S&F Ref: 834393 AUSTRALIA PATENTS ACT 1990 COMPLETE SPECIFICATION FOR A STANDARD PATENT Name and Address Canon Kabushiki Kaisha, of 30-2, Shimomaruko 3-chome, of Applicant : Ohta-ku, Tokyo, 146, Japan Actual Inventor(s): Stephen James Hardy, Tuan Quang Pham Address for Service: Spruson & Ferguson St Martins Tower Level 35 31 Market Street Sydney NSW 2000 (CCN 3710000177) Invention Title: Paper fingerprint matching with compensation for small rotation The following statement is a full description of this invention, including the best method of performing it known to me/us: 5845c(10467x4 I) -1 PAPER FINGERPRINT MATCHING WITH COMPENSATION FOR SMALL ROTATION FIELD OF INVENTION The present invention relates to authenticating a physical medium of a document and, in particular, to a method of compensating for small rotation when using alpha masked correlation for determining the degree of resemblance between the digital signature of a 5 reference physical medium and the digital signature of a candidate physical medium. DESCRIPTION OF BACKGROUND ART The ability to establish the authenticity of a physical medium is critical to a large number of businesses and administrations. Administrative certificates, licences, forms, bank notes, company letterheads, coupons and vouchers all convey an important value in 10 being original. Various techniques already exist to facilitate document authentication. Most conventional certification systems require a physical addition to a document, for example in the form of a bar code, a hologram or a watermark. However, sophisticated counterfeiting techniques often eventually defeat these methods. Over the last decade, scientists have investigated a different approach to 15 authentication based upon the microscopic characteristics of the paper of the document. During the manufacturing process of paper, the structure of plant fibres is chemically (or mechanically) broken down, diluted and washed out from cellulose. The resulting pulp is then drained through a fine-web screen to form a fibrous sheet. Finally, the water is removed through pressing and drying using regular wire mesh screens. It has been 20 observed that the screens produce a significant imprint onto the paper during the manufacturing process. As a result, fibrous surfaces, such as paper, cardboard and other media having enough structural unicity, such as for instance plastic, present a complex 1046664_1 834393_speci -2 rough appearance on a microscopic level. The surface singularity of paper is not much affected by time, crushing or abrasion. The number of possible arrangements of the paper visual characteristics and the fluctuations in the pattern imprinted by the use of screens thus make each area of a sheet of 5 paper unique and therefore identifiable. Likewise, the complexity of the paper visual characteristics makes it very difficult to counterfeit. Examining the visual characteristics of paper in order to establish the authenticity of a document (or to detect counterfeiting) is used in so-called document fingerprinting methods. Typically, at an early stage of a document's lifecycle, a unique digital signature, or so-called fingerprint, is recorded based 10 on the surface appearance of a selected area of the document. That digital signature is then later compared to a digital signature obtained from a corresponding area in a candidate document. Some document fingerprinting methods exploit the unique speckle pattern produced by shining laser light onto plain paper, allowing a test to be developed for matching a strip 15 of paper with its previously-scanned "fingerprint". Other techniques have achieved similar results at lower cost without requiring the use of coherent light, making it possible to authenticate a paper fingerprint created using a standard scanner or photocopier. Evaluating the matching between two paper fingerprints is typically done using statistical correlation. The random distribution of paper fibres appears interestingly similar 20 to noise images. As a result, correlation provides a very sharp peak when the digital signatures match. Prior techniques have been developed for extracting a digital signature from a printed part of a document, based upon the microscopic non-uniformity of toner. The electro-photographic process comprises several steps amongst which the development, the 1046664_1 834393_speci -3 transfer and the fusing (or fixing), wherein the toner passes through electromagnetic fields, causing uncontrollable scattering of toner gains, and resulting in an unevenness of linear patterns only visible under a microscope. Although ink-jet printing technologies are substantially different, a similar artefact occurs due to the non-uniform penetration of ink. 5 Therefore, even when printing the same pattern twice, the distribution of individual toner points cannot be fully controlled and varies substantially. It has been shown that it is possible to exploit the local non-evenness of printed patterns in order to extract a reliable fingerprint. For instance, the fingerprint can consist of a feature vector computed from a mosaic of tiles that maps a certain area along the edges of the printed pattern, wherein each 10 tile conveys a toner density value. However, depending on the purpose and usage conditions of a document, an area that served to extract a reference digital signature may undergo substantial modification or alteration after the digital signature was extracted. Examples of modifications include further printing, stains, perforations, stamps, seals, official marks, signatures, etc. Consider 15 for example a company letterhead paper, or an administrative form, being printed upon or filled in after it has been created. In this instance, the modified area is likely to have a strong influence when evaluating the matching of the digital signatures and naturally, the larger the modified area, the stronger its influence. The applicant has observed that when printing two different documents with the 20 same piece of text a high correlation value is obtained due to the similarity of the printed areas, even though the sheets of paper are different. Therefore, printed parts may be responsible for false positives. When a large part of a document is printed after the reference digital signature is extracted, a low correlation value is obtained due to the 1046664_1 834393_speci -4 dissimilarity of the printed areas, even though the paper should be treated as authentic. In this case, the printing is responsible for a false negative. Some prior attempts to eliminate the effects of markings subsequent to extraction of the reference digital signature operate by detecting and 'filling' printed areas with an 5 alternative colour or pattern. Such attempts fail when such areas occupy a substantial portion of the total surface, typically more than half. A more recent technique for comparing digital signatures comprising a significant amount of markings associates a number of weight values with different parts of the fingerprints in order to quantify the contribution of each part in the correlation. This 10 technique typically uses two alpha-mask images al and a2 to encode contribution weights for each digital signature fi and f2, and is thus known as alpha-masked image matching. Each alpha mask al and a2 is a non-negative-valued image that associates a weight to each pixel of the corresponding digital signature. An alpha mask value of 0 at a particular pixel signifies that the pixel is removed from consideration when determining the match strength 15 value. The masks are obtained by applying a threshold to the original image followed by a dilation operator. Each pixel of the input images with a value lower than, or equal to, a pre-set threshold, say 150, is assigned a value of 0 in the corresponding binary mask. Similarly, each pixel of the input images with a value greater than this threshold is assigned a value of 1 in the corresponding binary mask. Given a shift (xo, yo) relating the two 20 fingerprintsfi andf 2 to be compared, a match strength value can be determined statistically in various manners as for instance using an alpha-masked correlation as defined below: (f,(x, y) - f 2 (X - X, - YO)) 2a, (x, y)a 2 (X - XO, y - Y 0 ) e(x 0 ,y 0 )= " y a,(x,y)a 2 (x -xO,y - yO) 1046664_1 834393_speci -5 It has also been demonstrated that the match strength value can equivalently be determined using normalized correlation as expressed below: 2Z f, (x, y)f 2 (x - x., y - yo)a, (x, y)a 2 (X - Xo, - YO) E(xo,y 0 =o X, C a,(x,y)a 2 (X - X 0 , -y- Y 0 ) xly Both alpha-masked correlation and normalized correlation can be achieved in the 5 spatial or in the spectral domain as more adequate. Finally the fingerprint images can be background subtracted, for instance using normalized averaging, in order to compensate for the differences in gain and bias that can result in creating the fingerprints with two different scanners. Pure alpha-masked image matching is however not robust to rotation or scaling. 10 This is especially a problem when the document to be authenticated is processed through the scanner's feeding tray (a.k.a. sheets feeders) or when the document, when manually placed on the scan bed, is not perfectly aligned with respect to the original scan conditions. Although the documents can be realigned by means of image registration techniques prior to determine a degree of matching, a small difference in alignment occasionally remains. 15 This difference translates into a few degrees rotation (typically less than 3 degrees) or a small scaling that relate the original and the candidate fingerprints. In the second case, the amount of scaling remains limited and its actual value is dependent upon the scanner's resolution. As a consequence, the match strength obtained when comparing two documents is likely to drop significantly, resulting in an authentic document being 20 misclassified. It has been observed that slight misalignments can make the classification of authentic documents less reliable, in particular compared to forgeries. 1046664_1 834393_speci -6 SUMMARY OF THE INVENTION Accordingly, the invention provides a method of establishing the authenticity of a physical medium comprising steps for: (a) Retrieving a reference digital signature previously determined based upon the 5 optical characteristics of an area of a reference physical medium; (b) Determining a candidate digital signature based upon visual characteristics of an area of said physical medium, said area of said physical medium being at least partly corresponding to said area of said reference physical medium; (c) Selecting a plurality of patches of said reference digital signature; 10 (d) Selecting a plurality of patches of said candidate digital signature, wherein each of said patches of said candidate digital signature at least partly corresponds to one of said patch of said reference digital signature; (e) Determining a shift estimate that relates each of said patch of said reference digital signature and said corresponding patch of said candidate digital signature; 15 (f) Determining the parameters of a two-dimensional projective transform that relates said patch of said reference digital signature and said corresponding patch of said candidate digital signature based on said correlation values; and (g) Determining an authenticity measure using said parameters. 20 BRIEF DESCRIPTION OF THE DRAWINGS One or more embodiments of the invention will now be described with reference to the following drawings, in which: Fig. 1 shows a schematic flow diagram of a method of compensating for small rotation and/or scaling when performing alpha-masked fingerprint matching 1046664_1 834393_speci -7 Fig. 2 illustrates by way of example the process of selecting a plurality of patches for estimating a value of rotation and/or scaling; and Fig. 3 is a schematic block diagram representation of a general purpose computer upon which the arrangements described may be practiced. 5 DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION Disclosed is a method of establishing the authenticity of a physical medium. The method retrieves a reference digital signature, such as a fingerprint, previously determined based upon the optical characteristics of an area of a reference physical medium. A 10 candidate digital signature is then determined based upon visual characteristics of an area of the physical medium, the area of the physical medium being at least partly corresponding to the area of the reference physical medium. A plurality of patches of the reference digital signature are then selected as are plurality of patches of the candidate digital signature, wherein each of the patches of the candidate digital signature at least 15 partly corresponds to one of the patches of the reference digital signature. The method then determines a shift estimate that relates each of the patches of the reference digital signature and the corresponding patch of the candidate digital signature. Various parameters of a two-dimensional projective transform that relates said patch of the reference digital signature and the corresponding patch of the candidate digital signature based on said 20 correlation values are then determined, followed by a determination of an authenticity measure using the parameters. The method of establishing the authenticity of a physical medium described herein may be implemented using a computer system 300, such as that shown in Fig. 3 wherein the processes of Figs. 1 and 2 may be implemented as software, such as one or more 1046664_1 834393_speci -8 application programs executable within the computer system 300. In particular, the steps of the method are effected by instructions in the software that are carried out within the computer system 300. The instructions may be formed as one or more code modules, each for performing one or more particular tasks. The software may also be divided into two 5 separate parts, in which a first part and the corresponding code modules performs the authentication methods and a second part and the corresponding code modules manage a user interface between the first part and the user. The software may be stored in a computer readable medium, including the storage devices described below, for example. The software is loaded into the computer system 300 from the computer readable medium, 10 and then executed by the computer system 300. A computer readable medium having such software or computer program recorded on it is a computer program product. The use of the computer program product in the computer system 300 preferably effects an advantageous apparatus for authentication of physical media. As seen in Fig. 3, the computer system 300 is formed by a computer module 301, 15 input devices such as a keyboard 302 and a mouse pointer device 303, and output devices including a printer 315, a display device 314 and loudspeakers 317. An external Modulator-Demodulator (Modem) transceiver device 316 may be used by the computer module 301 for communicating to and from a communications network 320 via a connection 321. The network 320 may be a wide-area network (WAN), such as the 20 Internet or a private WAN. Where the connection 321 is a telephone line, the modem 316 may be a traditional "dial-up" modem. Alternatively, where the connection 321 is a high capacity (eg: cable) connection, the modem 316 may be a broadband modem. A wireless modem may also be used for wireless connection to the network 320. 1046664_1 834393_speci -9 The computer module 301 typically includes at least one processor unit 305, and a memory unit 306 for example formed from semiconductor random access memory (RAM) and read only memory (ROM). The module 301 also includes an number of input/output (1/O) interfaces including an audio-video interface 307 that couples to the video 5 display 314 and loudspeakers 317, an I/O interface 313 for the keyboard 302 and mouse 303 and optionally a joystick (not illustrated), and an interface 308 for the external modem 316 and printer 315. In some implementations, the modem 316 may be incorporated within the computer module 301, for example within the interface 308. The computer module 301 also has a local network interface 311 which, via a connection 323, 10 permits coupling of the computer system 300 to a local computer network 322, known as a Local Area Network (LAN). As also illustrated, the local network 322 may also couple to the wide network 320 via a connection 324, which would typically include a so-called "firewall" device or similar functionality. The interface 311 may be formed by an EthernetTM circuit card, a wireless Bluetooth TM or an IEEE 802.11 wireless arrangement. 15 The interfaces 308 and 313 may afford both serial and parallel connectivity, the former typically being implemented according to the Universal Serial Bus (USB) standards and having corresponding USB connectors (not illustrated). Storage devices 309 are provided and typically include a hard disk drive (HDD) 310. Other devices such as a floppy disk drive and a magnetic tape drive (not illustrated) may also be used. An optical 20 disk drive 312 is typically provided to act as a non-volatile source of data. Portable memory devices, such optical disks (eg: CD-ROM, DVD), USB-RAM, and floppy disks for example may then be used as appropriate sources of data to the system 300. The components 305, to 313 of the computer module 301 typically communicate via an interconnected bus 304 and in a manner which results in a conventional mode of 10466641 834393_speci -10 operation of the computer system 300 known to those in the relevant art. Examples of computers on which the described arrangements can be practised include IBM-PC's and compatibles, Sun Sparcstations, Apple MacTM or alike computer systems evolved therefrom. 5 Typically, the application programs discussed above are resident on the hard disk drive 310 and read and controlled in execution by the processor 305. Intermediate storage of such programs and any data fetched from the networks 320 and 322 may be accomplished using the semiconductor memory 306, possibly in concert with the hard disk drive 310. In some instances, the application programs may be supplied to the user 10 encoded on one or more CD-ROM and read via the corresponding drive 312, or alternatively may be read by the user from the networks 320 or 322. Still further, the software can also be loaded into the computer system 300 from other computer readable media. Computer readable media refers to any storage medium that participates in providing instructions and/or data to the computer system 300 for execution and/or 15 processing. Examples of such media include floppy disks, magnetic tape, CD-ROM, a hard disk drive, a ROM or integrated circuit, a magneto-optical disk, or a computer readable card such as a PCMCIA card and the like, whether or not such devices are internal or external of the computer module 301. Examples of computer readable transmission media that may also participate in the provision of instructions and/or data include radio or 20 infra-red transmission channels as well as a network connection to another computer or networked device, and the Internet or Intranets including e-mail transmissions and information recorded on Websites and the like. The second part of the application programs and the corresponding code modules mentioned above may be executed to implement one or more graphical user interfaces 1046664_1 834393_speci -11 (GUIs) to be rendered or otherwise represented upon the display 314. Through manipulation of the keyboard 302 and the mouse 303, a user of the computer system 300 and the application may manipulate the interface to provide controlling commands and/or input to the applications associated with the GUI(s). 5 The method 100 of determining alpha-masked image matching that provides compensation for small rotation or scaling is now described with reference to Fig. 1. The method 100 is preferably executed as the software application in the computer system 300. The fingerprints may be scanned or otherwise input to the system 300 as bitmap images for example. A match strength value 120 is first determined by comparing the reference and 10 the candidate fingerprints using one of the alpha-masked image matching techniques 110 described in the background section. If the document is authentic and the two fingerprints were obtained under the same alignment conditions, the match strength is sufficient and the method ends in 170. If the document is not authentic, the match strength is typically very low. However, as described earlier, a weak match strength can also be obtained when the 15 document is authentic but the two fingerprints were slightly misaligned when scanned. For example, a slightly rotated authentic fingerprint can produce a similar match strength to that of a forged fingerprint. For this reason, the method determines in step 130 an estimation of the rotation or scaling and uses this estimation to re-align the two fingerprints relatively, by applying to one of them the inverse rotation and/or scaling. A second match 20 strength value is then determined in 140 based on the transformed fingerprints. If this second match strength value is again unsatisfying, the document is classified as fake 160. Conversely, if the second match strength is sufficient, the document is classified as authentic 170. 1046664_1 834393_speci -12 The step 130 of compensating for small rotation and/or scaling is now described in more details with reference to Fig. 2. Note that in this diagram the relative rotation difference is exaggerated for improved visibility. We assume that a first match strength value was obtained as described in 120 for a shift (xo, yo) relating a first fingerprint 210 5 (extracted from a reference document A) and a second fingerprint 220 (extracted from a corresponding area in a candidate document B). In practice, this correspondence is difficult to achieve precisely, and the areas used for extracting the fingerprints only partially match, thereby inducing a shift (xo, yo). In this instance, a rotation of a few degrees and/or a small value of scaling further relate the two documents and thereby the two fingerprints. If 10 document B is authentic, the fibrous structures of the two fingerprints will better match if they are realigned relatively. An estimate of the values of rotation and scaling is obtained by examining several pairs of corresponding patches from each of the fingerprints, within the overlapping area 230. In one embodiment of this invention, the patches are selected at the four corners of the overlapping area. The size of the selected patches can be set 15 arbitrarily (for instance 64x64 pixels), or relative to the size of the overlapping area (for instance half of its height and length respectively). A shift estimate is then determined for each of the selected patches using alpha-masked image matching, as represented in this example with four vectors i,, 9 2 , i 3 , i4 . Denote (x,,y) the end-point coordinates of vectors 9, fori e [l;4], and (x,,y,) the coordinates of the centre of each corresponding 20 patch. The four shift values can then be used to determine an estimate 240 of the rotation 0 that relates the two fingerprints for instance as follows: 0= atan { Y j 1046664_1 834393_speci -13 In this instance, the value of 0 is the average angle for each of the four displacements, with respect to the centre of the overlapping area. The estimated rotation value can be used to re-align the two fingerprints, as illustrated in 250, prior to recomputing the match strength in step 140. 5 The four shift values can similarly be used to determine an estimate of the scaling that relates the two fingerprints for instance as follows: /4 S = '= 4 0 1 The method can be further extended to estimate the six parameters of an affine transform relating the two fingerprints: 10 = + yy b2 b2 yY AY The displacement vectors can further be used with a standard least squares minimisation algorithm such as the Levenberg-Marquardt algorithm to find the parameters of a perspective transformation (b,,,b, b 1 3 , b 2 , b 2 ,b2 3 ,bA 2 b) that relate the two fingerprints: blx+b12 y+b13 15 b 3 x+ b 32 y+1 _ b 2 x+b 22 y+b 23 b 3 x+b 32 y+I Various patch distributions can be used for estimating a transform. The minimum number of patches and their spatial distribution depends upon the type of transform. Although rotation can be determined using only one patch, several patches will provide more robustness. Scaling estimation typically requires a minimum of three patches. The 1046664_1 834393_speci -14 further away the patches from the centre of the overlapping area, the more accurate the transformation parameters can be estimated. Alternatively, if the whole surface of any of the fingerprints was covered with ink or altered in any way, the patches can preferably be selected in areas with fewer markings. 5 Industrial Applicability The arrangements described are applicable to the computer and data processing industries and particularly for the authentication of documents. The foregoing describes only some embodiments of the present invention, and modifications and/or changes can be made thereto without departing from the scope and 10 spirit of the invention, the embodiments being illustrative and not restrictive. In the context of this specification, the word "comprising" means "including principally but not necessarily solely" or "having" or "including", and not "consisting only of". Variations of the word "comprising", such as "comprise" and "comprises" have correspondingly varied meanings. 15 1046664_1 834393_speci