METHOD FOR AUTHENTICATING A VISUAL ITEM USING A STORED DIGITAL FINGERPRINT AND AN AUTHENTICATION DEVICE, AUTHENTICATION DEVICE AND DIGITAL FINGERPRINT GENERATION DEVICE

Abstract

The present invention relates to the technical field of optical detection of the authenticity or inauthenticity of visual items. In particular, the invention relates to the technical field of encryption methods and authentication methods, respectively, of a digital representation of a visual item using an encryption device and an authentication device.

Description

METHOD AND DEVICE FOR THE AUTHENTICATION OF A VISUAL ITEM Technical field The present invention relates to the technical field of optical detection of the authenticity or inauthenticity of visual items. In particular, the invention relates to the technical field of encryption methods and authentication methods, respectively, of a digital representation of a visual item using an encryption device and an authentication device. Background of the invention The problems of tampering with documents or valuables are well-known and growing daily. For example, valuable documents such as an identification card, passport, driver's license, etc., generally contain a photo of the holder and text containing personal information such as name, date of birth, etc. Various techniques can be used to manipulate these types of documents. In some situations, manipulation can involve replacing the visual item, for example, the photograph on an ID card. This substitution may involve only minor modifications, such as using a transformation process, or it may be a complete replacement of the photograph. All of these techniques can be very difficult to detect with the naked eye. In fact, to the human eye, the replacement of the photograph may be undetectable, and to an automatic optical reader, the transformation technique can be quite effective. Regarding these technical problems, several prior art solutions have been proposed. The most commonly used involve material-based security features. These material-based security features can be based on security inks, for example. However, these solutions often involve physically adding features to the visual item or at least to the medium that carries that visual item—that is, the document of value. This can lead to a design problem or an increased cost to produce that document of value. It may also require some modification to the supply chain. 2088382 of 55 Beyond the realm of valuable documents, the manipulation of other types of visual items is also a problem. For example, paintings or collectible cards can be targeted. In this case, the use of material-based security features remains a solution. However, this solution can be costly and cannot be directly implemented on the visual item itself in some situations. Furthermore, the use of material-based security features typically involves the use of dedicated reading devices configured to assess whether or not the valuable document has been altered. These dedicated reading devices are also expensive and cannot always be owned by the end customer, for example. Therefore, it is an object of the invention to solve at least partially some of these technical problems. Summary of the invention According to one aspect, the present invention relates to a method for authenticating a visual article VI using a stored digital fingerprint F0 and an authentication device AD, said authentication device AD ​​comprising at least one optical unit OPT1 and at least one processing unit CPU1, said stored digital fingerprint Fo being previously generated from an authentic visual article AVI, said method comprising the following steps: a. acquiring, using said optical unit OPT1, at least a plurality of images I of an area comprising the visual article VI to be authenticated, said optical unit OPT1 being in communication with said processing unit CPU1, each image It of the plurality of images I comprising at least partially a digital representation of said visual article VI, the image It being acquired in a time t; and b. for each image It of the plurality of images I, generating, by means of said processing unit CPU1, a spatially corrected image Ict by correcting the image It based on at least one spatial feature to calibrate at least partially a digital representation of said visual article VI (50) so that it is in the same perspective as that of said authentic visual article AVI (10), creating a plurality of spatially corrected images Ic; and c. for each spatially corrected Ict image of the plurality of 2088382 of 55 spatially corrected images Ic: i. extract, by means of said processing unit CPU1, a plurality of features; and ii. generate, by means of said processing unit CPU1, a digital fingerprint Ft of the digital representation of the visual article VI from the spatially corrected image Ict using at least a part of said plurality of extracted features; and iii. calculate, by means of said processing unit CPU1, at least one distance metric D(Ict) between said digital fingerprint Ft and said stored digital fingerprint F0; and iv.calculate, using said processing unit CPU1, a first likelihood function L(Ict|H) of authenticity H of the visual article VI from its digital representation of the spatially corrected image Ict based on said calculated distance metric D(Ict) and a second likelihood function L(Ict|G) of non-authenticity G of the visual article VI from its digital representation of the spatially corrected image Ict based on said calculated distance metric D(Ict); and. v. compute, by means of said processing unit CPU1, a probability P(H) that the visual item VI is authentic using the first likelihood function L(Ict|H) and the second likelihood function L(Ict|G); and wherein the probability P(H) is updated as each spatially corrected image Ict of the plurality of spatially corrected images is processed and the authentication of the visual item VI is confirmed, by means of said processing unit CPU1, if the probability P(H) is greater than a predetermined threshold. The present invention allows for the detection of the authenticity of a visual item based on a stored digital fingerprint. The present invention allows for the authentication of a visual item based on a plurality of images of said visual item and a stored digital fingerprint. The present invention can be used with respect to any type of visual article. The present invention advantageously uses likelihood functions to update a probability of authenticity, said update being based on 2088382 of 55 in each new image of the visual article that is processed by the invention. The present invention allows the use of a plurality of images of a visual item to compute the probability of authenticity of said visual item based on a stored digital fingerprint. Advantageously, each new processed image of the visual item updates the probability of authenticity. The present invention eliminates the need for material-based security features to help a user determine whether a visual item is authentic. Only a few images of a visual item and a stored digital fingerprint of the authentic visual item are required to confirm its authenticity. According to another aspect, the present invention relates to an AD authentication device configured to authenticate a visual item VI using a stored digital fingerprint F0, said stored digital signature Fo being previously generated from an authentic visual item VI, said AD authentication device comprising: a. an optical unit OPT1 configured to acquire at least a plurality of images I of an area comprising the visual item VI to be authenticated, each image It of the plurality of images I comprising at least partially a digital representation of said visual item VI, the image It being acquired in a time t; and b. a CPU1 processing unit configured to: i. for each image It of the plurality of images I, generate a spatially corrected image Ict by correcting the image It based on at least one spatial feature to calibrate at least partially a digital representation of said visual article VI (50) so that it is in the same perspective as that of said authentic visual article AVI (10), creating a plurality of spatially corrected images Ic; and ii. for each spatially corrected image Ict of the plurality of spatially corrected images Ic: (1) extract a plurality of features; and (2) generate a digital fingerprint Ft of the digital representation of the visual item VI from the spatially corrected image Ict using at least a part of said plurality of features 2088382 of 55 extracted; and (3) calculate at least one distance metric D(Ict) between said fingerprint Ft and said stored fingerprint Fo; and (4) calculate a first authenticity likelihood function L(Ict|H) H of the visual article VI from its spatially corrected image digital representation Ict based on said calculated distance metric D(Ict) and a second non-authenticity likelihood function L(Ict|G) G of the visual article VI from its spatially corrected image digital representation Ict based on said calculated distance metric D(Ict); and (5) compute a probability P(H) that the visual item VI is authentic using the first likelihood function L(Ict|H) and the second likelihood function L(Ict|G), updating the probability P(H) as each spatially corrected image Ict of the plurality of spatially corrected images is processed;and (6) confirm that the visual item VI is authentic if the probability P(H) is greater than a predetermined threshold.; This allows a user to verify the authenticity of a visual item using the present invention. According to another aspect, the present invention relates to a digital fingerprint generating device GD configured to generate a digital signature Fo of an authentic visual item AVI, said digital fingerprint Fo being configured to be stored, said digital signature generating device GD comprising: a. an OPT2 optical drive configured to acquire at least one image I of the authentic visual article AVI and / or a download drive configured to download at least one image I of the authentic visual article AVI from at least one server; and b. a CPU2 processing unit configured to: i. preferably, generate a spatially corrected image Ic by correcting image I based on at least one spatial feature; and ii. extract a plurality of features from said image I and / or 2088382 of 55 preferably from said spatially corrected image Ic; and iii. generate the digital fingerprint Fo of the digital representation of the authentic visual article VI from image I, and / or preferably from the spatially corrected image Ic, using at least a part of said plurality of extracted features; and c. Preferably, an SU2 storage unit configured to store the Fo fingerprint in at least one of the following forms: a QR code, a data matrix, a barcode, a serial number, a watermark, a digital watermark, metadata, data stored in memory, or data stored on a server. This allows a fingerprint to be generated from an authentic visual item using a single image. Before providing a detailed review of the embodiments of the invention below, some optional features that may be used in conjunction with or as an alternative will be listed below: According to one example, the probability computation stage P(H) comprises the following stages: to. b. and c. Establish, using the CPU1 processing unit, a prior probability P0(H) based on a predetermined set of rules, and for each spatially corrected image Ict, calculate, using the CPU1 processing unit, a posterior probability Pt(H) as follows: =¿(Zct|H)Po(H)=___________¿(Zct|H)Po(H)___________t() P( / ct) L(Zct|H)Po(H) + L(Zct|G)(1-Po(H)) Calculate, using the CPU1 processing unit, the probability P(H) by computing a weighted moving average as follows: Σ^οΚ^-^) P(H) = ΣΓ=οH where w is the number of spatially corrected images Ict of the plurality of spatially corrected images Ic and {y¿} (i = 0,...w) are predetermined weights. This allows the probability P(H) of authenticity to be computed based on each new corrected Ict image. This allows the probability P(H) to be computed from a previous probability P0(H), this previous probability P0(H) is updated using each new 2088382 of 55 corrected image Ict through the first and second likelihood functions. Each update applied to the probability P(H) is weighted advantageously; this allows varying the impact of an Ict-corrected image, for example, based on a quality score. According to an example, the probability P(H) is computed from a probability distribution p(P(H)) and P(H) is related to at least one descriptive statistic of the probability distribution p(P(H)), and preferably P(H) is a mathematical expectation of p(P(H)) as follows: P(H) = fp(y)y dy Jo Using a probability distribution allows for the calculation of more descriptive statistics for P(H). According to one realization, this allows us to calculate a confidence interval. This confidence interval can be used to make a better decision than just a probability estimate based on points P(H), because using a probability distribution p(P(H)) allows us to have both the probability P(H) and an uncertainty in this probability P(H). Preferably, the confidence interval can be calculated using percentiles. For example, the 90% confidence interval for P(H) lies between the 5th and 95th percentiles of the probability distribution p(P(H)). According to one realization, the computation stage of the probability P(H) comprises the following stages: to. b. c. establish, using the CPU1 processing unit, a prior probability distribution p0(P(H)) based on a predetermined set of rules, and sample, using the CPU1 processing unit, K independent prior probability values ​​{Po,1,, , P0,K} from the prior probability distribution p0(P(H)); and use each of these K prior probabilities {P0,k}fc_i K, calculate, using the CPU1 processing unit (42), K posterior probabilities {Pt,fc(H)}K as follows: £( / ct|H)Po(Hf___________L(IctlH)Po(^)___________t,k() PUG) L( / ct|H)Po(H) + L( / ct|6)(1-Po(H)) 2088382 of 55 d. Use the K posterior probabilities {Pt,k(H)}fc_i K, adapting, by means of the CPU1 processing unit, a new posterior probability distribution pt(P(H)), and e. Calculate, using the CPU1 processing unit, the probability distribution p(P(H)) by computing a weighted moving average as follows: P(P(H)) = Zr=oatPt-i(P(H)) ΣΓ=ο^ where w is the number of spatially corrected images Ict of the plurality of spatially corrected images Ic and {aj (i = 0,...w) are predetermined weights. This allows the probability P(H) of authenticity to be computed based on each new corrected Ict image. This allows computing a probability distribution p(P(H)) from a previous probability distribution p0(P(H)), said previous probability distribution p0(P(H)) is updated using each new corrected image Ict through the first and second likelihood functions. Each update applied to the probability distribution p(P(H)) is advantageously weighted; this allows varying the impact of an Ict-corrected image, for example, based on a quality score. According to an example, the calculation stage of the first likelihood function L(Ict|H) based on the distance metric D(Ict) comprises the following stages: a. adapting, by means of the CPU1 processing unit, a probability distribution or probability density function PD1 with respect to a distance metric D(SD1) of at least one set of training data SD1, said training data SD1 corresponding to authentic visual items; and b. For each image Ict, calculate, using the CPU1 processing unit, the first likelihood function L(Ict|H) as a probability or probability density given by PD1 to the distance metric D(Ict); and, the calculation stage of the second likelihood function L(Ict|G) based on the distance metric D(Ict) comprises the following stages: c. adapt, using the CPU1 processing unit, a probability distribution or probability density function PD2 of a metric 2088382 of 55 distance D(SD2) from at least one set of SD2 training data, said SD2 training data corresponding to non-authentic visual items; and d. For each image Ict, calculate, using the CPU1 processing unit, the second likelihood function L(Ict|G) as a probability or probability density given by PD2 to the distance metric D(Ict). This allows for an improved calculation of the probability P(H) by calculating the first and second likelihood functions based on the distance metric. According to an example, the calculation stage of the first likelihood function L(Ict|H) based on the distance metric D(Ict) comprises the following stages: a. adapting, by means of the CPU1 processing unit, a probability distribution PD1 of a distance metric D(PD1) from at least one set of training data SD1, said training data SD1 corresponding to authentic visual items; and b. For each image Ict, calculate, using the CPU1 processing unit, the first likelihood function L(Ict|H) as probability density given by PD1 to the distance metric D(Ict); and, the calculation stage of the second likelihood function L(Ict|G) based on the distance metric D(Ict) comprises the following stages: c. adapting, by means of the CPU1 processing unit, a probability distribution PD2 of a distance metric D(PD2) from at least one set of training data SD2, said training data SD2 corresponding to non-authentic visual items; and d. For each image Ict, calculate, using the CPU1 processing unit, the second likelihood function L(Ict|G) as the probability density given by PD2 to the distance metric D(Ict). This allows for improved computation of the probability distribution p(P(H)) by calculating the first and second likelihood functions based on the distance metric. According to one example, the first likelihood function L(Ict|H) is generated at least partially by an artificial intelligence algorithm A1 using at least one training dataset SD1, comprising 2088382 of 55 said first likelihood function L(Ict|H) at least one subfunction SF1 defined by a probability density that the visual item VI is authentic, said subfunction SF1 being generated by the artificial intelligence algorithm A1 that uses said training data SD1, said training data SD1 corresponding to authentic visual items, said first subfunction SF1 corresponding to a mathematical model of authentic visual items; and the second likelihood function L(Ict|G) is generated at least partially by an artificial intelligence algorithm A1' using at least one training data set SD2, said second likelihood function L(Ict|G) comprising at least one subfunction SF2 defined by a probability density that the visual item VI is not authentic, said subfunction SF2 being generated by the artificial intelligence algorithm A1' using said training data SD2, said training data SD2 corresponding to non-authentic visual items, said subfunction SF2 corresponding to a mathematical model of non-authentic visual items. The use of an artificial intelligence algorithm allows the present invention to be trained to optimize the first and second likelihood functions. This allows the first and second likelihood functions to be adapted to different situations. According to an example, the visual article VI is transported by a medium ME and the first likelihood function L(Ict|H) is generated at least partially using an artificial intelligence algorithm A2 configured to generate at least one linear combination of mathematical models of the media for each spatially corrected image Ict based on at least a plurality of mathematical models of the media, and the second likelihood function L(Ict|G) is generated at least partially using an artificial intelligence algorithm A2' configured to generate at least one linear combination of mathematical models of the media for each spatially corrected image Ict based on at least a plurality of mathematical models of the media. This allows the nature of the medium to be considered in the calculations of the present invention. This allows us to extract some characteristics of the medium and then, 2088382 out of 55 classify this medium based on training data. According to one example, the artificial intelligence algorithms A2 and A2' comprise the following stages: a. extract, using the CPU1 processing unit, feature vectors of local binary patterns in pixels among a plurality of pixels from each spatially corrected image Ict of the plurality of spatially corrected images I; and b. calculate, using the CPU1 processing unit, the histogram of the feature vectors of local binary patterns in at least a portion of each spatially corrected image Ict of the plurality of spatially corrected images Ic; and c. to enable, by means of the CPU1 processing unit, a classifier that generates the medium from feature vectors of local binary patterns based on a plurality of mathematical models of the media; and d. using said classifier, generating, by means of the CPU1 processing unit, at least one probability for each mathematical model of the media from the plurality of mathematical models for each spatially corrected image Ict from the plurality of spatially corrected images Ic that the visual item VI is transported by a certain type of media; and e. generate, using the CPU1 processing unit, at least one linear combination of mathematical models of said media for the spatially corrected image Ict. This allows the nature of the medium to be considered in the calculations of the present invention. This allows us to extract some characteristics of the medium and then classify this medium based on training data. This allows for improved accuracy of the invention based on the nature of the medium. According to one example, the method comprises, before the step of extracting the plurality of features of a spatially corrected image Ict from the plurality of spatially corrected images Ic, a step for calculating, by means of the processing unit CPU1, a quality score of each image Ict from the plurality of spatially corrected images Ic, and 2088382 of 55 only if said quality score is greater than a predetermined threshold, the stage of extracting the plurality of features from said Ict image is executed. This allows the corrected images to be classified based on their quality score. This allows you to remove corrected images that have a poor quality score, for example. This allows you to promote corrected images that have a good quality score, for example. This allows weighting the probability Pt(H) and / or the probability distribution pt(P(H)) in the computation stage of the probability P(H). According to an example, the distance metric D(Ict) is calculated using a matrix Q as a mathematical operator between the stored fingerprint F0 and the fingerprint Ft. This matrix Q is generated using an artificial intelligence algorithm A3. This AI algorithm A3 is configured to generate this matrix Q based on at least two training datasets SD3 and SD4, such as: a. the matrix Q maximizes the distance metric D(Ict) using said SD3 training data, said SD3 training data corresponding to authentic visual items; and b. The Q matrix minimizes the distance metric D(Ict) using said SD4 training data, said SD4 training data corresponding to non-authentic visual items. This allows for the consideration of complex interactions when calculating the distance metric D(Ict) between the generated fingerprint Ft and the stored fingerprint Fo. According to one example, the default set of rules is established based on an A4 artificial intelligence algorithm configured to generate respectively a prior probability P0(H) or a prior probability distribution p0(P(H)) based on at least one of these parameters: a. a reputation score based on the nature of the visual item and / or the location of the visual item and / or metadata related to the visual item and / or a sender of the visual item and / or a sender of a medium that carries the visual item, a uniform distribution law, and / or based on at least one of the following processes: 2088382 of 55 b. a decision tree; and c. forests consisting of decision trees. This allows for the consideration of various parameters to establish the default set of rules. According to one example, the stored fingerprint Fo comprises a vector V0 comprising a predetermined number N of elements, and the fingerprint Ft comprises a vector Vt comprising a number M of elements, where M < N and where M is a function of the spatially corrected image quality score Ict. This allows adapting the calculation of the distance metric D(Ict) based on the length of the generated fingerprint Ft. This allows adapting the calculation of the distance metric D(Ict) based on, for example, the Ict corrected image quality. Brief description of the drawings The aims, objects, as well as the technical features and advantages of the invention will be best gleaned from the detailed description of an embodiment of the invention, which is illustrated by the following figures in which: - Figure 1 is a general schematic view of an authentication method according to an embodiment of the present invention. - Figure 2 is a flowchart of an authentication method according to an embodiment of the present invention. - Figure 3 is a schematic view of an authentication device according to an embodiment of the present invention. - Figure 4 is a general schematic view of a fingerprint generation method according to an embodiment of the present invention. - Figure 5 is a flowchart of a fingerprint generation method according to an embodiment of the present invention. - Figure 6 is a schematic view of an example of a fingerprint generating device according to an embodiment of the present invention. - Figure 7 is a schematic view of an example implementation of the present invention. Figure 8 is a schematic view of another example implementation of 2088382 of 55 the present invention. - Figure 9 is a schematic view of another example of implementation of the present invention. - Figure 10 is a schematic view of another example of implementation of the present invention. - Figure 11 is a schematic view of a feature vector extraction step from a mathematical transformation of an image according to an embodiment of the present invention. - Figure 12 is a schematic view of a first likelihood function L(Ict|H) of authenticity H of a visual article based on a distance metric D(Ict) and of a second likelihood function L(Ict|G) of non-authenticity G of the visual article based on a distance metric D(Ict) according to an embodiment of the present invention. - Figure 13 is a schematic representation of the relationship between the predetermined threshold, the estimated rate of false authentic detection events, and the rate of false non-authentic detection events. The drawings are provided for illustrative purposes only and do not limit the invention. They are representations of the principle intended to facilitate understanding of the invention and are not necessarily to scale for practical applications. Detailed description The present description is described here in detail by reference to non-limiting realizations illustrated in the drawings. The present invention relates to a method for authenticating a visual item using a stored digital fingerprint. This stored digital fingerprint has been generated according to a fingerprint generation method described below. The stored digital fingerprint is generated from an authentic visual item. Advantageously, this stored digital fingerprint is generated by an issuer, preferably an official issuer, accredited to issue a stored digital fingerprint, i.e., to establish the authenticity of an authentic visual item. For example, this issuer could be a government agency or a public administration, etc. The present invention relates to a solution for verifying the authenticity of a visual item. To do this, a digital fingerprint is generated from the visual item to be authenticated and compared with a digital fingerprint. 2088382 of 55 stored from the authentic visual article, preferably issued by an official and / or accredited issuer. According to one embodiment, and as described below in this document, the issuer of the stored fingerprint may use a fingerprint generation device to generate said stored fingerprint. According to one embodiment, and as described below in this document, a user, controller, or inspector may use an authentication device to authenticate whether a visual item is authentic or not using such stored fingerprint. The present invention can be used in many different cases. Some of these are described in detail below. One such case is a valuable document, such as an identification card, comprising the photograph of the owner of said identification card and a stored fingerprint corresponding to the fingerprint of the photograph of the owner of the identification card, said fingerprint, also called the stored fingerprint, having been generated by the official issuer of said identification card. According to one embodiment, to authenticate a person's identification card, a user employs an authentication device to extract the stored fingerprint and generate a new fingerprint from the photograph on the identification card. The generated fingerprint is then compared to the stored fingerprint to notify the user whether the photograph on the identification card is authentic. According to one embodiment, the authentication device is configured to execute an authentication method comprising several stages to compare a generated fingerprint with the stored fingerprint. As described below, the visual item can be selected from various items, such as a valuable document, an identity photo, a painting, a trading card, etc. The present invention is described below in this document using Figures 1 to 13 as schematic illustrations of some embodiments of the present invention. The authentication method According to one embodiment, the present invention relates to a 2088382 of 55 method to authenticate 200 a visual item VI 50. Said authentication method 200 uses at least the stored fingerprint F011. Said authentication method 200 is executed by an AD 40 authentication device described in Figure 3. As described below in this document, said AD 40 authentication device comprises at least: a. an OPT1 41 optical drive, and b. a CPU1 42 processing unit. Advantageously, said CPU1 42 processing unit and said OPT1 41 optical unit are in communication with each other, i.e., the OPT1 41 optical unit can send data to the CPU1 42 processing unit and / or the CPU1 42 processing unit can send data to the OPT1 41 optical unit. According to one embodiment, said OPT1 41 optical unit can be a camera, a camera system, an optical sensor, an array of optical sensors, a network of optical sensors, etc. According to one embodiment, such an AD 40 authentication device can be a smartphone, a computer, a laptop, a dedicated reader, etc. As described above, and as illustrated in Figure 4, for example, such stored fingerprint F0 11 has been previously generated from an authentic visual article AVI 10, preferably using a fingerprinting device GD 20, which advantageously uses a fingerprinting method 100 described below in this document. According to one embodiment and as illustrated in Figures 1 and 2, the authentication method 200 comprises at least the following steps: a. acquiring 210, preferably using said optical unit OPT1 41, at least a plurality of images I 44 of the visual article VI 50 to be authenticated, preferably of an area comprising the visual article VI 50 to be authenticated, advantageously each image It of the plurality of images I 44 comprises at least partially a digital representation of said visual article VI 50, preferably the image It is acquired in a time t; and b. For each image It of the plurality of images I 44, generate 220, by means of said processing unit CPU1 42, a corrected image 2088382 of 55 Ict; said Ict-corrected image is preferably spatially corrected; all these Ict-corrected images form a plurality of Ic-corrected images, preferably a plurality of spatially corrected Ic images; and c. for each corrected image Ict of the plurality of corrected images Ic: i. extract 230, by means of said processing unit CPU1 42, a plurality of features; and ii. generate 240, by means of said processing unit CPU1 42, a digital fingerprint Ft 45 of the digital representation of the visual article VI 50 of the corrected image Ict that uses at least a part of said plurality of extracted features; and iii. calculate 250, by means of said processing unit CPU1 42, at least one distance metric D(Ict) between said digital fingerprint Ft 45 and said stored digital fingerprint F0 11; and iv.calculate 260, using said processing unit CPU1 42, a first likelihood function L(Ict|H) of authenticity H of the visual article VI 50 from its digital representation of the corrected image Ict based on said calculated distance metric D(Ict) and a second likelihood function L(Ict|G) of non-authenticity G of the visual article VI 50 from its digital representation of the corrected image Ict based on said calculated distance metric D(Ict); and v. compute 270, by means of said processing unit CPU1 42, a probability P(H) that the visual article VI 50 is authentic using the first likelihood function L(Ict|H) and the second likelihood function L(Ict|G), preferably the probability P(H) is updated each time a new fingerprint Ft 45 is generated, i.e., each time steps 230 to 260 are executed with respect to a new corrected image Ict; and. d. confirm 280 the authentication of the visual item VI 50, by means of said processing unit CPU1 42, if the probability P(H) is greater than a predetermined threshold. According to one embodiment, each image It of the plurality of images I 44 comprises an area comprising at least a part of the visual article VI 50 to be authenticated, i.e., each image It of the plurality of images I 44 comprises at least partially a digital representation of 2088382 of 55 said visual item VI 50. According to one embodiment, each image It is a frame extracted from a set of frames from a video, advantageously acquired by the optical unit OPT1 41, preferably in real time. In fact, according to one embodiment, the optical unit OPT1 41 can be configured to acquire a video of an area comprising at least a portion of the visual item VI 50 to be authenticated, and each image It is derived from said video. According to one embodiment, these images It are selected based on a random selection from said video and / or correspond to a continuous portion of frames from said video. According to another embodiment, based on the quality score discussed below, some images It can be removed from the plurality of images I 44 and / or other images It can be extracted from the video and added to the plurality of images I 44. According to one embodiment, the image It of the plurality of images I 44 is acquired during the processing of the image It-1 by the processing unit CPU1 42, said processing comprising the steps of generating 220 a corrected image Ic, extracting 230 a plurality of features, generating 240 a fingerprint Ft-1 45, calculating 250 the distance metric D(Ict-1), calculating 260 the first likelihood function L(Ict-1|H) of authenticity H of visual item VI 50 and a second likelihood function L(Ict-1|G) of non-authenticity G of visual item VI 50 and computing 270 the probability P(H) that visual item VI 50 is authentic. In fact, according to this realization, the stage of generating the corrected image 220 is running while the stage of acquiring the plurality of images 210 is still running. According to one implementation, at least some of the following steps of the authentication method 200 are executed in real time when the It image is acquired: a. generate a corrected image Ic from image It-1, and b. extract 230 a plurality of features from the corrected image It-1, and c. generate 240 a digital fingerprint Ft-1 45, and d. Calculate the distance metric D(Ict-1) 250, and e. Calculate the first likelihood function L(Ict-1|H) of authenticity H of visual item VI 50 and a second likelihood function L(Ict-1|G) of non-authenticity G of visual item VI 50, and 2088382 of 55 f. compute 270 the probability P(H) that the visual item VI 50 is authentic. According to one embodiment, prior to acquiring the It image, the computation stage 270 is run based on the It-1 image. According to one embodiment, the processing steps with respect to image It can be performed before and / or during and / or after the acquisition of image It+1. According to one embodiment, the plurality of images I 44 can be acquired from at least one digital file, for example, by downloading at least one digital file. According to this embodiment, the optical unit OPT1 41 may comprise a module configured to download at least one image of at least one visual item, preferably using a QR code as a link to download said at least one image of at least one visual item. According to one realization, the OPT1 41 optical unit can be at least a smartphone camera. According to one embodiment, the step of generating 220 corrected Ict images for each It image from the plurality of I44 images is configured to identify at least one spatial feature in each digital representation of the visual item of the It image, and then use this spatial feature to spatially correct the It image in order to create a plurality of spatially corrected Ic images. Several methods of spatially correcting the It images can be used and are well known to those skilled in the art. These corrections may be necessary due to the orientation of the OPT1 optical unit with respect to the VI50 visual item. These corrections may be linked to perspective misalignment, for example. According to one embodiment, the step of generating an Ict-corrected image 220 may comprise a step of detecting a contour, i.e., an edge, and / or at least a portion of a contour and / or an edge, and / or at least a mark on the visual article VI 50 and / or the medium 51 carrying it, enabling the CPU1 42 processing unit to determine the spatial orientation of the visual article VI 50 with respect to the optical unit OPT1 41 and / or the spatial orientation of the optical unit OPT1 41 with respect to the visual article VI 50. According to one embodiment, the CPU1 42 processing unit 2088382 of 55 can use the OPT1 41 optical unit and at least one sensor to evaluate the orientation in space of the OPT1 41 optical unit with respect to the visual article VI 50, such as a gyroscope and / or an accelerometer. According to one embodiment, the detection of an edge, i.e., a contour, in the digital representation of visual article VI 50 may comprise the detection of at least three of the four corner points of an edge, preferably the detection of all four corner points of an edge or contour. This detection may be performed using, for example, the TensorFlow Lite implementation of the EfficientNet deep learning model, which is common among subject matter experts. Since this detection uses deep learning, it may be trained using different sets of images and / or edges and / or spatial orientations. According to one embodiment, the detection of an edge, i.e., a contour, in the digital representation of visual article VI 50 may comprise the detection of at least a portion of the corners of a polygon. According to one embodiment, the step of generating a corrected or spatially corrected Ict image 220 may comprise a step of estimating the projective transform between the coordinate system of the optical unit OPT1 41 and the world coordinate system, i.e., the coordinate system in which the visual article VI 50 is located. Once this projective transform has been estimated, a step of deforming the It image into the Ict image is executed, preferably by means of the CPU1 42 processing unit, to generate a spatially corrected Ict image. According to one embodiment, the main objective of correcting the It images is to calibrate the digital representation of the visual item VI 50 so that it is in the same perspective as the perspective of the actual visual item during the stage of generating the stored digital fingerprint F0, as described below. Preferably, correcting these It images allows obtaining corrected Ict images that are spatially oriented in a virtual plane parallel to the lens plane of the OPT1 optical unit. According to one realization, the stage of generating a corrected Ict image may involve correcting some of the characteristics of the It image such as brightness, contrast, saturation, etc. According to one realization, the stage of extracting 230 a plurality of 2088382 of 55 features of each Ict-corrected image may comprise the extraction of at least one feature vector from each Ict-corrected image, preferably from each digital representation of the visual item VI 50 comprised by each spatially corrected Ict image. For example, this feature vector may be based on the discrete cosine transform (DCT) and may be computed in local regions, for example, a rectangle, of each Ict-corrected image. For example, each region is resized to a predefined size, then a DCT is applied to that region, preferably a two-dimensional DCT or even a three-dimensional DCT if the digital representation of the visual item comprises three-dimensional data, then the frequency response of the upper left portion of the DCT transform is taken and used to form such a feature vector.This is the feature vector of a region, also called the feature of that region. The same process is repeated for different regions, and the present invention allows for the extraction of a final vector that is advantageously the concatenation of the feature vectors of all the regions. According to one embodiment, each value of the feature vector is quantified using the median / percentiles of the feature vector. In this way, the feature vector is converted into a binary vector. For example, considering a feature vector [1,2,3,4,5,6,7,8,9,10,11,12], if the median, which is equal to 6.5 in this example, is used to quantify the feature vector, i.e., to divide the feature vector into two subranges, then the feature vector becomes [0,0,0,0,0,0,1,1,1,1,1,1], which is a binary vector.For example, if the following percentiles are considered: between 0% and 25%, between 25% and 50%, between 50% and 75%, and between 75% and 100% to divide said feature vector into 4 subintervals, then the feature vector becomes [0,0,0,1,1,1,2,2,2,3,3,3], preferably then said feature vector can be binarized again into [00, 00, 00, 01, 01, 01, 10, 10, 10, 11, 11, 11]. According to one embodiment, these different regions can be all the regions of the Ict-corrected image; that is, the entire Ict image is considered at this stage 230. According to another embodiment, only one region can be used if the Ict-corrected image is relatively small, for example, no more than a few hundred pixels by a few hundred pixels. 2088382 of 55 This example, this single region refers to the entire corrected Ict image. According to another embodiment, the Ict-corrected image can be divided into a grid, thus defining different regions of the Ict-corrected image. All of these regions can be used to generate the final vector, or only a predetermined number of regions can be randomly selected to form a subset used to generate that vector. When there are several regions to consider, the feature vector of the Ict-corrected image for each region is concatenated to form a single final vector representing all the considered feature vectors. According to one embodiment, and as described below, a region of the Ict-corrected image can be ignored during this step 230, such a region comprising, for example, the stored fingerprint Fo 11.In fact, as discussed below in this document, when the stored digital fingerprint Fo 11 is found in the visual article itself, the region comprising it is not considered by the stage of extracting 230 said plurality of features. According to one embodiment, the step of generating a digital fingerprint Ft 45 of the digital representation of the visual item VI 50 from an Ict-corrected image is configured to use at least a part of the extracted features, i.e., from the final vector, to generate said Ft 45 fingerprint. Advantageously, the final vector of an Ict-corrected image is the digital fingerprint Ft 45 of said Ict-corrected image, i.e., of the digital representation of the visual item VI 50 from said Ict-corrected image. According to one embodiment, the step of generating the Ft 45 fingerprint 240 may comprise a step of selecting a subset of features from said plurality of features, i.e., a subset of elements from said final vector. According to one embodiment, said step of selecting a subset of features may comprise a step of reducing the dimension of the final vector, i.e., of the Ft 45 fingerprint, to a predetermined number of bytes. According to one embodiment, the step of selecting a subset of features from the plurality of features can be based on an artificial intelligence algorithm A0. Preferably, this artificial intelligence algorithm A0 is configured to select a low-dimensional subset of the final high-dimensional vector that best 2088382 of 55 discriminates between authentic and inauthentic visual items. According to one realization, the A0 artificial intelligence algorithm is configured to solve a one-zero trace ratio optimization problem; this method is well known to experts in the field. According to another embodiment, said artificial intelligence algorithm A0 is configured to use a Random Forest algorithm to permute the importance of the feature vectors. To train the A0 artificial intelligence algorithm, training data can be used. This training data can comprise two sets of verification attempts: one set contains authentic verification attempts, where the visual item is authentic, and the other set contains inauthentic verification attempts, where the visual item is not authentic. According to one embodiment, the authentication method 200 may comprise, prior to the step of extracting 230 the plurality of features from a corrected Ict image, a step of calculating a quality score for each Ict image using the CPU1 processing unit 42. Preferably, only if this quality score is greater than a predetermined threshold is the step of extracting 230 the plurality of features from this Ict image executed. This allows considering only Ict images where the quality of the digital representation of the visual item VI 50 is sufficient for the extraction 230 of the features used for generating the digital fingerprint Ft 45. According to one implementation, this quality score can be based on several parameters. For example, these parameters may include one of the following: Ict corrected image brightness, Ict corrected image contrast, Ict corrected image saturation, scan distance between the OPT1 41 optical device and the VI 50 visual item to be authenticated, scan perspective between the OPT1 41 optical device and the VI 50 visual item to be authenticated, OPT1 41 optical unit characteristics, Ict corrected image resolution, etc. This quality score allows for the removal of Ic-corrected images that have, for example, poor resolution or poor quality that would negatively impact the calculation of the D(Ict) distance metrics. At the same time, this quality score allows for weighting, as described below. 2088382 of 55 below in this document, the Ic corrected images based on their quality score to favor Ic corrected images with a better quality score. According to one embodiment, the stored digital fingerprint Fo 11 can take the form of a barcode, QR code, data matrix, number, phrase, watermark, metadata, data stored in memory, data stored in smart card memory, etc. According to one embodiment, the stored digital fingerprint F0 11 can be stored on a server. For example, in this case, the valuable document comprising the visual item VI 50 can contain any type of mark or readable data that allows the authentication device AD ​​40 to download the stored digital fingerprint F0 11 from the server. It should be noted that the authentication method 200 may comprise a step of acquiring the stored fingerprint F0 11, preferably by means of the optical unit OPT1 41, then preferably a step of decoding said stored fingerprint F0 11 by means of the processing unit CPU1 42. These two steps may be executed directly one after the other or indirectly one after the other. According to one embodiment, the stage of acquiring the stored digital fingerprint F0 11 is executed at the same time as the stage of acquiring 210 the plurality of images I 44 of the visual article VI 50. According to one embodiment, the stage of decoding the stored fingerprint F0 11 is executed before the stage of generating the fingerprint Ft 45. According to another embodiment, the AD 40 authentication device may comprise a COM1 communication unit configured to communicate with at least one server and to download at least one stored fingerprint F0 11 from said server. According to one embodiment, the AD 40 authentication device may comprise a COM1 communication unit configured to communicate with at least one smart card where the fingerprint F0 11 is stored. For example, such a smart card may carry such a visual item VI 50. According to one embodiment, the AD 40 authentication device may comprise a COM1 communication unit configured to 2088382 of 55 4 communicate with at least one RFID tag (radio frequency identification tag) where the stored digital fingerprint Fo 11 is stored. For example, this RFID tag can be transported with this visual item VI 50 in the same medium 51. According to one embodiment, the stored digital fingerprint Fo 11 comprises a vector V0. This vector V0 comprises a predetermined number N of elements, such as a predetermined number of bytes, for example. The digital fingerprint Ft 45 comprises a vector Vt. This vector Vt comprises a number M of elements. Preferably, the number M is less than or equal to the number N. Advantageously, the number M is a function of the quality score of the corrected image Ict. This allows generating a digital fingerprint Ft 45 that comprises a series of elements linked to the quality score of the corrected image Ict. In this way, if the quality level is low, then the length of the digital fingerprint Ft 45, and therefore its complexity, is reduced; that is, the digital fingerprint Ft 45 is less sensitive to the details of the corrected image Ict. In the case of a high quality score, then the entire length of the digital fingerprint Ft 45 can be considered.According to one embodiment, for the calculation of the distance metric D(Ict) described below in this document, the stored fingerprint Fo 11 can be truncated in order to consider the same number of elements, i.e., the same length, as the generated fingerprint Ft 45. For example, if the generated fingerprint Ft 45 comprises only 32 bytes, for example, due to the corrected image quality score Ict, then the processing unit CPU1 42 calculates the distance metric D(Ict) based on these 32 bytes and the first 32 bytes of the stored fingerprint F0 11. Figure 11 illustrates an example of fingerprint generation. As stated above, the fingerprint ft 45 can be based on a DCT of a digital representation of a visual item VI 50. These DCTs transform a digital representation of the visual item VI 50 into a two-dimensional array, where the upper left encodes low-frequency information, i.e., the low-detail features of the corrected image Ict, and the lower right encodes high-frequency information, i.e., the high-detail features of the corrected image Ict. According to one realization, the ft 45 fingerprint is constructed from the top left of the DCT coefficient matrix as illustrated 2088382 of 55 in Figure 11. For example, if the processing unit considers l χ l parts χχ χα of this matrix, then the fingerprint Ft 45 has / χ l bits, or — bytes. As illustrated in Figure 11, according to one embodiment, the extraction of elements to generate the vector Vt, i.e., the fingerprint Ft 45, from the matrix is ​​performed based on a nonlinear reading of the matrix. According to Figure 11, block 1 is taken, then block 2, then block 3, and so on to generate the vector Vt as Vt = [block 1, block 2, block 3, block 4, block 5, block 6, ...]. As discussed later in this document, selecting the matrix element in this order allows concentrating the low-frequency part of the matrix at the beginning of the vector Vt.In fact, the Vt vector is not constructed by reading the matrix line by line or column by column, but by selecting a zigzag block from the upper left block of the matrix to the lower right block. Therefore, if the image quality score (Ict) is below a certain threshold, the ft 45 fingerprint comprises a smaller number of elements. These elements are related to the low-frequency information of the corrected image (Ict), and thus to the upper left part of the matrix. Consequently, the Vt vector can be easily truncated by considering only the first elements, i.e., those related to the low-frequency information of the corrected image (Ict). Preferably, based on the quality of the ICT-corrected image, which is determined by its quality score, not all the information contained in this matrix is ​​necessarily used. In fact, for example, if the ICT-corrected image is of very good quality, the entire matrix can be used—for example, a 24x24 matrix—resulting in a generated Ft 45 fingerprint with a length of 72 bytes. However, if the ICT-corrected image is of lower quality, only a portion of the matrix is ​​used, preferably the upper left portion—for example, the upper left of the 24x24 matrix—resulting in a generated Ft 45 fingerprint with a length of 48, 32, or even 16 bytes, which is preferably determined by the quality index. According to one implementation, the quality score can also consider whether the corrected image Ict contains noise such as chirps / dust in the display medium. Advantageously, to calculate the distance metric D(Ict) between the stored fingerprint Fo 11, which has N elements, for example 72 bytes, and the generated fingerprint Ft 45 of the image 2088382 of 55 corrected Ict that has M elements, for example, less than 72 bytes, only the corresponding first M elements of the vector Vo are considered, and only the top left part of the matrix Q is used, as described below in this document. To facilitate this adaptable length, the cells in l times l blocks are arranged in a zigzag pattern as illustrated in Figure 11 to transform the 2D DCT matrix into the 1D vector Vt, i.e., the fingerprint Ft 45. Advantageously, this allows us to know that the elements to the left of the vector Vt encode low-frequency information and the elements to the right encode high-frequency information. Preferably, if the vector Vt has to be cut to a specific length, preferably based on a quality score, the CPU1 42 processing unit only has to cut at the specified byte number from the left, for example. As described below in this document, and due to the method 100 for generating the stored fingerprint Fo 11, the stored fingerprint F0 11 is calculated using a larger part of the array to account for the higher frequencies, i.e., more details of the authentic visual item AVI 10. According to one embodiment, the method uses a 24x24 block, resulting in a vector Vo, i.e., the stored fingerprint Fo 11, of 576 bits, i.e., 72 bytes. As is usually the case, the digital representation of the authentic visual item AVI 10 is a high-quality digital image, i.e., the image I used to generate the stored fingerprint F0 11 is a digital image with the highest quality score, for example. According to one embodiment, after generating the digital fingerprint Ft 45 from a corrected image Ict, the authentication method comprises the step of calculating, using the processing unit CPU1 42, at least one distance metric D(Ict) between said digital fingerprint Ft 45 and said stored digital fingerprint Fo 11. This distance metric D(Ict) can be calculated using different solutions. According to a preferred embodiment, this distance metric D(Ict) is a Mahalanobis distance metric. A Mahalanobis distance metric between two vectors x and y can be written as follows: oQ(x,y) = A [xi-yi, ...,xD-yD] '41.1 41.2 41.0' 42.1 42.2 ··· 42, O : :--.0 .4o,i 4d,2 4d,d. r«i-—yi i 1 r^i — yi] l ... I== [«i-yi>.,«D-yD]Ql ··· IL«D-yoJ JL*o—yoJ Advantageously, in the implementation of the invention, the vector x can 20883822727 of 55 can be replaced by the stored fingerprint Fo 11 and the vector and can be replaced by the generated fingerprint Ft 45. According to one embodiment, the distance metric D(Ict) is preferably calculated using a matrix Q as a mathematical operator between the stored fingerprint Fo 11 and the fingerprint Ft 45. Preferably, the matrix Q is a positive semidefinite matrix. According to one realization, if the matrix Q is an identity matrix, then the distance metric degenerates into a Euclidean distance; if the matrix Q is a diagonal matrix, then the distance degenerates into a weighted Euclidean distance; if the matrix Q is a general matrix, then the distance metric can capture very complex interactions between different feature dimensions and the distance metric is a Mahalanobis distance metric. Advantageously, the present invention comprises a training stage for an artificial intelligence algorithm A3 configured to optimize said matrix Q using at least one dataset SD3 and one dataset SD4. The dataset SD3 preferably contains authentic verification attempts, where the visual item is authentic, and the dataset SD4 preferably contains inauthentic verification attempts, where the visual item is not authentic. This training stage is configured to optimize the matrix Q such that: a. the distance metric D(Ict) between the generated fingerprint Ft 45 and the stored fingerprint F0 11 is as small as possible using said SD3 dataset, i.e., the images in the SD3 dataset corresponding to authentic visual items; and b. The distance metric D(Ict) between the generated fingerprint Ft 45 and the stored fingerprint F0 11 is as large as possible using an SD4 dataset, i.e., the SD4 dataset images corresponding to non-authentic visual items. In fact, the main objective of this A3 artificial intelligence algorithm is to obtain a matrix Q that: a. Maximize the distance metric D(Ict) using said SD3 training data; and b. Minimize the distance metric D(Ict) using said SD4 training data. 2088382 of 55 According to one embodiment, this training stage can use a method called the trace-relationship optimization problem or an improvement of trace-relationship optimization with sparse representation constraints. Using one of these techniques allows this training stage to optimize the Q matrix. According to one embodiment, the step of calculating a first authenticity likelihood function L(Ict|H) of the visual article VI 50 from its digital representation of the corrected image Ict based on said calculated distance metric D(Ict) comprises the following steps: a. adapt a probability distribution or probability density function PD1 of a distance metric D (SD1) from at least one training dataset SD1, said training dataset SD1 corresponding to authentic visual items; and b. For each image Ict, calculate the first likelihood function L(Ict |H) as the probability or probability density given by PD1 to the distance metric D(Ict); According to another embodiment, the first likelihood function L(Ict |H) can be calculated as the probability density given by PD1 to the distance metric D(Ict). According to one embodiment, the first likelihood function L(Ict |H) can be partially generated by an artificial intelligence algorithm A1 using at least the first training dataset SD1. According to this embodiment, the first likelihood function L(Ict |H) can comprise at least one subfunction SF1. Preferably, this subfunction SF1 is defined by a probability density function for the visual item VI 50 to be authentic. Advantageously, this subfunction SF1 can be generated by the artificial intelligence algorithm A1 using the dataset SD1. According to one embodiment, this first subfunction SF1 corresponds to a mathematical model of authentic visual items, i.e., a mathematical model of the probability distribution or probability density function with respect to the authentication of authentic visual items. According to one embodiment, when the visual article VI 50 is transported by a means 12, the present invention can consider the nature of the means 12 in order to optimize its execution, i.e., in order to optimize the computation of the probability P(H). For example, the visual article VI 50 may be a photograph printed on paper or a plastic sheet, or it may 2088382 of 55 can be displayed on a screen, for example. Depending on the nature of the medium 12, the generated fingerprint Ft 45 may be different. The present invention is preferably configured to take into account the nature of the medium 12 in its calculations. According to one embodiment, the first likelihood function L(Ict|H) can be partially generated using an artificial intelligence algorithm A2 configured to generate at least one linear combination of mathematical media models for each corrected image Ict based on at least a plurality of mathematical media models. According to this embodiment, the artificial intelligence algorithm A2 is configured to classify medium 12 carrying visual article VI 50 according to a linear combination of mathematical media models. Such an artificial intelligence algorithm A2 may comprise the following stages: a. extract feature vectors of local binary patterns in pixels among a plurality of pixels of each Ict-corrected image; and b. calculate the histogram of the feature vectors of local binary patterns in at least a portion of each Ict-corrected image; and c. training a classifier that generates the medium from the feature vectors of local binary patterns based on a plurality of mathematical models of the media, preferably the artificial intelligence algorithm has been trained to identify each mathematical model from said plurality of mathematical models of the media; and d. using said classifier, generating at least one probability for each mathematical model of the media from the plurality of mathematical models for each corrected Ict image that the visual item VI 50 is transported by a certain type of media; and e. generate at least one linear combination of mathematical models of said media for the Ict corrected image. According to one embodiment, the first likelihood function L(Ict |H) can be partially generated using an AI algorithm A5. This AI algorithm A5 can be configured to classify the quality of each corrected image Ict from the plurality of corrected images Ic. According to one embodiment, this AI algorithm A5 can use well-known solutions such as a TensorFlow Lite model, via a convolutional neural network (CNN). 2088382 of 55 According to one embodiment, the second likelihood function L(Ict |G) can be partially generated by an artificial intelligence algorithm A1' using at least the training dataset SD2. According to this embodiment, the second likelihood function L(Ict |G) can comprise at least one subfunction SF2. Preferably, this subfunction SF2 is defined by a probability density function for the visual item VI 50 not being authentic. Advantageously, this subfunction SF2 can be generated by the artificial intelligence algorithm A1' using the dataset SD2. According to one embodiment, this second subfunction SF2 corresponds to a mathematical model of non-authentic visual items, i.e., a mathematical model of the probability distribution or probability density function with respect to the authentication of non-authentic visual items. According to one embodiment, the second likelihood function L(Ict|G) can be partially generated using an AI algorithm A2' configured to generate at least one linear combination of mathematical media models for each corrected image Ict based on at least a plurality of mathematical media models. According to this embodiment, the AI ​​algorithm A2' is configured to classify the medium carrying the visual item according to a linear combination of mathematical media models. The AI ​​algorithm A2' can comprise the same stages as the AI ​​algorithm A2. According to one embodiment, the second likelihood function L(Ict|G) can be partially generated using an artificial intelligence algorithm A5'. This artificial intelligence algorithm A5' can be configured to classify the quality of each corrected image Ict from the plurality of corrected images Ic according to the artificial intelligence algorithm A5 described above. Figure 12 illustrates, according to one realization, the first likelihood function L(Ict|H) and the second likelihood function L(Ict|G). In this figure, the distance metrics D(Ict) are plotted on the abscissa and the probability density on the ordinate. As illustrated, the first likelihood function L(Ict|H) covers an area corresponding to the small distance metric D(Ict), while the second likelihood function L(Ict|G) covers an area corresponding to the large distance metric D(Ict). These two likelihood functions The probability distributions 2088382 of 55 have been designed to correspond to the case of an authentic visual item for the first likelihood function L(Ict|H) and a non-authentic visual item for the second likelihood function L(Ict|G). In fact, for an authentic visual item AVI 10, the distance metric D(Ict) has to be small and, therefore, the probability distribution (probability density function) represented by the first likelihood function L(Ict|H) has to cover smaller distance metrics D(Ict) than the second likelihood function L(Ict|G). Conversely, for a non-authentic visual item, the distance metric D(Ict) has to be large and therefore the probability distribution (probability density function) represented by the second likelihood function L(Ict|G) has to cover larger distance metrics D(Ict) than the first likelihood function L(Ict|H). With respect to these probability functions, according to one implementation and as described above, each corrected image Ict will be analyzed by the CPU1 processing unit. Event H corresponds to the detection of a genuine visual item, and event G corresponds to the detection of a counterfeit visual item. Therefore, the probability P(H) is the probability that visual item VI 50 is genuine, and the probability P(G) = 1 - P(H) is the probability that visual item VI 50 is counterfeit. The probability distribution p(P(H)) is the probability distribution of the probability that visual item VI 50 is genuine, and the probability distribution p(P(G)) is the probability distribution of the probability that visual item VI 50 is counterfeit.Based on that, the first likelihood function L(Ict|H) corresponds to the probability of observing the corrected image Ict if the visual item VI 50 is authentic, and the second likelihood function L(Ict|G) corresponds to the probability of observing the corrected image Ict if the visual item VI 50 is not authentic. According to one embodiment, the first likelihood function and / or the second likelihood function can be generated at least partially based on the characteristics of the OPT1 41 optical unit. In fact, for example, based on the nature of the OPT1 41 optical unit, such as its model, the likelihood functions can be trained using an artificial intelligence algorithm. According to this embodiment, for each camera model, a specific first likelihood function and / or a specific second likelihood function can be considered. 2088382 of 55 According to one embodiment, the step of computing the probability P(H) that the visual item VI 50 is authentic using the first likelihood function L(Ict |H) and the second likelihood function L(Ict |G) may comprise the following steps: to. b. c. establish a prior probability Po(H) based on a predetermined set of rules, and for each spatially corrected image Ict, calculate a posterior probability Pt(H) as=£( / ct|H)Po(H)=___________£( / ct|H)Po(H)___________t() PUG) L( / ct|H)Po(H) + L( / ct|6)(1-Po(H)) and, calculate P(H) by computing a weighted moving average as follows: P(H) = ΣΓ=οH where w is the number of corrected images Ict of the plurality of corrected images Ic and {y¿} (i=0,...w) are predetermined weights. According to one embodiment, each predetermined weight applied to each of the Ict-corrected images is a function of the quality score of that Ict-corrected image. Advantageously, each predetermined weight {y¿} (i=0,...w) applied to each of the Ict-corrected images is a function of the quality score of the Ict-corrected image, preferably corresponding to the predetermined weight {y¿} (i=0,...w). Preferably, the higher the quality score, the higher the weight, and the lower the quality score, the lower the weight. This allows more weight to be given to Ict-corrected images with a higher quality score than to those with a lower quality score. According to one embodiment, this prior probability Po(H) is established based on a predetermined set of rules. This predetermined set of rules can be established based on an AI4 algorithm, for example, preferably using at least one decision tree. This AI4 algorithm can be advantageously configured to generate a prior probability Po(H) based on at least one of these parameters: a reputation score based on the nature of the visual item and / or the location of the visual item and / or metadata related to the visual item and / or the sender of the visual item and / or 2088382 of 55 the issuer of a medium that transports the visual article, a uniform distribution law, etc. For example, and as described later in relation to the implementation of the identification card, the above probability can be set according to the card issuer, preferably using historical data, for example for each issuer and / or country, the above probability can be set at the percentage of authentic identification cards issued by this issuer and / or country over a predetermined number of times, such as a predetermined number of past years. According to one implementation, the A4 artificial intelligence algorithm uses a decision tree process and / or a forest process consisting of a decision tree process to generate the previous probability P0(H). For example, using historical data, a decision tree model can be trained to generate the previous probability for a given issuer and / or a given country, which was not present in the historical data. According to one embodiment, the probability P(H) can be computed from a probability distribution p(P(H)) as described below. In this case, the probability P(H) is related to at least one descriptive statistic of that probability distribution p(P(H)). Preferably, the descriptive statistics of that probability distribution p(P(H)) may include: the mean, median, mode, range, IQR (interquartile range), variance, standard deviation, a surface, a moment, etc. According to another embodiment, the step of computing a probability P(H) that the visual item VI 50 is authentic using the first likelihood function L(Ict |H) and the second likelihood function L(Ict |G) may comprise the following steps: a. establish, using the CPU1 42 processing unit, a previous probability distribution p0(P(H)) based on a predetermined set of rules, and b. sample, using the CPU1 42 processing unit, K independent values ​​of previous probability {Po,i,...,P(o, k)} from the previous probability distribution p0 (P(H)); and c. Using each of these K previous probabilities {P(o,k)}(k=i...K), calculate, using the CPU1 processing unit (42), K probabilities 2088382 of 55 d. and. later {P <t,k)(H)}<k=i.K> : L(IctlH)Po(^) ___________L(Jct\H)Po(H)____________t¡k() P(jct) L( / ct|H)Po(H) + L( / ct|6)(1-Po(H)) use the K posterior probabilities {P(t,k)(H)} <k=i.k), adaptando, mediante la unidad de procesamiento cpu1 42, una distribución probabilidad posterior pt(p(h)), y calcular, (42), p(p(h)) computando un promedio móvil ponderado siguiente manera:p(P(H)) = ΣΓ=ο«ίΡί-ί(Ρ(^)) ΣT=oαι weight where w is the number of spatially corrected images Ict of the plurality of spatially corrected images Ic and {az} (i=0,...w) are predetermined weights. As described above, according to one implementation, each default applied to each of the Ict corrected images is a function of the quality score of that Ict corrected image. Advantageously, each predetermined weight {a¿} <i=0,...w) aplicado a cada una de las imágenes corregidas ict es función la puntuación calidad imagen corregida ict, preferentemente correspondiente al peso predeterminado {az} <i="0,...w)." preferentemente, cuanto mayor calidad, el y menor predeterminado. esto permite dar más crédito ic que tienen mejor peor calidad.According to one realization, the predetermined set of rules used to establish the above probability distribution p0(P(H)) can be based on the A4 artificial intelligence algorithm as discussed above. According to one embodiment, the authentication of the VI 50 visual item is confirmed if P(H) is greater than a predetermined threshold. For example, if the probability of authenticity P(H) is greater than 75%, preferably 85%, and advantageously 95%, then the VI 50 visual item is considered authentic. According to one embodiment, this predetermined threshold can be defined according to the use case of the present invention. Preferably, this predetermined threshold determines the balance between 2088382 out of 55 possible false matches and false non-matches, i.e., between a false confirmation of authenticity and a false confirmation of non-authenticity. For example, in use cases where a false match is a more serious problem than a false non-match, such as in a high-security facility, a stricter threshold, i.e., a higher threshold value, can be used. As another example, in use cases where a non-false match is a more serious problem than a false match, such as in a low-security facility, a less strict threshold, i.e., a lower threshold value, can be used. According to one embodiment, such a default threshold can be determined using a mathematical relationship between the false match rate, the false non-match rate, and the default threshold value. For example, Figure 13 illustrates such a mathematical relationship.Figure 13 illustrates a graph showing the expected false match rate and the expected false non-match rate based on the threshold. This mathematical relationship is advantageously calculated using the SD1 and SD2 datasets described above, preferably using an AI algorithm A6 that utilizes these SD1 and SD2 datasets. According to one implementation, the default threshold can be found by consulting the acceptable false match rate and the acceptable false non-match rate. According to one implementation, this predetermined threshold can be based on the predetermined set of rules. According to one realization, the probability P(H) is equal to at least one descriptive statistic of that probability distribution p(P(H)). P(H) is advantageously the expected value of p(P(H)), such that: P(H) = r1 xx J0p(y) ydy According to one embodiment, the present invention relates to a computer program comprising instructions that, when the program is executed by the CPU1 42 processing unit, causes the CPU1 42 processing unit to perform authentication method 200, i.e., the steps of authentication method 200. Preferably, the CPU1 42 processing unit is configured to control the OPT1 41 optical unit. According to one embodiment, the present invention relates to a computer-readable storage medium comprising instructions 2088382 of 55 that, when the program is executed by the CPU1 42 processing unit, causes the CPU1 42 processing unit to perform authentication method 200, i.e., the stages of authentication method 200. According to one embodiment, the authentication process can be implemented by a smartphone through an application, for example, which can be downloaded from an app store, for example. The present invention makes it possible to determine whether a visual item, preferably of any type, is authentic or not based on a stored digital fingerprint, preferably a certified stored digital fingerprint, generated by an issuer, preferably a certified issuer. The use of multiple images allows a user to determine whether a visual item is authentic or not without the need for a dedicated reader, but simply by using an application on their smartphone. The present invention cleverly uses several techniques combined in an advantageous way that allow for a high level of accuracy in assessing whether a visual item is authentic or not. The use of artificial intelligence allows us to overcome several technical problems such as the optical conditions for acquiring multiple images, the nature of the medium that carries the visual item, and the nature of the visual item itself, for example. The authentication device According to one embodiment, the present invention relates to an AD 40 authentication device configured to execute the authentication method 200 described above in this document. According to a preferred embodiment, and as described in Figures 1 to 3, said authentication device AD ​​40 is configured to authenticate a visual item VI 50 using a stored fingerprint Fo 11 associated with said visual item VI 50. Preferably, said authentication device AD ​​40 comprises: a. an OPT1 41 optical unit configured to acquire at least a plurality of images I 44 of an area comprising the visual article VI 50 to be authenticated, preferably each image It of the plurality of images I 44 comprising at least partially a digital representation of said visual article VI 50; and b. a CPU1 42 processing unit configured for: 2088382 of 55 i. for each image It, generate 220 corrected images Ict by preferably correcting the image It based on at least one spatial feature, creating the plurality of corrected images Ic; and ii. for each corrected image Ict: or extract 230 a plurality of features; I generate 240 a digital fingerprint Ft 45 of the digital representation of the visual item VI 50 using at least a part of said plurality of extracted features; I compute 250 at least one distance metric D(Ict) between said fingerprint Ft 45 and said stored fingerprint F0 11; I compute 260 the first likelihood function L(Ict |H) and the second likelihood function L(Ict |G) as described above; I compute 270 the probability P(H) that the visual item VI (50) is authentic using the first likelihood function L(Ict |H) and the second likelihood function L(Ict |G); and iii. confirm 280 that the visual item VI 50 is authentic if the P(H) is greater than a predetermined threshold. c. advantageously, a COM1 communication unit configured to download the stored digital fingerprint F0 11 if necessary and / or to send a message as to whether the visual item VI 50 is authentic or not; and d. advantageously, a display unit configured to display at least confirmation that the visual item VI 50 is authentic and / or to display the corrected Ict image during the execution of authentication method 200; and e. preferably, a power unit configured to power the AD 40 authentication device. According to one embodiment, the AD 40 authentication device is a smartphone and / or a laptop and / or a tablet. As described in Figures 1 to 3, and according to one embodiment, the user of the authentication device AD ​​40, for example, a smartphone running an application designed to execute authentication method 200, takes a plurality of photographs, preferably a video, advantageously in real time, of a visual item VI 50 to be authenticated, for example, an identification card. In this example, the phone's camera The 2088382 of the 55 smart device, i.e., the OPT1 41 optical unit, acquires a plurality of ID card images, including the ID card owner's photograph, as well as the stored fingerprint F0 11 encoded in the form of a barcode, for example. Then, the smartphone's CPU1 41 processing unit, i.e., the AD 40 authentication device, executes the steps of authentication method 200 in order to generate a fingerprint Ft 45 for each corrected Ict image, and compare it with the stored fingerprint F0 11.According to one embodiment, the user's smartphone can display what the optical device OPT1 41, i.e., the smartphone's camera, is seeing, preferably in real time, allowing the user to move the smartphone relative to the visual item in order to optimize the acquisition of multiple images, i.e., to increase the quality score of the corrected images Ic, as well as to optimize the relative position of the smartphone with respect to the visual item VI 50, i.e., the ID card. Then, preferably, the processing unit CPU1 42 uses the smartphone's display unit to notify the user whether the photograph located on the ID card corresponds to the photograph used by the official issuer of said ID card to generate the stored digital signature Fo 11. According to one embodiment, the display unit can display the corrected image Ict and / or its fingerprint Ft 45 and / or the image of the stored fingerprint Fo 11, for example, the barcode that encodes said stored fingerprint F0 11, and / or the decoding of the stored fingerprint F0 11 and / or the probability P(H). According to one embodiment, the display unit of the AD 40 authentication device can display information, preferably in real time, about image quality, and / or messages to the user to move the AD 40 authentication device relative to the VI 50 visual item in order to increase the quality score of the corrected images Ic. For example, these messages may comprise at least one of the following: a word, a phrase, a drawing, a figure, a symbol, etc. According to one embodiment, said CPU1 42 processing unit comprises at least one processor configured to execute at least one set of instructions, preferably stored in memory. Said 2088382 of 55 memory is preferably a non-transient memory. Such memory advantageously stores a computer program as described above. Method of generating fingerprints According to one embodiment, the present invention relates to a fingerprint generation method 100 configured to generate a digital signature F0 11 of an authentic visual item AVI 10, preferably using a fingerprint generation device GD 20 disclosed herein, said fingerprint generation device GD 21 comprising at least one optical unit OPT2 21, one processing unit CPU2 22, and preferably one storage unit SU2. Advantageously, said fingerprint Fo 11 is configured to be stored, this storage being able to use different forms as described above. According to one embodiment, the fingerprint generation method 100 uses several features similar to the authentication method 200, advantageously with respect to the generation of the fingerprint Ft 45. As will be explained, according to one embodiment, because the stored fingerprint F0 11 is generated from a visual item 10 considered authentic by the issuer, only one image I is sufficient to generate the stored fingerprint F0 11, whereas in the case of the fingerprint Ft 45, the visual item VI 50 may not be authentic; therefore, several images must be taken and considered to assess a probability P(H) that the visual item VI 50 is authentic. According to one embodiment, and as described in Figures 4 to 6, said digital signature generating method 100 comprises the following stages: a. acquiring 110, preferably by means of the OPT2 optical unit 21, at least one image I of the authentic visual article AVI 10, preferably said image I being a high-resolution image, i.e., its resolution is preferably greater than 300 dpi, advantageously greater than 720 dpi; according to one embodiment, said acquisition step may comprise a step of downloading said image I in the form of a digital file; advantageously, the image may be downloaded from a server and / or received by email; and b. Preferably, generate 120, using the processing unit 2088382 of 55 CPU2 22, a corrected image Ic, preferably a spatially corrected image Ic, correcting image I based on at least one spatial feature, preferably according to the process described above; and c. extract 130, by means of the CPU2 22 processing unit, a plurality of features from said image I and / or from said corrected image Ic, preferably using the methods described above; and d. generate 140, by means of the CPU2 22 processing unit, the digital fingerprint f0 11 of the digital representation of the authentic visual article AVI 10 of image I and / or the corrected image Ic, using at least a part of said plurality of extracted features; and e. Preferably, storing the digital fingerprint f0 11 in at least one of the following forms using the storage unit SU2: a QR code, a data matrix, a barcode, a serial number, a watermark, a digital watermark, metadata, data stored in memory, etc. According to one embodiment, said storage unit can be a computer, a server, a printer, a display, etc. Preferably, said storage unit SU2 is configured to store the digital fingerprint F0 11 in a location accessible to the authentication device AD ​​40, preferably for retrieving and / or downloading it. Advantageously, the storage unit SU2 is a printer, and the digital fingerprint F0 11 is printed as a QR code on the same medium as the authentic visual item.For example, with respect to the use case of the ID card, the owner's photograph can be the authentic visual item and the stored fingerprint F0 11 can be printed near the authentic visual item; it should be noted that the visual item, i.e., the ID card owner's photograph, for example, is considered authentic by the issuer when the security document or medium comprising the visual item and its stored fingerprint F0 11 is issued. As described above, the optical unit OPT2 21 is configured to acquire at least one image I of the authentic visual item AVI 10. According to one embodiment, the optical conditions of this. 2088382 of 55 acquisition allow the acquisition of a single image I of the authentic visual article AVI 10, preferably a high-resolution image I. Advantageously, said optical unit OPT2 21 is a scanner. According to one embodiment, at least one image I of the authentic visual article AVI 10 can be acquired from at least one digital file, for example, by downloading at least one digital file. According to this embodiment, the optical unit OPT2 21 may comprise a module configured to download at least one image I of at least one authentic visual article AVI 10. According to this embodiment, the optical unit OPT2 21 may comprise a module configured to download at least one image of at least one authentic visual article AVI 10, preferably using a QR code as a link to download said at least one image of said at least one authentic visual article AVI 10. As described above, the CPU2 22 processing unit is configured to: a. preferably generating a corrected image Ic, preferably if necessary, by correcting image I, advantageously based on at least one spatial feature; according to one embodiment, the acquisition stage of image I may require correction of said image due to misalignment, for example, and / or due to brightness, contrast, saturation conditions; and b. extracting 130 a plurality of features from said image I and / or from said corrected image Ic, preferably said spatially corrected image Ic; said extraction step 130 may use the same steps as the extraction step 230 described above; and c. generating the digital fingerprint Fo 11 of the digital representation of the authentic visual article AVI 10 from the corrected image Ic using at least a part of said plurality of extracted features; preferably, said step of generating the digital fingerprint Fo 11 comprises steps similar to the step of generating the digital fingerprint Ft 45 described above. According to one embodiment, the present invention relates to a computer program comprising instructions that, when the program is executed by the CPU2 22 processing unit, cause the CPU2 22 processing unit to perform the fingerprint generation method 2088382 of 55 100, that is, the stages of the fingerprint generation method 100. Preferably, the CPU2 22 processing unit is configured to control the OPT2 21 optical unit. According to one embodiment, the present invention relates to a computer-readable storage medium comprising instructions that, when the program is executed by the CPU2 22 processing unit, cause the CPU2 22 processing unit to perform the fingerprint generation method 100, i.e., the steps of the fingerprint generation method 100. The present invention allows for the easy generation of a digital fingerprint of a visual item, and this fingerprint can then be used to compare it with another digital fingerprint generated from a visual item. If these two fingerprints are sufficiently close to each other, then these visual items are in fact the same visual item; that is, the visual item is authentic. Fingerprint generating device According to one embodiment, the present invention relates to a fingerprint generating device GD 20. Said fingerprint generating device GD 20 is configured to execute the fingerprint generation method 100 described above. Said fingerprint generating device GD 20 is advantageously configured to generate the digital signature F0 11 from an authentic visual item AVI 10, and preferably to store said fingerprint F0 11, advantageously in at least one of the forms discussed above. According to one embodiment, the GD 20 digital signature generating device comprises: a. an OPT2 21 optical unit configured to acquire at least one image I of the authentic visual article AVI 10 and / or a download unit configured to download at least one image I of the authentic visual article AVI 10 from a server; according to one embodiment, said OPT2 21 optical unit may comprise a module configured to download at least one image I of the authentic visual article AVI 10; and b. a CPU2 22 processing unit configured for: i. preferably generate 120 an Ic corrected image, preferably a spatially corrected Ic image by correcting the 2088382 of 55 image I based on at least one spatial feature; and ii. extract 130 a plurality of features from said image I and / or from said corrected image Ic; and iii. generate 140 the digital fingerprint Fo 11 of the digital representation of the authentic visual article VI 10 using at least a part of said plurality of extracted features; and c. preferably, an SU2 storage unit configured to store 150 F0 10 digital fingerprints in at least one of the following forms: a QR code, a data matrix, a barcode, a serial number, a digital watermark, metadata, data stored in a memory, for example, a smart card, etc.; and d. Preferably, a COM2 communication unit configured to send and / or receive data over at least one communication network; said COM2 communication unit is advantageously configured to send the stored digital fingerprint F0 11 to a server, preferably a secure server that allows an AD 40 authentication device to download it if required; and e. advantageously, a display unit configured to display at least image I of the authentic visual article AVI 10 and / or to display fingerprint F0 11 and / or the stored fingerprint F0 11, i.e., fingerprint F0 11 in its stored form; and f. preferably, a power unit configured to power the GD 20 fingerprint generation device. According to one embodiment, the GD 20 fingerprint generation device can be a computer connected to a scanner. In another embodiment, it can be a mobile device such as a smartphone, for example. Preferably, this fingerprint generation method 100 has several stages in common with the authentication method 200. In particular, the stages that differ are based primarily on the fact that, in the case of generating the stored fingerprint F0 11, the visual item in question is authentic, and preferably the optical conditions for acquiring an image of this authentic visual item allow for a high-resolution digital representation of said authentic visual item, i.e., an image I with 2088382 out of 55 the highest possible quality score, whereas in the case of generating the Ft 45 fingerprint, the visual item considered VI 50 may not be authentic, and the conditions for acquiring the images may not be perfect, resulting in a more advanced process for generating the Ft 45 fingerprint that allows compensating for these conditions. According to one embodiment, said CPU1 42 processing unit comprises at least one processor configured to execute at least one set of instructions, preferably stored in memory. Said memory is preferably non-transient memory. Said memory advantageously stores a computer program as described above. First example: Identification card, also called ID card According to a first example, and as illustrated in Figure 7, the present invention can be implemented to allow a user to authenticate a person's 10b identification card. According to one embodiment, this 10b identification card is issued by a government agency. To obtain this 10b identification card, the prospective owner must provide the official agency with a photograph of themselves. The issuer of the 10b identification card deems this photograph to be authentic. This photograph may be the visual element carried by the 10b identification card. According to one embodiment, this photograph of the prospective owner may be a digital file, preferably downloaded or received digitally by the official agency. The issuer uses fingerprint generation method 100 to generate the fingerprint F0 11 associated with the photograph. This fingerprint F0 11 is then preferably stored in printed form on the identification card, for example, using a barcode, a data matrix, a QR code, and / or a watermark, digital watermark, metadata, or data stored in memory, etc. According to one embodiment, a watermark may comprise steganographic data, which preferably encodes the stored fingerprint F0 11. According to one embodiment, to use a steganographic process to store the F0 11 fingerprint within at least a part of an article 2088382 of 55 visual, a predetermined area of ​​said visual item may be chosen to store said fingerprint Fo 11 in a steganographic form. Preferably, in this case, said stored fingerprint F0 11 is generated using an area other than the predetermined area in order to avoid any disturbance due to the addition of the stored fingerprint Fo 11 in its steganographic form within the visual item. Therefore, the CPU1 42 processing unit may be configured to avoid considering said predetermined area to calculate a fingerprint Ft 45. Preferably, said other area is considered by the CPU1 42 processing unit to generate said fingerprint Ft 45, and said predetermined area is used to extract said stored fingerprint F0 11.Advantageously, a watermark can be used in a predetermined area configured to be avoided by the CPU1 42 processing unit to generate such fingerprint Ft 45 and considered by the CPU1 42 processing unit to extract such stored fingerprint F0 11. For example, such predetermined area can be at least a part of the edge, or outline, of the visual item. According to another embodiment, the CPU1 42 processing unit can be configured to use an artificial intelligence algorithm A7 trained to identify steganographic data from a visual item. Preferably, this artificial intelligence algorithm A7 can be trained using a dataset comprising a plurality of visual items, a plurality of stored fingerprints F0, each stored fingerprint F0 of said plurality of stored fingerprints Fo being associated with a visual item of said plurality of visual items and a plurality of visual items comprising steganographic data. Each visual item comprising steganographic data corresponds to a visual item of the plurality of visual items, and each of these steganographic data encodes a stored fingerprint Fo of the plurality of stored fingerprints Fo.Using this dataset, the A7 artificial intelligence algorithm is advantageously capable of extracting steganographic data from a given visual item and generating an Ft fingerprint from it, preferably without considering the steganographic data in the generation of the Ft fingerprint. The A7 artificial intelligence algorithm is preferably capable of generating an Ft fingerprint from a visual item that includes its fingerprint. 2088382 of 55 stored Fo, encoding in steganographic data, said generated fingerprint Ft which is configured to correspond to said stored fingerprint Fo, that is, to said stored fingerprint Fo steganographically encoded. Therefore, the issuer uses a fingerprint generation device GD 20, such as a computer and a scanner, the computer being the CPU2 22 processing unit and the scanner being the OPT2 21 optical unit. The issuer uses the scanner to acquire at least one image I of the photograph of the future owner of the identification card. The CPU2 22 processing unit extracts at least a plurality of features from this image I based on the method discussed above. These features, in vector form, are then used to generate a fingerprint F0 11 associated with the photograph of the future owner, preferably intrinsically linked to that photograph. In fact, any modification of the photograph will generate a different fingerprint Fo 11.Next, the identity card is printed and carries said image I as well as said fingerprint Fo 11 which is now printed, i.e., stored on the identity card, for example, in the form of a QR code. Next, if a user wants to authenticate the ID card owner's photograph, they can use an AD 40 authentication device as described above, such as their smartphone, for example, using a dedicated application. Using their smartphone, the user captures multiple images of the photograph on the ID card; preferably, the user also captures a real-time video of this photograph. According to one embodiment, for each frame captured, or at least for some of the frames captured from this video, the authentication method 200 is applied to correct the I images, generating multiple Ic images. Advantageously, this correction stage is very useful.In fact, for the generation of the stored digital fingerprint Fo 11, the owner's photograph was fixed and perfectly aligned with the lenses of the OPT2 21 optical unit; in this case, this photograph was scanned using a scanner. Meanwhile, when the user is acquiring this plurality of images I, the position of the AD 40 authentication device, their smartphone, with respect to the position of the ID card cannot be perfect; there are some perspective issues that need to be corrected. 2088382 of 55 example. Therefore, each image It is spatially corrected so that the digital representation of the visual item VI 50, i.e., the photograph of the ID card holder, is on a virtual plane parallel to the plane of the lenses of the optical unit OPT1 41, the user's smartphone camera. Then, for each corrected image Ict, preferably spatially corrected images Ict, the smartphone's processing unit CPU1 42 extracts some features and generates a digital fingerprint Ft 45 associated with the image considered Ict. Based on this generated fingerprint Ft 45, the CPU1 42 processing unit, i.e., the user's smartphone, estimates a distance metric D(Ict) between this generated fingerprint Ft 45 and the stored fingerprint Fo 11. In fact, before or during the acquisition of the It images, the OPT1 41 optical unit acquires at least one image of the stored fingerprint Fo 11 and the CPU1 42 processing unit, if necessary, decodes the stored fingerprint Fo 11 from this Ict image. Then, as described above, a first likelihood function L(Ict|H) and a second likelihood function L(Ict|G) are calculated. The first likelihood function L(Ict|H) relates to the case that the photograph, i.e., visual item VI 50, is authentic, and the second likelihood function L(Ict|G) relates to the event that the photograph, i.e., visual item VI 50, is not authentic. Next, as described above, using these two likelihood functions, a probability P(H) is calculated that the visual article VI 50 illustrated in the image Ict is authentic based on a prior probability P0(H) and / or a prior probability distribution p0(P(H). Advantageously, for each new corrected Ict image considered, the probability P(H) is updated. Preferably, each Ict image does not have the same weight, i.e., the same impact, with respect to updating the probability P(H). In fact, as described above, and according to a preferred embodiment, depending on the quality score of each Ict-corrected image, its weight is not the same in calculating the average probability P(H). This allows for greater consideration of Ict-corrected images with a higher quality score, for example, one higher than a certain threshold. 2088382 out of 55 default, and have low consideration for Ict corrected images with a lower quality score, for example, lower than a default threshold. Ideally, in real time, the user can see on their smartphone screen confirmation or non-confirmation as to whether the photograph on the ID card is authentic or not. According to one embodiment, authentication method 200 may comprise a step of guiding the user to move the authentication device AD ​​40 in space in order to improve the quality of the I 44 images acquired by the optical unit OPT1 41. For example, the authentication device AD ​​40 may notify the user to move the optical unit OPT1 41 closer to the visual item VI 50 and / or to move the optical unit OPT1 41 farther away from the visual item VI 50. As another example, the authentication device AD ​​40 may notify the user to rotate the optical unit OPT1 41 relative to the visual item VI 50 to reduce or avoid, for example, any misalignment, defect, or perspective error. As yet another example, the authentication device AD ​​40 may notify the user that the lighting conditions are insufficient to properly capture the I 44 images of the visual item VI 50. According to one realization, after a predetermined number of Ict-corrected images and / or after a predetermined image acquisition time t: a. If the probability P(H) is less than a predetermined threshold, the processing unit notifies the user that the visual item VI 50 is not authentic; and b. If the probability P(H) is greater than a predetermined threshold, the processing unit confirms to the user that the visual item VI 50 is authentic. Second example: valuable document According to one embodiment, and as illustrated in figure 8, the visual item VI 50 can be transported by a document of value 10a such as a diploma for example. According to this use case, the visual item VI 50 can be a part or even the entire value document, depending on the use case. Regarding the identification card, the document of value is considered 2088382 of 55 as authentic by the issuer of said valuable document, such as a diploma, for example. Said issuer scans and / or downloads it from a server in digital form and generates a digital fingerprint Fo 11 of it or at least a part of it. This digital fingerprint can be printed F0 11 in a dedicated location on the diploma, for example, on the border or the back, or even in a region designed to be avoided by the authentication device AD ​​40 for the generation 240 of the digital fingerprint Ft 45, for example. According to one embodiment, the digital fingerprint Fo 11 can be stored on a server, and the diploma can include a QR code that allows, for example, the authentication device AD ​​40 to download this digital fingerprint F0 11 from said server during authentication method 200. This allows, for example, the entire valuable document to be considered for generating the digital fingerprint Ft 45. As before, authentication method 200 comprises acquiring a plurality of It images, generating an Ft 45 digital fingerprint for each of the corrected Ict images, preferably if their quality score is higher than a predetermined threshold, and notifying the user whether the visual item VI 50, i.e., the document of value in this case, is authentic or not. Third example: Painting According to a third example, and as illustrated in Figure 9, the present invention can be implemented in the field of art, for example, for the authentication of paintings. According to one embodiment, a painting 10c or at least a part of a painting may be a visual article VI 50 with respect to the present invention. Therefore, an authority can generate a stored digital fingerprint (F0 11) of a painting, or at least of a part of a painting. This stored digital fingerprint (F0 11) can then be stored on a server or located near the painting, on its outline, edge, frame, or even on the back of the painting. According to one embodiment, the OPT2 21 optical unit of the GD 20 fingerprint generation device can be a three-dimensional scanner and / or a camera configured to preferably acquire the relief of the paint, as well as its image. According to one embodiment, the stored fingerprint F0 11 can then be generated using transforms of 2088382 of 55 discrete three-dimensional cosines to extract a plurality of feature vectors forming the stored fingerprint Fo 11 as described above. According to one embodiment, the authentication method can be executed by an authentication device, as described above, wherein the OPT1 41 optical unit is a three-dimensional scanner and / or camera, and / or wherein the OPT1 41 optical unit is a two-dimensional scanner and / or camera. Fourth example: Trading cards According to a fourth example, and as illustrated in Figure 10, the present invention can be implemented in the fields of collectible items, such as, for example, trading cards. According to this definition, it should be noted that a visual item can be more than just a photograph; it can be a text and / or a combination of both. In fact, a visual item is simply an optical element that allows an optical unit to capture at least one photograph of it. Therefore, according to that example, the visual item can be a 10d collectible card, or at least a part of it that allows the rest of the card to carry the stored fingerprint F0 11 in printed form, for example. According to this example, the issuer of the 10d card acquires an image of it using a scanner, for example, or by downloading it from a server, for example, and then generates an F0 11 digital fingerprint that is stored on the card and / or on a server, as described above. The user who wants to authenticate a collectible card then uses his smartphone, for example, as an authentication device AD ​​40, and acquires a plurality of images I of the collectible card, preferably through a real-time video, and the processing unit CPU1 42 of the smartphone generates for each of the considered images Ict a digital fingerprint Ft 45 which is compared with the stored digital fingerprint Fo 11 using the distance metric D(Ict), this allows calculating the first and second likelihood functions which update a probability P(H) that the collectible card is authentic. Fifth example: NFT Authenticity According to a fifth example, the present invention may 2088382 of 55 can be implemented in the field of non-fungal tokens, also called NFTs, with respect to the domain of art, for example. In fact, according to this example, if the NFT is a photograph or even a three-dimensional object, the present invention can be applied to generate a stored digital fingerprint F 11 of an authentic visual item represented by an NFT. The present invention can then be implemented to authenticate whether the visual item represented by an NFT is authentic or not. In fact, based on the NFT process, an NFT is always authentic, but the visual item, 2D or 3D, that is represented by said NFT may not be authentic. An NFT is simply a smart contract associated with a digital item. This digital item can be a visual item such as a photograph, a video, or even a three-dimensional object. The present invention can be used to generate the stored digital fingerprint F0 11 of a visual item represented by an NFT to allow a user to authenticate that said NFT is related to a genuine visual item. In this case, the medium carrying said visual item can be a digital display, a smartphone screen, a computer screen, a tablet screen, etc. According to one embodiment, the stored fingerprint Fo can be stored in the metadata of an NFT. Typically, an NFT comprises metadata, for example, a link to a server where the visual item represented by the NFT is stored. Having the stored fingerprint Fo in the metadata of an NFT allows a user to authenticate the visual item represented by an NFT without needing to access the authentic visual item stored on a server. According to this embodiment, the stored fingerprint Fo can be used as a compressed version of the visual item represented by the NFT, for example, when the link to the server where the visual item represented by the NFT is stored is no longer active and / or live; that is, it is dead. In this case, the user can again use their smartphone as an AD 40 authentication device to run authentication method 200 to determine whether the displayed visual item is authentic or not. Sixth example: video authenticity According to another example, the present invention can be applied to a 2088382 of 55 video. In fact, a video comprises a plurality of frames, each of these frames may be or may comprise a visual item, therefore, the present invention may be used to authenticate a video, or at least a part of a video. For example, each frame of the video can be authenticated. According to another embodiment, only some of the video frames are used in accordance with the present invention to authenticate the entire video. According to another embodiment, a stored digital fingerprint is generated for the entire video based on a plurality of sub-stored digital fingerprints with respect to each or some of the frames of the video. According to one embodiment, authentication method 200 can be applied to a plurality of frames of a video. Therefore, each frame considered can allow the authentication device to compute the probability P(H) that the video is authentic. As described above, the present invention can be applied to a wide range of use cases. These described examples are not a limitation of the present invention. The invention is not limited to the embodiments described above and extends to all embodiments covered by the claims. References: Authentic visual article 10a Valuable document 10b Identification card 10c Paint 10d Collectible Card Stored digital footprint Medium that transports the authentic visual item Digital signature generating device Optical unit of the recording device Recording device processing unit Recording device screen Image of the authentic visual article Digital footprint generated Stored digital footprint Authentic visual item with its stored digital fingerprint 2088382 of 55 Authentication device Optical drive of the authentication device Authentication device processing unit Authentication device screen Plurality of images in the visual article Digital footprint generation Extracted fingerprint Visual item to be authenticated Medium that transports the visual item to be authenticated Registration method Acquire at least one image of the authentic visual item. Generate a spatially corrected image. Extract a plurality of features. Generate a fingerprint. Fingerprint storage Authentication method Acquire a plurality of images of a visual item Generation of a spatially corrected image Extraction of a plurality of features Generating a digital footprint Calculation of at least one distance metric D Calculation of a first likelihood function L(Ict|H) of authenticity H of the visual item VI and a second likelihood function L(Ict|G) of non-authenticity G of the visual item VI. Computation of a probability P(H) that the visual item VI is authentic. Confirmation that visual item VI is authentic 2088382 of 55 CLARKE MODET & CO. (ARGENTINA) SA - 30540437455 Digitally signed by PORTALTRAMITES - INPI Date: 2022.12.19 10:31:59 -03:00 Reason: Digitally Signed by the INPI Location: Buenos Aires, Argentina 2088382

Claims

1. A method for authenticating (200) a visual article VI (50) using a stored fingerprint F0 (11) and an authentication device AD ​​(40), said authentication device AD ​​(40) comprising at least one optical unit OPT1 (41) and at least one processing unit CPU1 (42), said stored fingerprint F0 (11) being previously generated from an authentic visual article AVI (10), said method (200) comprising the following steps: a) acquiring (210), using said optical unit OPT1 (41), at least a plurality of images I (44) of an area comprising the visual article VI (50) to be authenticated, said optical unit OPT1 (41) being in communication with said processing unit CPU1 (42), each image It of the plurality of images I (44) comprising at least partially a digital representation of said visual article VI (50), the image It being acquired in a time t;characterized in that the method (20) further comprises the following steps: b) for each image It of the plurality of images I (44), generating (220), by means of said processing unit CPU1 (42), a spatially corrected image Ict by correcting the image It on the basis of at least one spatial feature to calibrate at least partially a digital representation of said visual article VI (50) to be in the same perspective as that of said authentic visual article AVI (10), creating a plurality of spatially corrected images Ic; and c) for each spatially corrected image Ict of the plurality of spatially corrected images Ic: i) extracting (230), by means of said processing unit CPU1 (42), a plurality of features;and ii) generating (240), by means of said CPU1 processing unit (42), a digital fingerprint Ft (45) of the digital representation of the visual article VI (50) from the spatially corrected image Ict using at least a part of said plurality of extracted features; and iii) calculating (250), by means of said CPU1 processing unit (42), at least one distance metric D(Ict) between said fingerprint Ft (45) and said stored fingerprint F0 (11);and iv) calculating (260), using said processing unit CPU1 (42), a first likelihood function L(Ict|H) of authenticity H of the visual article VI (50) from its digital representation of the spatially corrected image Ict based on said calculated distance metric D(Ict) and a second likelihood function L(Ict|G) of non-authenticity G of the visual article VI (50) from its digital representation of the spatially corrected image Ict based on said calculated distance metric D(Ict); and v) computing (270), using said processing unit CPU1 (42), a probability P(H) that the visual article VI (50) is authentic using the first likelihood function L(Ict|H) and the second likelihood function L(Ict|G);and wherein the image It of the plurality of images I (44) is acquired during the processing of the preceding image It-1 by the processing unit CPU1 (42), said processing comprising performing steps b), c)i), c)ii), c)iii), c)iv) and c)v) on the preceding image It-1; and wherein the probability P(H) is updated as each spatially corrected image Ict of the plurality of spatially corrected images is processed and the authentication of the visual article VI (50) is confirmed by said processing unit CPU1 (42) if the probability P(H) is greater than a predetermined threshold. Fourteen claims follow;