversion is performed as the conversion unit that converts the image data into the data representing the luminance and position of the pixel that is a non-zero element has been described, but the present invention is not limited to this. In addition, the non-zero element and the coordinate position data of the non-zero element may be converted based on various other conversion methods.

  Further, in the above-described embodiment, the case where the cache memory 30 is applied as the work memory for temporarily storing data has been described. However, the present invention is not limited to this, for example, other than the cache memory. Various work memories can be widely applied.

  Further, in the above-described embodiment, feature point extraction processing, feature point connection processing, and registration authentication information RC registration are performed as arithmetic means for executing arithmetic processing corresponding to the data held in the work memory. Although the case where the process is performed has been described, the present invention is not limited to this, and by performing various other processes other than the feature point extraction process and the feature point connection process, the non-zero element to be authenticated and the non-zero element Registration processing to the flash memory may be performed on the data based on the coordinate position data of the element, and the non-zero element to be authenticated and the non-zero element without performing the feature point extraction process and the feature point connection process The coordinate position data may be registered.

  Further, in the above-described embodiment, a predetermined point extraction process is performed on the sparse matrix conversion vascular line data DE, and a plurality of feature points are respectively obtained from the vascular line based on the sparse matrix conversion vascular line data DE. The case of extraction has been described. For example, point extraction processing called a Harris corner, a lattice-shaped image having a predetermined size is superimposed on a binary blood vessel image, and a point on a blood vessel line intersecting with the lattice is extracted. An arbitrary point on the blood vessel line in the value blood vessel image may be extracted. Even in this case, the same effects as those of the above-described embodiment can be obtained.

  Furthermore, in the above-described embodiment, all the pixels (Pixels) on the line segment AB generated between the search reference point A and the search candidate point B are determined, and the determination result is a cross-correlation larger than a predetermined threshold value. When the value is obtained, the case where the search candidate point B selected at this time and the search reference point A are connected has been described. However, the present invention is not limited to this, and each feature point is determined by a method such as a relative neighborhood graph. Then, these line segments may be determined as described above, and various other feature point connection processing methods may be used.

  Furthermore, in the above-described embodiment, the case where the flash memory FM is applied as the storage means for storing each data converted by the conversion means has been described. However, the present invention is not limited to this, for example, the HDD ( Various other storage means such as Hard Disk Drive) and Memory Stick (registered trademark of Sony Corporation) can be widely applied. Further, this storage means may be configured to be connected via a transmission line as a separate body from the authentication device 1.

  Further, in the above-described embodiment, the A / D conversion unit 22A (FIG. 3), the filter unit 22B (FIG. 3), the binarization unit 22C (FIG. 3), and the thinning unit 22D are applied to the blood vessel image signal SA. Although the case where feature points are extracted after performing various preprocessings sequentially is described, the present invention is not limited to this, and the processing content of the preprocessing may be changed as necessary. .

  Furthermore, in the above-described embodiment, a binary blood vessel line (dark hatched portion) as shown in FIG. 9 and an image (thin hatching) obtained by dilating the binary blood vessel line with a morphological processing dilation (Dilation) The portion) is described as binary blood vessel image data DD. However, the present invention is not limited to this. For example, a binary blood vessel line may be used as binary blood vessel image data DD. It may be binary blood vessel image data subjected to.

  Further, in the above-described embodiment, the case where the feature point data DF and the connection state data DG (FIG. 8B) are registered in the flash memory FM as the registration authentication information RC has been described. Not limited to this, the sparse matrix converted blood vessel line data DE may be registered in the flash memory FM.

  The present invention can be used in a field using a technique for identifying a living body.

It is a basic diagram which shows the whole structure of the authentication apparatus by this Embodiment. It is a block diagram which shows the structure of a signal processing part. It is a functional block diagram with which it uses for description of the process of an authentication part. It is a basic diagram which shows before and after sparse matrix conversion. It is a table showing reduction of the data amount based on a sparse matrix conversion process. It is a graph which shows the relationship of the data ratio with the ratio of a non-zero element, and the image data of a blood vessel line. It is an approximate line figure showing before and after extraction of a feature point. It is an approximate line figure showing before and after connection of a feature point. It is an approximate line figure used for explanation of connection of a feature point. It is a table showing reduction of the processing time by sparse matrix conversion. It is a basic diagram with which it uses for description of the collation process with a reconfigure | reconstructed blood vessel line and the imaged blood vessel line. It is a conceptual diagram with which it uses for description of the mode in a cache memory. It is a conceptual diagram with which it uses for description of the state in the conventional cache memory.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Authentication apparatus, 2 ... Blood vessel imaging part, 3 ... Signal processing part, 14 ... Camera part, 15A, 15B ... Near infrared light source, 16 ... Macro lens, 17 ... Near infrared light Transmission filter, 18 ... CCD imaging device, 20 ... control unit, 21 ... blood vessel imaging drive unit, 22A ... A / D conversion unit, 22B ... filter unit, 22C ... binarization unit, 22D ... Thinning section, 23 …… Authentication section, 30 …… Cache memory, 31 …… Sparse matrix conversion section, 32 …… Vessel line processing section, 33 …… Verification processing section, 34 …… Matrix calculator, IF …… Interface , FM: Flash memory, FG: Finger, COM (COM1 and COM2) ... Execution command, SA (SA1-SAn) ... Blood vessel image signal, DA (DA1-DAn), DB (DB1-DBn) ... Blood vessel image data, DC (DC1-D n), DD: binary blood vessel image data, DE: sparse matrix transformed blood vessel line data, DF: feature point data, DG: connection state data, DH (DH1 to DHn): reconstructed blood vessel line data, IM1... Binary blood vessel image, IM2... Blood vessel line, IM3... Reconstructed blood vessel line, RC (RC1 to RCn)... Registration authentication information, D1.

Claims (8)

In an image processing apparatus that processes image data obtained by imaging an authentication target,
Conversion means for converting the image data into data representing the brightness and position of a pixel that is a non-zero element;
A work memory that temporarily holds the data;
An image processing apparatus comprising: an arithmetic unit that executes arithmetic processing corresponding to the data with respect to the data held in the work memory. The conversion means is
The image processing apparatus according to claim 1, wherein the image data is converted into matrix data by performing sparse matrix conversion. The above image data is
The image processing apparatus according to claim 1, wherein the image processing apparatus is image data based on a blood vessel unique to a living body. The conversion means is
Converting a plurality of the image data into each of the data,
Storage means for storing each of the data converted by the conversion means,
The computing means is
Each of the data stored in the storage means is supplied to the work memory for each predetermined unit, and an arithmetic process corresponding to the data is performed on the data of the predetermined unit held in the work memory. The image processing apparatus according to claim 1. In an image processing method for processing image data obtained by imaging an authentication target,
A first step of converting the image data into data representing luminance and position of a pixel that is a non-zero element;
A second step of temporarily storing the data in the work memory;
An image processing method comprising: a third step of executing a calculation process corresponding to the data with respect to the data held in the work memory. In the first step,
The image processing method according to claim 5, wherein the image data is converted into matrix data by performing sparse matrix conversion. In the first step,
Converting a plurality of the image data into each of the data,
In the second step,
Each of the data is temporarily stored in the work memory for each predetermined unit,
In the third step,
The image processing method according to claim 5, wherein an arithmetic process corresponding to the data is executed on the data of the predetermined unit held in the work memory. For an image processing apparatus that processes image data obtained by imaging an authentication target,
A first step of converting the image data into data representing luminance and position of a pixel that is a non-zero element;
A second step of temporarily storing the data in the work memory;
A program for causing the data stored in the work memory to execute a third step of executing arithmetic processing corresponding to the data.