JPH02501098A - Storing and retrieving identification information - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] Memory and search for specific information le yue a1 east The present invention digitally stores information identifying an individual, e.g., an image of the individual's face and signature. It relates to storing and retrieving information.

Such information may be used, for example, when a buyer provides a credit card to a seller when purchasing something. Can be used when passing a code. Previously, sellers had to certify the buyer's signature on the card. or compare the buyer's face to the photo on their driver's license. They also try to verify the identity of the purchaser.

One digital technique for verifying a person's identity was introduced on January 13, 1987. No. 4,636,622 issued to It involves digitally storing the person's fingerprints in a central location. Ru. If it becomes necessary to perform identification, use the fingerprints of the person Take it again with a machine that can convert it into digital data. This data is central The fingerprint is sent to the station, where it is compared with the stored fingerprint. If those fingerprints If there is a match, a signal will be sent to that remote station.

Mitsushige F and about 1 A general feature of the invention is a method for compressing facial images into a digital bit stream. This method converts the image into an array of raw pixel values. and downscaling this image to generate a smaller array of pixel values. and reducing the grayscale range of pixel values within the small array. obtaining a data word representing the pixel value; a type of code in which the data word is encoded as a shorter set of bits. therefore encoding the data words of said bit stream. I'm reading.

A preferred embodiment has the following features.

The method is suitable for verifying the identity of a given group of individuals at a large number of locations. the digital bit stream, and the method then further capturing information about facial images that identify each individual for compression into a facial image; This compressed bit stream is placed on a pocket-sized recording medium to be given to a person. recording the recorded bit stream and verifying the identity of the individual; providing a device at each location for restoring images from the system; There is.

The recording medium can be a magnetic strip or a printed strip, either is also part of a plastic card. Bit stream central information In the case of memory, each individual has his or her own compressed bit stream at a central location. It is possible to provide a key to the storage address recorded in the and this bit stream is routed to its local device based on its key information. sent to the This key is stored on a pocket-sized recording medium given to the individual. recorded as

This bit stream is no longer than 5000 bits.

The above image is reduced to a small array and specified by the user based on the observation of the image. up to a certain extent. Before encoding, each digital word is The zero method is also the difference between two of the vixel values of the ray scale. remapping the pixel values to a coarse range of pixel values. this Remapping focuses more on tones in the middle of a scale than on tones at either end of the scale. More sensitive.

As a result, individuals can be quickly and accurately identified based on their facial and signature images. It can be done. Images of faces and signatures can be placed on cards that individuals can easily carry but are difficult to forge. be remembered. Instead, access to images of faces and signatures by individuals is limited to all images. Can be limited by central storage. Required to memorize images of faces and signatures The total amount of bits considered is small, and the time required to transmit these bits is The amount and amount of space required to store these bits is reduced.

Other advantages and features will be apparent from the following description of the preferred embodiments and from the claims. There will be.

beautiful snow We first briefly describe the drawing.

FIG. 1 is a block diagram of an identification system.

FIG. 2 is an acquisition terminal or unit.

FIG. 3 is a diagram of a captured frame illustrating cropping.

FIG. 4 is a flowchart of a routine for processing facial images.

FIG. 5 is a diagram of scale processing.

FIG. 6 is a flowchart of a routine for processing signatures.

FIG. 7 shows the creation of an identification baguette.

FIG. 8 is a block diagram of a display terminal or unit.

Figures 9 and 10 show the respective flows of the routine for displaying facial images and signatures. It is a chart.

FIG. 11 is a diagram of bilinear interpolation processing.

FIG. 12 is a diagram of a credit card.

1- Sac Referring to Figure 1, identifying information (i.e., an individual's facial image and signature) is digitally recorded. In the system 10 for uploading and searching, images and signatures are so-called captured end red stamps. It is captured by the terminal numbered 12. Identify the individual whose image and signature were captured Therefore, the display terminal or terminal 14 is said to correspond to the individual to be identified. Display the previously memorized image and signature.

Capture terminal 12 is connected to central station 18 by dedicated telephone line 16 .

The concave line 16 conveys the captured image and digital data representing the signature, and the central stage is maintained in the normal storage device 20 of the application 18. The central station 18 is A conventional communication device 22 for receiving and transmitting digital data and monitoring the communication device. and, in particular, a conventional data processing device 24 for processing the stored data.

The display terminal 14 displays stored images and signatures used to identify individuals. To receive the data, the central station 18 is connected to a regular dialing line. Talkative! ! 26.

A combined capture and display terminal 28, which may perform both capture and display functions, includes a It is connected to the central station 18 via a dialing line 30. different viewing edge The terminals 32 are connected to the central station 18 via a satellite link 34.

A freestanding capture and display unit 36,38 may also be provided. The capture unit 36 is not connected to the central station 18, but instead e.g. The image and signature information are stored on the magnetic strip of the card. Similarly, the display unit 38 Transfer signatures and images to credit cards without being connected to central station 18 View directly from the tribe.

A wide variety of combinations of terminals and units (beyond the exemplary combination shown in Figure 1) You can set a limit. The combination can be made through any combination of available communication links. A large number of acquisition terminals, display terminals and acquisition units with communication to a central station. and display unit.

Such systems have a wide range of application. One example is credit card This is a system for identifying users. In such systems, each authorized The image and signature of the credit card holder may be stored on the credit card itself or centrally. 0 individuals to use credit cards held on application storage devices With each presentation, the shopkeeper projects an image and signature that is said to belong to the individual, and Confirm that it is a person. The display shows credit card stripes. data that is sent directly from a central station or from a central station. I can do it. Other information stored on your credit card may also be used for this purpose. For example, the identification number stored on the card may be returned to the display terminal. used at the central station to indicate stored images and signatures to be sent. can be done.

Referring to FIG. 2, at the capture terminal 12 or capture unit 36, the normal Black video camera 40 (for example, Model No. manufactured by...) is suitable. arranged to capture a facial image of the individual 42 using suitable lighting (not shown). . The second monochrome video camera 44 (model...) has the signature of the individual 42. 46 images. Signature 46 is written on a piece of paper and then can be held in place by a carrier (not shown).

The output of each camera 40.44 is a conventional raster scan analog composite This is an NTSC video signal. The NTSC video signal is a normal frame graph. is sent to the bar 48. The operation of the camera 40.44 and frame grabber 48 is carried out by microphones. The microprocessor controls and operates an interactive terminal 52 connected to the microprocessor. (according to the user's instructions), facial images and signature images are Sent to mule 48. Interactive terminal 52 includes a personal computer monitor and A menu-driven keyboard can be used.

Frame grabber 48 samples the NTSC signal, digitizes it, and digitizes each For the image, there are 307,200 entries in data storage 54 with 480 rows and 640 columns. Send a series of raw pixel values. Each raw pixel value is 8 bits long and its value is generated by the frame grabber, and the set of image pixel values consists of all 256 All possible 8-bit pixel values are distributed appropriately.

The array of raw pixel values representing both the face image and the signature are processed and compared. With a relatively small number of bits, it can be stored on the magnetic strip of a regular credit card. images and signatures in a sufficiently small number of bits to be sent over the wire in a very short time. generate a compressed bit stream that captures the Processing of raw pixel values is input from the user under the control of a program stored in a regular program storage device. This is done by the microprocessor 50 based on the power.

Referring to FIG. 3, the raw pixel values generated by camera 40 are Contains information about both the face and the background surrounding the face, and therefore contains a fraction of the raw pixel values. Frame 60 is a cropped frame large enough to contain all of the important information. larger than the frame 62.

Referring to Figure 4, the first step in processing a facial image is to Cropping the frame to include only those pixel values in the frame ( 70), which has the widest height and width, is cut out by the user. 0, which is indicated via the interactive terminal 52 in the usual manner, e.g. The borders of the removed frame 62 are moved using the normal cursor until the appropriate border is reached. regulated using the control system. Therefore, the user must View ratio (ratio of height to width) and vixes present in the cut frame You can command both the total number of files (defined by the absolute value of their height and width). Ru. The main effect of cropping is that the face image is separated from irrelevant column pixels. This eliminates the need for subsequent processing of the irrelevant vixels.

The next step is to scale the image (72), each by a scale factor. The goal is to reduce the number of pixels in Dean John. The user - select the scale factor via the active terminal 52, and if the same scale factor is dimension. This scaling is done by a simple averaging process. . Referring to FIG. 5, for example, a cropped frame for a scale factor of 4 Each region 80 including 4×4=16 pixels 81 in 82 is the region 80 A single pixel in image 86 is scaled by averaging the 16 pixel values in One pixel is mapped to 84 pixels.

Referring again to Figure 4, the pixel values of the scaled image are then reduced to 8 bits. The data is then remapped to 4 bits (74).

Remapping is done by changing the tone (shade) of an 8-bit pixel value to a 4-bit pixel value. This is done by assigning to the tone as follows. Remapping is the focus biased, that is, tones at intermediate scales than at the extremes of the scale be sensitive to.

The remapping table appears as follows.

8-bit pixel, remapped 4-bit value ffl, "Uth" 128-139 B 14O-1519 The cropped, scaled, and remapped pixel values are then of the form Data compression is performed based on the Shannon-Fari code of Eq. (76), ( An example of Shannon-Fano encoding is “D1g1talI” by Pratt. sage Processing″ W i l e y + New Yor k, 1978. P, 741. ) First, each The difference between the pixel value and the previous pixel value is calculated. (Here, the pixel value is They are obtained one by one, column by column. ) The output of the encoding process is a pixel A bitstream representing a codeword in the code based on the calculated difference value of the values It is a system. If this difference is zero (which is the most favorable situation) ), the codeword is a single "1" bit. If the difference is +1, -1 or ÷2 If so, the codeword will be one of three different 3-bit codewords ( 001.010.011), and if the difference is any other value, the code The word is a special 3-bit code that follows a 4-bit value equal to the magnitude of the actual difference. word (000).

The result of the encoding process is a relatively short stream of bits, e.g. less than 5000 bits. It is mu. The actual number of bits depends on cropping, scaling factors, and remapping. Image quality and final image quality depend on each stage of processing, including tucking techniques and encoding. There is a trade-off between the number of bits in the ream 0. For example, the number of bits in the ream is 5 The quality of the image can be reduced by scaling the image to make it sharper. descend. In most cases, it is recorded on the magnetic stripe of a typical credit card. It is possible to obtain recognition-quality facial images with a final bit stream short enough to It would be Noh.

Referring to Figure 6, in processing a signature image, the image is first compared to a face image. is cropped (90) in the same way as.

The next step (unlike processing facial images) is to equivalence the cropped image ( 92), that is, mapping each 8-bit pixel value to 1-bit black or be converted to a white value. The threshold for choosing between black or white is It is selectable by an operator using conversational terminal 52 (FIG. 1).

The equivalenced image is then processed by the microprocessor 50 to calculate the total number of pixel values. Automatically determine scaling factor to compress to target number It is scaled (94) in the same way as the face image except that it is scaled (94).

The scaled pixel values are subjected to two stages of data compression. 1st stage The technique called run-length unpredictable differential quantization (“Image analysis” by Huang et al. "Bandwidth Compression" Gordon J, Beach.

Two-dimensional run-length encoding (described in N.Y., 1972) (96), the objective is to recognize a two-dimensional region that is black. First one The line is run-length coded in a conventional manner.The line's run-length code is The difference between and the run length sign of the next line is determined. The line obtained by this process The pair difference value is next, and the entry followed by the 000 code is the position difference value or run length. using the same sign as for the face image, except that the difference value is either The length for each entry is the length of the signature. calculated during compression, so the lengths are unique to a given signature and each length is It changes from one signature to the next. The result is an encoded It is a bit stream. For example, the bit stream is about 1400 bits. However, the actual number is short for cropping, equivalence, scaling and encoding. It depends and changes.

The same image quality and bit stream length trade-off (as for facial images). There is a f.

Referring to Figure 7, the signature and image bit stream is generated after being compressed. , an identification packet 90 is assembled. This packet has a header (packet identification packet, caller identifier, description of packet content and size, and identification of the encoding technique used. signature bit stream, facial image bit stream, stream, a compressed version of literary information entered by the user (e.g. Credit limits for credit card holders and packet consistency later Contains checksums for all packets used for inspection.

In the case of an independent supplementary device, referring again to Figures 1 and 2, the identification packet The cut is from data memory 54 (which was stored by microprocessor 50). from the packet to a magnetic stripe recording device, which inserts the packet into a plastic cartridge. is recorded on the magnetic stripe 57 of the card 59.

On the other hand, in the case of an acquisition terminal connected to a central station, the identification packet From the data storage device 54, a normal It is sent to the communication unit 61. At the central station, the packet is included in the ladder stored in the storage device 20 at a later retrievable location based on the identifier will be paid. That same identifier can also be clicked in some way so that it can be retrieved later. Can be recorded as bits on a credit card.

Referring to FIG. 8, in the display terminal or unit 14.38, the Display packet (for standalone display unit 38) ) from the magnetic stripe 59 via the magnetic stripe reader 102, or central office via communication unit 104 (in case of display terminal 14); , and is loaded into data storage unit 100 . In each case, the packet (based on the program stored in the program storage device 108). ) by the microprocessor 106 and via the keyboard 110. controlled under the direction of the user. Capture terminal or unit used Microprocessor 106 uses the reverse procedure of Regenerate the image and cropped signature pixel array, then add the regular frame Display-pixels combined for loading into system buffer 112 ・Generate an array. A typical video monitor 114 (including drivers) The facial image of the individual whose identity is being checked and Give signature indication. The user, on the other hand, uses the macro to view the textual information from the packet. Can command the microprocessor.

Referring to FIG. 9, when reproducing a facial image, the microprocessor 106 For the first time, a facial image bit stream was created by Shannon-Fa. no) decoding 120), the 0th order to recover an array of 4-bit pixel values; , this 4-bit pixel value is mapped to an 8-bit pixel value according to the table below. 4-bit pixel map method for the xel value, 8-bit pixel This image is To open the existing pixel values using bilinear interpolation (122), It is expanded by adding more pixel values.

Referring to FIG. 11, in bilinear interpolation, the enlarged Each new vixel 130 in cell array 132 is The four adjacent vixel values in array 142 are 134.136.138.140. Calculated as a weighted average, this weight is placed in the expanded array 132. Based on the distance of the new pixel 130 from each of the original pixels. Ru.

Bilinear interpolation was introduced in 1979 by Prentice Hall. tice-Hall), Castleman's L1 Ward 9) I - Discussed in WMC $9υ1.

Following the expansion, Bixel, for example, p. 322 of the Pratt reference. control using convolutional filtering techniques of the type described in The last emphasis is given.

Referring to Figure 10, the signature image is regenerated from the compressed signature bitstream. To do this, microprocessor 106 first sends the pit stream to S hannon-F ano decoding and run-length decoding (142), regenerating the thresholded and scaled pixel array do. This array is similarly expanded using bilinear interpolation (144).

Referring to FIG. 11, an identification packet is recorded on a conventional credit card 150. In one technique, the magnetic stripe 152 is 4 in. Treated as having a book track. 3 out of 4 tracks are ANSI Matches NS, recorded at track 1.2 (210 bits per inch (bpi)) ) and track 3 are used for conventional purposes. Track O is two and a half The track is divided into 154 and 156 tracks, and a zero value bit is recorded on one half track. , 2600 bits per input due to magnetic flux changes that record single-value bits on other half tracks. Alternatively, all four tracks may be recorded with be written outside the ANSI standard at a higher bit rate to include the identification packet. You may also record.

to 1 In an example of a system used to identify credit card holders, Each person to whom a card is issued must use a terminal to capture an image of the “image” and signature. It is necessary to capture and accumulate the information. The compressed identification packet is issued deposited directly on your credit card, at a central office, or both. The key to search the accumulated packets is also stored on the credit card (probably (on a separate track).

When an individual presents a card to buy something, the seller moves the stripe to Pass through the terminal's Strive Leader. The person's image and signature are credit-limited. Displayed on the seller side along with information such as degree. Identification packet is credit card based on the storage location read directly from the card or read from the card and sent to the central office. The information is then retrieved from the central office via telephone lines.

Other embodiments are within the scope of the following claims.

For example, this system can be used for various applications other than the credit card described. can be done. It also captures and searches other forms of identification information, including fingerprint information. You can also do this.

The trimming process involves installing a zoom lens on the video camera so that the facial image fills the entire screen. This can be done by having the user adjust the zoom lens to fill the area. Wear. This way, the original raw pixel values will not contain any unwanted background pixels. do not have.

Identification information may be collected by means other than the magnetic stripe on a credit card, e.g. holographically on a card or in an optically readable stripe. I can do it.

Other data compression techniques are used to ensure that identification packets and The specific technology can be displayed on Heragu.

Identification The information identifying the packet may be stored independently of the packet itself, e.g. on a card. It is stored somewhere, and the packet and the information that identifies it act as a lock and key. The card may be prevented from being used in a manner that does not involve detection of both the key and the key. stomach.

Other techniques for processing and redisplaying signature images can be used. Lanre Following the step of pixel decoding, the threshold scaled pixels are Either scale the array up (without interpolation) or scale it up and flatten the edges. Smoothing may also be performed.

Engraving (no changes to the content) IG 1 FIG 7 FIG12 Procedural amendment □ January to Sunday, 1990, the country Yoshi, Commissioner of the Patent Office 1) Takeshi Moon 1.Display of the incident PCT/US88102503 2. Name of the invention 3. Person who makes corrections Relationship to the incident: Patent applicant address Name: ETC (2 others) 4. Agent Address: Shin-Otemachi Building, 206-ku, 2-2-1 Otemachi, Chiyoda-ku, Tokyo 5. Date of amendment order: November 7, 1999 ■Date of sending) (2) Power of attorney and translation Letter (3 letters each) international search report

Claims (14)

[Claims] 1. A method for compressing facial images into a digital bit stream, the method comprising: Converts an image to an array of row pixel values and emits a smaller array of pixel values. steps to reduce the image to create a gray scale range of pixel values. decreasing in the smaller array; deriving a data word indicating the pixel value; and More frequently occurring data words are encoded as shorter sets of bits the data word into the bit stream according to a code in the format and encoding the method. 2. Claim 1, wherein the bit stream is not longer than 5000 bits. How to put it on. 3. The image is further reduced in size to the extent that the image is identified by the user based on his/her perception of the image. The method of claim 1, wherein the method is reduced to a small array. 4. Each of said data words has a reduced gray scale pixel value of 2 2. The method of claim 1, comprising a difference between two. 5. further remapping said pixel values to a coarser range of pixel values. 2. The method of claim 1, comprising: 6. The remapping operation is more tonally sensitive in the middle of the scale than at the ends of the scale. 6. The method according to claim 5, wherein the method is sensual. 7. Used to verify the identity of a given group of individuals in multiple locations. for further compression into the digital bit stream. a step of capturing information about facial images for identifying each individual; said digital bit stream on a pocket-sized recording medium that can be and recording the recorded bit to confirm the identity of the individual. a device at each of said locations for playing back said images from said stream; 2. The method of claim 1, comprising the step of providing. 8. 8. The method of claim 7, wherein the recording medium comprises a magnetic strip. 9. 8. The method of claim 7, wherein the recording medium comprises a printed strip. 10. 8. The strip according to claim 7, wherein the strip is part of a plastic card. Method. 11. 8. The method of claim 7, wherein the facial image further includes a signature image. 12. For use in verifying the identity of a given group of individuals at multiple locations The method is applied to the digital bit stream, the method further comprising applying pressure to the digital bit stream. capturing information about said facial images to identify each individual; Capturing information about a signature image that identifies a person and converting the signature information for inclusion in a stream; a stage for recording the digital bit stream for the individual at a central location; floor and receiving the bit stream for the selected individual from the central location; and identifying the face from the bit stream to confirm the identity of the individual. providing a local device at each said location to reproduce the signature; The method according to claim 1. 13. the digital bit stream for the individual at the central location; providing each individual with a key to a recorded storage address; and communicating the bit stream to the local device based on the key information; 13. The method of claim 12, further comprising: 14. The key is stored in bits and pieces on a pocket-sized storage medium given to the individual. 14. The method of claim 13, wherein the method is stored as: