# US 11,823,521 B2 · Image processing method for an identity document > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,823,521 B2 - Document ID: US11823521B2 - Status: Granted - Kind: B2 - Application number: 17/604,962 - Filed: 2020-04-20 - Published: 2022-08-18 - Granted: 2023-11-21 - Earliest claimed: 2019-04-19 - Assignees: Surys - Applicants: Surys - Inventors: Marc-Michel Pic; Mohammed Amine Ouddan; Gaël Mahfoudi; Xavier Pic - Web page: https://patentgrep.com/patent/US11823521B2 ## CPC - G07D 7/00: Testing specially adapted to determine the identity or genuineness of valuable papers or for segregating those which are unacceptable, e.g. banknotes that are alien to a currency - G06F 18/2414: {Smoothing the distance, e.g. radial basis function networks [RBFN]} - G06V 10/764: using classification, e.g. of video objects - G06V 10/7747: {Organisation of the process, e.g. bagging or boosting} - G06V 10/7796: {based on specific statistical tests} - G06V 10/817: {by voting} - G06V 10/82: using neural networks - G06V 30/194: References adjustable by an adaptive method, e.g. learning - G06V 30/245: {Font recognition} - G06V 30/40: Document-oriented image-based pattern recognition - G06V 30/10: Character recognition - G07D: HANDLING OF COINS OR VALUABLE PAPERS, e.g. TESTING, SORTING BY DENOMINATIONS, COUNTING, DISPENSING, CHANGING OR DEPOSITING - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## Abstract An image processing method, for an identity document that comprises a data page, comprises acquiring a digital image of the page of data of the identity document. The method further comprises assigning a class or a super-class to the candidate identity document via automatic classification of the digital image by a machine-learning algorithm trained beforehand on a set of reference images in a training phase; processing the digital image to obtain a set of at least one intermediate image the weight of which is lower than or equal to the weight of the digital image; applying discrimination to the intermediate image using a discriminator neural network; and generating an output signal as output from the discriminator neural network, the value of which is representative of the probability that the candidate identity document is an authentic document or a fake. ## Claims ### Claim 1 A method for processing an image of a candidate identity document that comprises a page of data comprising a plurality of glyphs of a plurality of fonts, the method comprising a step of: acquiring a digital image of the page of data of the candidate identity document, each digital image being able furthermore to be extracted from a video stream recorded by said communicating object; characterized in that it further comprises steps of: assigning a class or a super-class to the candidate identity document via automatic classification of said digital image by a machine-learning algorithm trained on a set of reference images in a training phase during which each glyph is learnt individually for each font of each class, extracting a set of glyphs from the digital image and storing each glyph in the form of an individual image, processing each individual image in order to obtain a set of at least one intermediate image, each intermediate image corresponding to one individual glyph, each intermediate image being obtained via at least one step comprising processing each individual image with a recurrent neural network, wherein processing each individual image with the recurrent neural network includes obtaining as output, by cosegmentation, a set of intermediate images that each correspond to a segmentation of the glyph obtained at the end of a glyph-recognition step, discriminating between each glyph of each font of said intermediate image using a respective discriminator neural network, each individual glyph being considered to be probably authentic when it corresponds to a glyph learnt in a training phase, and generating an output signal by combining the output signals of the discriminator neural networks, the value of which is representative of the probability that the candidate identity document is an authentic identity document or a fake. ### Claim 2 (depends on 1) The method of claim 1, wherein the step of assigning a class or a super-class is carried out by an unsupervised learning algorithm or by a pre-trained classifier neural network. ### Claim 3 (depends on 2) The method of claim 2, wherein the step of assigning a class or a super-class is carried out by the pre-trained classifier neural network, and the pre-trained classifier neural network is a convolutional neural network. ### Claim 4 (depends on 1) The method of claim 1, further comprising, prior to the step of generating the output signal, steps of: extracting a set of at least one zone among: a zone corresponding to all or some of the machine-readable zone (MRZ) of the candidate identity document, a zone corresponding to the top banner (TOP) of the candidate identity document, and a zone corresponding to the exterior part (EXT) of the candidate identity document, each zone comprising glyphs, of at least one given font, individually extracting each glyph from the set of at least one zone, via optical glyph recognition, and comparing each extracted glyph to its corresponding glyph learnt in the training phase. ### Claim 5 (depends on 4) The method of claim 4, further comprising, when the set of at least one zone comprises variable data, a step of blurring the variable data in the training phase. ### Claim 6 (depends on 1) The method of claim 1, wherein the step of automatic classification comprises assigning a predetermined class depending on at least one of the following characteristics: the nature of the identity document; the country or state of origin of the identity document; the type of identity document; the version or generation number of the identity document. ### Claim 7 (depends on 1) The method of claim 1, further comprising a step of data augmentation in the training phase. ### Claim 8 (depends on 1) The method of claim 1, wherein the output signal is one among: a majority vote, a qualified vote, and a weighted vote of the signals output from the discriminator neural networks. ### Claim 9 (depends on 1) The method of claim 1, further comprising a step of inspecting optical security characteristics of the identity document. ### Claim 10 (depends on 1) A computer program stored in a non-transitory computer readable medium, the computer program comprising program-code instructions for executing the steps of the method of claim 1, when said program is executed on a computer. ### Claim 11 (depends on 1) The method of claim 1, wherein acquiring a digital image of the page of data of the candidate identity document comprises acquiring the digital image of the page of data of the candidate identity document with a communicating object.