Method for encoding biometric data using two-dimensional symbolism and integration thereof into an identity document

Abstract

The invention relates to a method (100) for encoding biometric data represented by a reference image (PI0) using two-dimensional symbolism, which method is designed to be implemented by a system for producing an identity document (10). Such a method comprises a step (130) of downsampling the reference image (PI0) in order to produce a first resampled image (PI1i) prior to the implementation of a step (140) of compressing the latter. The method further comprises a step (150) of upsampling the compressed image (PI2i) in order to produce a second resampled image (PI3i) having the same dimensions as the reference image (PI0) so as to produce a measure (SSIMi) of the similarity between the second resampled image (PI1i, PI3i) and the reference image (PI0). The encoding (180) of the compressed image (PI2i) in the form of a single two-dimensional symbol (20) is implemented only if (120y) the similarity measure (SSIMi) satisfies (170) a determined selection criterion.

Description

[0001] A method for encoding biometric data using two-dimensional symbolism and integrating this symbolism into an identity document.

[0002] The invention relates to the field of physical or digital identity documents. More specifically, it relates to identity documents containing two-dimensional symbolism encoding biometric information characterizing an individual, including an image, for example a photograph of a face, a fingerprint, an iris.

[0003] Originally designed to ensure the traceability of a part or product in the automotive sector, symbols in the form of barcodes are also used in the field of identity documents. This type of symbology allows for the encoding of various textual and / or graphic identification and / or biometric data, representing them as one or more machine-readable visuals or symbols. Such symbols can then be displayed on the screen of an electronic device or printed on physical media to create, for example, a national identity card, or more generally, any form of identification or proof of identity. In the field of identity, the symbology used is described as "two-dimensional" or "2D."It is indeed common to use a code commonly called a "QR code" (short for "quick response"), meaning that the content of this graphic code can be quickly decoded after being read by a suitable reader installed on an electronic device such as a smartphone or tablet. Each symbol is a two-dimensional barcode. Such a symbol consists of modules or square dots, usually black, arranged in a quadrilateral, generally a square, which is white. These modules define the information that the symbol represents or encodes. A 2D symbol can contain or represent more information than a one-dimensional symbol in the form of a linear barcode, that is, one made up of a single series of lines of varying thicknesses. There are various 2D symbol schemes.Examples include, but are not limited to, the QR code as defined by the ISO / IEC 18004 standard, the Flashcode, and the 2D-Doc. The latter format is a two-dimensional barcode-type symbolism similar to that used in QR codes, but unlike the latter, the information contained in a 2D-Doc symbol is secured using a digital signature based on asymmetric cryptography. A 2D-Doc symbolism relies on a standardized technology (ISO / IEC 16022) and offers enhanced authentication, ensures code integrity, and consequently, the non-repudiation of signed data, while retaining the properties of a printable or displayable graphic symbol that is automatically readable by a machine designed for this purpose.

[0004] In the field of identity, it is known that such 2D symbolism encodes biometric data in the form of an image, for example, a portrait of an individual holding a national identity card. Figure 1 illustrates a first example of a physical identity document 10, which consists of a rectangular card with a front 10A and a back OB. Each of the two sides 10A and 10B of the identity document 10 illustrated in Figure 1 is printed. The front 10A contains a high-definition, high-resolution image of a portrait of the holder and their personal information 12. This information 12 includes, for example, the holder's surname, first name, date of birth, nationality, etc. The back 10B contains other information about the holder, such as their residential address 13. This back 10B also contains a two-dimensional symbol 20 within a zone 14.Such a two-dimensional symbol 20 is advantageously a QR code arranged to represent, for example, an image similar to portrait 11 but of lower definition or in a compressed form of said portrait 11, or even other biometric data of said bearer.

[0005] A 2D symbol has a maximum encoding capacity per symbol. For example, a Denso QR code (ISQ / IEC18004) preferably has a maximum encoding capacity of 2953 bytes and comprises 177 x 177 dots or modules. For a physical or digital identity document, the information represented by a two-dimensional symbol consists of civil status information and biometric data, usually a photograph.

[0006] As an example, Figure 2 illustrates different two-dimensional symbols of the same symbology, in this case QR codes, according to different encoding capacities. It is clear from Figure 2 that the larger the size of the encoded information, the greater the definition or complexity of the two-dimensional symbol. Thus, according to the example illustrated in Figure 2, symbol 20a has a low EC encoding capacity (21 x 21 modules) and large modules or dots. In contrast, symbols 20b and 20c have increasing EC encoding capacities, exceeding that of symbol 20a. We can visually observe that the size of the modules or dots of symbol 20b (33 x 33 modules) and even more so that of the modules or dots of symbol 20c (177 x 177 modules) decrease sharply until they reach printing or display limits.With the 20c symbol, it becomes difficult to correctly distinguish the symbol's modules if the symbol's dimensions are small. It is therefore generally necessary to increase the display or printing dimensions of such a symbol to maintain its legibility, at the expense of the rest of the information conveyed by the physical or digital identity document. Indeed, known methods of printing, displaying, or engraving two-dimensional symbologies offer a more or less precise definition of the symbolism used to encode biometric data. It is therefore not possible to reduce the area dedicated to a 2D symbol below a certain limit. Similarly, the vast majority of known readers designed to decode 2D symbolisms have varying reading capabilities, requiring a minimum size for the dots or modules below which the reader cannot decode the symbolism.

[0007] To find a reasonable compromise between the bulk of two-dimensional symbols on an identity document and the usability of the information conveyed after decoding said symbols by known readers, some identity document issuers intentionally limit or restrict the resolution of the encoded image. As a result, the dimensions of the areas dedicated to 2D symbols are controlled and reasonable in relation to the rest of the information on the identity document. However, the image encoded by a 2D symbol, for example a portrait, due to its low resolution or due to excessive or inappropriate compression or resampling rates used to compress the image and reduce its resolution, sometimes proves to be of insufficient visual quality for effective use in identification processes, as the image may be considered too blurry or contain too many artifacts.The term “artifacts” refers to visually distracting features, distortions, or details that spontaneously appear in the compressed image due to lossy compression of an original image, and that are foreign to the original image.

[0008] The technique described in document FR 3 013 257 takes the opposite approach to this method of unreasonably reducing the resolution of an encoded image to be represented by a single 2D symbol. It thus provides a relevant solution to the problem posed by the maximum encoding limit of a symbol, for example, one based on QR codes. This document teaches the possibility of combining several QR codes (ISO / IEC 18004) to include a large amount of biometric data distributed across several juxtaposed two-dimensional symbols on an identity document. A document produced according to this method, similar to the one already described in connection with Figure 1, is illustrated in Figure 3.Thanks to this technical training, it is possible to encode an image, for example, a person's portrait, with a relevant resolution—that is, one greater than the encoding capacity of a single two-dimensional symbol—by distributing the information contained in the image across several 2D symbols. Thus, by reading multiple symbols jointly encoding the image, followed by concatenating the partial information respectively represented between the read symbols, an image can be reconstructed with a relevant and sufficient resolution to allow a security officer to identify the bearer of the identity document. As shown in Figure 3, area 14 on the reverse side 10B of document 10 contains a plurality of two-dimensional barcodes 20, spaced at an interval 21, unlike area 14 of document 10 illustrated in Figure 1.The two-dimensional barcodes 20 are aligned so that their adjacent sides extend parallel to each other, continuing into zone 14. The encoding capacity of the symbology used to produce document 10, as shown in Figure 3, is therefore ten times greater than that used to produce document 10 in Figure 1. However, this technical approach has drawbacks. On the one hand, it becomes necessary to complicate the encoding and decoding process to compose, juxtapose, and use a plurality of two-dimensional symbols on the same identity document. On the other hand, such juxtaposition results in significant clutter on the document, as shown in Figure 3. In this example, the plurality of two-dimensional symbols 20 occupies almost half (zone 14) of the reverse side of document 10. Furthermore, such a graphic arrangement of the identity document can be aesthetically unappealing.Finally, this technique requires complicating the process of decoding the encoded identification data, requiring an ordered reading of the symbols followed by a controlled concatenation of the data distributed within said symbols to reconstruct the image.

[0009] Figures 4 and 5, taken from document FR 3 013 257, present the respective functional architectures of a system 30 for producing an identity document, such as a national identity card, and of a system 40 for automatically reading symbols used to encode biometric identification data, such as a portrait, a fingerprint, or an iris scan. The identity document production system 30 illustrated in Figure 4 produces and prints two-dimensional symbols to encode an image resulting from the capture of a subject's biometric information by a capture device, for example, a camera 36 capable of taking photographs of a face or iris, or even a scanner 37 designed to capture a digital representation of the subject's fingerprints.The system 30 includes an information processing unit 31, for example in the form of a microcomputer forming a terminal connected to a communication network 32 (internet or intranet). This information processing unit 31 includes an input / output human-machine interface (a screen 34 and a keyboard 35) for entering information, particularly concerning the civil identity of the individual commonly referred to as the "bearer" for whom the identification document is being issued. This unit 31 is connected via the network 32 to a remote database 33 designed to store, for each individual, information relating to their civil status, biometric information such as a portrait, fingerprints, or iris scan. This database 33 may also contain security information concerning the individuals, indicating, for example, whether an individual is wanted and / or prohibited from entering certain territories.The information processing unit 31 is connected to an output device, such as a printer 38 or a computer screen 34, configured to print or display two-dimensional symbols on a produced identification document, such as the symbols 20 on the reverse side 10B of document 10 illustrated in Figure 3. The information processing unit 31 comprises one or more microprocessors or microcontrollers 31U cooperating with one or more (internal or remote) data and program memories 31M. To control, in particular, the various peripherals 34, 35, 33, 36, 37, 38 mentioned above, such a program memory 31M of the information processing unit 31 contains one or more computer programs whose program instructions trigger the implementation of appropriate functional processes. The term "memory" refers to any computer memory, whether volatile or not.Non-volatile memory is a type of computer memory whose technology retains its data even when no electrical power is supplied. It can contain data resulting from input, calculations, measurements, and / or program instructions. The main non-volatile memories currently available are electrically writable, such as EPROM (Erasable Programmable Read-Only Memory), or electrically writable and erasable, such as EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, SSD (Solid-State Drive), etc. Non-volatile memories are distinct from volatile memories, whose data is lost when power is removed.The main volatile memories currently available utilize RAM ("Random Access Memory" according to Anglo-Saxon terminology or also called "vm" memory), DRAM (dynamic vm memory, requiring regular updating), SRAM (static vm memory requiring such updating during an electrical under-powering), DPRAM or VRAM (particularly suited to video), etc.

[0010] Figure 5 illustrates a system 40 for reading two-dimensional symbols printed on an identity document 10 produced by a system 30 as shown in Figure 4, and for identifying a subject. Such a system 40 includes, for example, a symbol reader 46, for symbols such as two-dimensional barcodes. This reader 46 consists, for example, of a mobile phone equipped with a camera and suitable reading software. This reading software is adapted to detect and read two-dimensional symbols. The system 40 further includes a terminal 41 connected to the remote database 33, mentioned in connection with Figure 4, and to the reader 46, for example, via a wireless connection such as Wi-Fi or any other wireless or wired proximity communication protocol 43. A fingerprint scanner 47 can also be connected to the terminal 41.The latter is arranged to compare the civil status and biometric information decoded by the reader 46 with the reference information contained in the database 33. Depending on the distance or measure in terms of dissimilarity or similarity between said reference data (recorded in the database 33) and that read by the reader 46, said terminal 41 can accept or reject the identification of the bearer of the document 10 read by the reader 46. The quality of the information encoded by the symbols and decoded by the reader 46 is paramount to allow such a comparison with a reference (from the database 33) by the terminal 41.Regardless of the solution chosen to symbolically represent identity data (whether through a sometimes excessive reduction in the resolution of the encoded image or by distributing it across multiple symbols), issuers of identity documents must resign themselves to reducing the space allocated to the symbolic representation of identification data. Ultimately, a reduction or compression of biometric data is necessary to encode it in an increasingly constrained space, with the aforementioned drawbacks of such compression. This compression often proves inappropriate or excessive for certain subsequent applications, indiscriminate regardless of the data being encoded, or poorly controlled. Figure 6 illustrates the impact of such inappropriate compression.Figure 6 shows four subjects, S1 to S4, each with a portrait captured to produce a high-resolution PIO reference image (350 x 450 pixels in this case) with a seven-ninth aspect ratio (e.g., a 45 mm x 35 mm passport photograph as required by the International Civil Aviation Organization, ICAO). For comparison purposes, Figure 6 also shows, for each of the four subjects, a second SCPI image compressed in a standardized and undifferentiated manner, for example, using a known lossy compression algorithm such as AVI F (AV1 Image File Format) or WebP, the quantization of which has been established, for example, dichotomously, to obtain an SCI image whose data volume can be encoded as a single two-dimensional symbol, such as a QR code.As an example, Figure 6 illustrates compressed images with respective sizes on the order of a thousand bytes. Although a compression algorithm was chosen for its excellent compression capabilities while preserving high visual quality, it is easy to see that these SCPI-compressed images contain blur or artifacts that degrade identification, or even make automatic identification impossible. To emphasize the degradations caused by producing such second compressed images using state-of-the-art methods, Figure 6 also presents, for each of the four subjects S2 to S4, areas of interest (eye, nose, mouth) before and after compression, as well as enlargements of these areas of interest.Certain particularly characteristic features of a subject almost completely disappear in the second SCPI images, resulting in high dissimilarities calculated by implementing a method for measuring the similarity between the first PIO images and the second SCPI images using an identification system similar to the 40 system illustrated in Figure 5. Indeed, the details of the iris of the right eye (area IA) of subject S2 are completely blurred in the compressed SCPI image, as shown in the EIA enlargement of this area of ​​interest. Similarly, the outline of the lips of subject S1 disappears completely after compression. The same is true for the outline of the nose of subject S2, which appears totally blurred after compression. Alternatively, the right eye (area IA) of subject S3 contains numerous artifacts in the compressed SCPI image, as shown in the EIA enlargement of this area of ​​interest.The same applies to the outline of the lips of the mouth of said subject S3. As indicated in said figure 6, the area of ​​interest IA relating to the right eye or the mouth of subject 4, shows the combined presence of blur and artifacts, as highlighted by the EIA enlargements of said area IA.

[0011] The invention addresses the aforementioned drawbacks by providing a method for compressing, selecting, and encoding an individual's identification data, including an image such as a portrait, onto a single two-dimensional symbol, while preserving sufficient and relevant visual quality of said image after decoding the symbol for the purpose of identifying the individual. Among the numerous advantages offered by implementing the invention, we can mention that the method provides:

[0012] - an optimal adaptation of the compression and selection of an image whose symbolic representation must be produced, according to a determined symbolism, criteria for editing the identity document (maximum space devoted to printing or displaying the symbol encoding said compressed image) and criteria for reading said determined symbolism and minimum visual quality required of the image obtained after reading and decoding said determined symbolism;

[0013] - an optimal adaptation of the compression and selection of an image for which a symbolic representation must be made, depending on the content of said identification data (framing, non-essential and / or extraneous information);

[0014] - a plurality of arrangements or variants of implementation of such a process of compression, selection and encoding of an individual's biometric data depending on the implementation capacity of the machine or system implementing such a process, or even the maximum duration granted or permitted for such implementation.

[0015] To this end, the invention provides a method for encoding biometric data using two-dimensional symbolism, the maximum encoding capacity of which for each two-dimensional symbol is predetermined. This method is designed to be implemented by an information processing unit within an identity document production system. It comprises:

[0016] - a step involving the collection of biometric data in the form of a reference image of predetermined dimensions,

[0017] - a step of producing a compressed image from said reference image;

[0018] - an encoding step of said compressed image in the form of a two-dimensional symbol of said dimensional symbolism.

[0019] To avoid altering the visual quality of the image obtained through compression, whatever the method, such a process includes:

[0020] - a subsampling step of the reference image according to a determined resampling rate to produce a first resampled image prior to the implementation of the compressed image production step, said compressed image production step consisting of the compression of said first resampled image;

[0021] - a step of upsampling the compressed image to produce a second resampled image of the same dimensions as the reference image;

[0022] - a step of producing a similarity measure between the second resampled image and the reference image.

[0023] Furthermore, the step of encoding the compressed image into a two-dimensional symbol is only implemented if the compressed image satisfies a determined selection criterion based on the similarity measure of the second resampled image associated with it.

[0024] To be able to adapt to the characteristics of a reference image and ultimately select a promising compressed image from a plurality of candidates produced, a process according to the invention includes a step of determining the resampling rate used to produce a first resampled reference image.

[0025] Thus, such a process can be arranged so that they are iterated sequentially until an iteration criterion is met:

[0026] - the step of determining the resampling rate;

[0027] - the step of subsampling the reference image;

[0028] - the stage of producing a compressed image;

[0029] - the step of oversampling said compressed image to produce a second resampled image;

[0030] - the step of measuring similarity between the second resampled reference image and the reference image.

[0031] According to an advantageous embodiment, the iteration criterion can be satisfied as soon as a compressed image satisfies a determined selection criterion which is based on the measurement of similarities of the second resampled image associated with it, said process therefore comprising a step of selecting said compressed image for the subsequent implementation of the step of encoding the selected compressed image in the form of a single two-dimensional symbol.

[0032] Alternatively, a method according to the invention can be adapted so that the compressed image and the similarity measurement between the second resampled image and the reference image, respectively produced at each of said iterations, can be stored as a data set in a data memory of the information processing unit. The iteration criterion can then consist of a minimum number of iterations to be achieved. The method can also include, to satisfy said iteration criterion, a step for selecting the compressed image for the subsequent implementation of the encoding step of the selected compressed image as a single two-dimensional symbol, said selection step consisting of:

[0033] - reading from the data memory the datasets containing respectively the compressed images and the similarity measures of the second resampled reference images respectively associated;

[0034] - the selection of the compressed image associated with the best similarity measure with respect to those respectively associated with the compressed images included in said read datasets.

[0035] In order for the said identity document produced to be directly examined by security personnel, the similarity measurement step may advantageously consist of calculating a measure of the visual quality of the second resampled image relative to the reference image.

[0036] To prevent suboptimal or biased determination of the resampling rate for producing a compressed image with a high similarity measure influenced by the presence of secondary or extraneous information in the reference image with respect to one or more regions of interest in the reference image, a method according to the invention may include a step of determining a region of interest within the reference image. In this case, the step of producing a similarity measure between the second resampled image and the reference image relates only to said region of interest.

[0037] To produce a compressed image with a volume smaller than the encoding capacity of a two-dimensional symbol while preserving as much as possible the structures of the reference image after compression, the quantification used for implementing the compressed image production step can be determined using a binary search algorithm so that the volume of compressed data is less than a predetermined target volume that is a function of the maximum encoding capacity of a two-dimensional symbol.

[0038] To define a relevant trade-off between finding an optimal compressed image and the time or hardware resources required to perform this search, the iteration criterion can consist of a predetermined number of iterations to be carried out. The step of determining the resampling rate used to produce a first resampled image can then consist of calculating this resampling rate so that its value changes at each iteration by a predetermined step within one or more predetermined ranges of possible values.

[0039] As an alternative or in addition to seeking such a compromise, the invention provides that such ranges of possible values ​​of the resampling rate can be chosen in such a way that said rate induces a measure of average similarity between the second resampled images produced during the iterations and the reference image which is greater than a predetermined threshold.

[0040] To benefit from feedback on previous encodings involving reference images and similar compression constraints, the quantification retained for the implementation of the compressed image production step can be established from a history of previous implementations of said biometric data encoding process under a two-dimensional symbolism by the information processing unit of an identity document production system, having respectively produced compressed images by an identical or similar compression algorithm from distinct reference images and having respective dimensions similar to those of the compressed image that said process seeks to select.

[0041] To further benefit from feedback on previous encodings involving reference images and similar compression constraints, a predetermined range of possible values ​​for the resampling rate can be established from a history of previous implementations of said biometric data encoding process under two-dimensional symbolism by the information processing unit of an identity document production system, having produced respectively compressed images by an identical or similar compression algorithm from distinct reference images and having respective dimensions similar to those of the compressed image that said process seeks to select.

[0042] Furthermore, the biometric data represented by a reference image can be associated with additional identification information. To encode such information jointly with said biometric data in the form of a single two-dimensional symbol, a method according to the invention may include a step of collecting identification data complementary to the biometric data. Moreover, the predetermined target volume for retaining the quantification used to implement the compressed image production step is then calculated to be less than or equal to the maximum encoding capacity of each symbol of the two-dimensional symbolism, minus the volume of said additional identification data.The encoding step of the selected compressed image in the form of a two-dimensional symbol can then be arranged so that said two-dimensional symbol jointly encodes said additional identification data and said selected compressed image.

[0043] To preserve the volume of data dedicated to the compression of the reference image that we seek to encode in a two-dimensional symbol, such a process according to the invention may advantageously include a step of reducing the additional identification data collected in order to retain only additional identification data of interest.

[0044] Similarly, when biometric data results from a capture phase producing a large image, a method according to the invention may include a step of producing the reference image of predetermined dimensions by lossless compression of a second image of larger dimensions, said second image representing said biometric data.

[0045] According to a second object, the invention relates to a method for producing a digital or physical identity document, said method being designed to be implemented by an information processing unit of an identity document production system. To this end, such a method comprises:

[0046] - a step of digitizing biometric data and producing an image representing said biometric data;

[0047] - a triggering step for the implementation of a process - according to the invention - for encoding biometric data represented by said image produced, under a two-dimensional symbolism;

[0048] - a step of integrating the two-dimensional symbol generated by said implementation of said biometric data encoding process into the identity document thus produced.

[0049] Such a process for producing an identity document may include a step of digitizing additional identification information and producing a set of additional identification data.

[0050] According to a third object, the invention relates to a computer program product comprising one or more program instructions executable by an information processing unit of an identity document production system, said program instructions being: - loadable into a memory of said information processing unit;

[0051] - designed so that their execution by said information processing unit causes the implementation of a biometric data encoding process under a two-dimensional symbolism and / or a process for producing an identity document in accordance with the invention.

[0052] To convey such a computer program product, the invention further relates to a computer-readable storage medium containing the instructions for said computer program product according to the invention.

[0053] According to a fifth object, the invention relates to a system for producing an identity document comprising:

[0054] - an information processing unit comprising a memory recording the program instructions of a computer program product conforming to the invention;

[0055] - a biometric data capture device;

[0056] - an output device for printing or displaying an identity document.

[0057] Other features and advantages will become clearer upon reading the following description and examining the accompanying figures, including:

[0058] - Figure 1, already described, illustrates a first known identity document presenting a single two-dimensional symbol encoding biometric data;

[0059] - Figure 2, already described, illustrates two-dimensional symbols according to their respective encoding capabilities;

[0060] - Figure 3, already described, illustrates a second known identity document presenting a plurality of two-dimensional symbols jointly encoding biometric data; - Figure 4, already described, illustrates a production system for a known identity document;

[0061] - Figure 5, already described, illustrates a known system for reading two-dimensional symbols present on an identity document and for identifying the bearer of the latter;

[0062] - Figure 6 illustrates the disadvantages induced by image compression techniques known prior to encoding such compressed images in the form of two-dimensional symbols;

[0063] - Figure 7 illustrates a functional description of an optimal image compression process to be encoded in the form of a predetermined symbolism in order to produce an identity document, for example, physical or digital;

[0064] - Figure 8 illustrates the advantages provided by the implementation of a process according to the invention on two initial examples of portraits intended to be compressed and then encoded in the form of two-dimensional symbols;

[0065] - Figure 9 illustrates the advantages provided by the implementation of a process according to the invention on two second examples of portraits intended to be compressed and then encoded in the form of two-dimensional symbols.

[0066] Figures 4 and 7 illustrate the general principle of the invention. This mainly relates to a method 100 for encoding biometric data represented by a PIO reference image in a two-dimensional symbology. Such a method 100 is designed to be implemented by an identity document production system 10 such as that illustrated by way of non-limiting example in Figure 4. To adapt the operation of such a system 30, suitable program instructions are written into a program memory 31M of the information processing unit 31 of said system 30. Such instructions are designed to trigger the implementation of the method 100 when executed by a microprocessor 31U of said information processing unit 31.

[0067] Such a process 100, like known processes, includes a step 101 of collecting biometric data in the form of an image PI. When this image originates directly from a capture device, such as a camera 36 or a scanner 37, the reference image PI can be of very high resolution. Traditionally, a process 100 may include a first step 102 of producing a reference image PI0 of predetermined dimensions, reduced by lossless compression of the larger PI image. By way of non-limiting example, such a step 102 may consist of implementing a known compression algorithm such as PNG (Portable Network Graphics), GIF (Graphics Interchange Format), TIFF (Tagged Image File Format), or equivalent algorithms.Thus, a biometric image PI0 depicting a portrait (examples of which are already described in connection with Figure 6) can have dimensions of 350 by 450 pixels, or any other predetermined dimensions. A method 100 according to the invention comprises, like known methods, a step 140 of producing a compressed image PI2i from said reference image PI0 followed by a step 180 of encoding said compressed image PI2i in the form of a two-dimensional symbol 20, for example, a two-dimensional barcode using known and suitable encoding techniques.Such a step 140 implements a known lossy compression algorithm such as AVI F (Anglo-Saxon acronym for "AV1 Image File Format") or WebP whose quantification is established, for example, in a dichotomous way so that the compressed image describes a volume of data less than the maximum volume of data that can be encoded by a two-dimensional symbol 20.

[0068] To address the technical problems raised by prior techniques (in this case, excessive degradation of the compressed image which does not allow the implementation of efficient visual or automatic identification procedures, and / or a multiplication of two-dimensional symbols integrated on the same identity document to represent large volumes of biometric data), a process 100 according to the invention includes, prior to the compression 140 of the reference image PI0, a subsampling step 120, 130 (i.e. resampling to dimensions smaller than those of the reference image PI0) of the latter to produce a first resampled image PI1 i.A method 100 according to the invention further comprises, after the implementation of step 140 of producing a compressed image PI2i by compressing the first resampled image P11i, a step 150 of upsampling said compressed image PI2i to produce a second resampled image PI3i having dimensions identical to those of the reference image PI0. Indeed, in order to ultimately encode only compressed images suitable for the intended use of the electronic or physical documents 10 that will respectively convey them via a two-dimensional symbol 20, a method 100 according to the invention comprises a step 160 of producing a similarity measure SSIMi between said second resampled image PI3i and the reference image PI0.According to an advantageous embodiment, such a similarity measure (SSIMi) can be performed by calculating a visual quality measure of the second resampled image (PI3i) relative to the reference image (PI0). For example, such step 160 can utilize the algorithm known by the acronym SSIM for "Structural SI Milarty." This algorithm was designed to measure the similarity of structures between two images, as the human eye is more sensitive to structural changes between images, unlike a pixel-by-pixel difference measurement between two images, as is possible with the PSNR (Peak Signal-to-Noise Ratio) algorithm, for instance. However, step 160 could implement the latter algorithm instead of the SSIM technique or any other technique capable of producing such a similarity measure.The objective of this step 160, coupled with step 150, is to be able to envision the subsequent use of the biometric data encoded by the two-dimensional symbol 20 after its decoding, for example by a system as shown in Figure 5. If the measured similarity SSIMi is deemed insufficient, the compressed image is rejected by the implementation of process 100 before its encoding as a two-dimensional symbol 20. The invention thus avoids the "blind" retention of a compressed image of low relevance, thereby ultimately preventing the production of an identity document (such as a document 10 shown in connection with Figure 1) incorporating said two-dimensional symbol 20 carrying biometric data unsuitable for the intended use of said identity document 10.Thus, a method 100 according to the invention includes a step 170 aimed at selecting a compressed image PI2s, for the purpose of encoding it 180 in the form of a single two-dimensional symbol 20, if and only if said compressed image PI2s satisfies a determined selection criterion based on the SSIMi similarity measurement of the second resampled image PI3i associated with it.

[0069] According to a first embodiment, such a selection criterion may consist of a predetermined minimum threshold that the SSIMi similarity measure of the second resampled image PI3i associated with the compressed image PI2i must reach after resampling the latter to recover dimensions similar to the reference image PI0. Such a threshold may be that of the SSIMi similarity between an image produced by compressing the reference image PI0 without prior resampling of the latter to smaller dimensions. Such a threshold is, for example, symbolized by a discontinuous horizontal line on the SP curve associated with said reference image PI0 or with a region of interest thereof PI0', as will be discussed later.

[0070] When a compressed image is associated with a similarity measure deemed insufficient (i.e., whose visual quality is too far removed from that of the reference image), it is rejected, and process 100 may involve a further iteration of the steps consisting of sequentially:

[0071] - determine 120 a new resampling rate T1 distinct from the previous one; - subsample 130 the reference image (PIO) according to said new resampling rate T1 and obtain a new first resampled image PI1 i;

[0072] - compress 140 of the said new first resampled image PI1 i and produce a new compressed image PI2i;

[0073] - upsample the latter by 150 to obtain a new second resampled PI3i image;

[0074] - measure 160 the SSIMi similarity of the latter with the PIO reference image;

[0075] - select the new compressed image PI2s, for its encoding 180 in the form of a single two-dimensional symbol 20, if and only if said new compressed image PI2s satisfies a determined selection criterion based on the SSIMi similarity measurement of the second resampled image PI3i associated with it.

[0076] The invention provides that said steps 120, 130, 140, 150, and 160 can be iterated as long as an iteration criterion is not met (a situation symbolized by link 120n in Figure 7). According to one embodiment, such an iteration criterion can be met as soon as a compressed image PI2i satisfies said selection criterion (a situation symbolized by link 120y in Figure 7). The process 100 then triggers the implementation of the encoding step 180 of said selected compressed image PI2s in the form of a single two-dimensional symbol 20.

[0077] Alternatively, the invention provides for other iteration criteria. Thus, sequences of implementations of said steps 120, 130, 140, 150, and 160 can be iterated sequentially or concurrently if the information processing unit 31 allows it. For each sequence (therefore for a given resampling rate value T1), the compressed image PI2i and the similarity measurement SSIMi between the second resampled image PI3i and the reference image P1, respectively produced and calculated, are recorded as a dataset {PI2i} in a data memory 31M of the information processing unit 31 implementing said process 100. The sequential iterations of said steps 120, 130, 140, 150, and 160 thus produce several datasets {PI2i}*, respectively associated with the compressed images PI2i.It is only when an iteration criterion is satisfied (situation symbolized by link 120y in Figure 7) that a method 100 according to the invention includes a selection step 170 of the compressed image PI2i for the subsequent implementation of the encoding step 180 of the selected compressed image PI2s in the form of a single two-dimensional symbol 20. Such a step 170 may consist of:.

[0078] - reading from the 31M data memory of the {Pi2i}* datasets containing respectively the PI2i compressed images and the SSIMi similarity measures respectively associated;

[0079] - the selection of the PI2i compressed image associated with the best of the SSIMi similarity measures with regard to those respectively associated with the compressed images included in said {PI2i}* data sets read.

[0080] An example of an iteration criterion may consist of a minimum number of iterations of the sequence of steps 120, 130, 140, 150 and 160. For example, such a minimum number of iterations may advantageously be equal to or a multiple of the number of cores of the microprocessor(s) 31U of the information processing unit 31 implementing said process 100. Thus, it is only at the end of the satisfaction of said iteration criterion (situation symbolized by link 120y in Figure 7) that the selection 170 of one of the compressed images taken from one of the datasets {PI2i} is accomplished for the subsequent implementation of the encoding step 180 of the selected compressed image PI2s in the form of a single two-dimensional symbol 20.Thus, as Figure 7 shows by way of example, the invention provides that said process 100 implements eight sequences of steps 120, 130, 140, 150, and 160, with a resampling rate Ti whose value increases within a range of values ​​between 20% and 90%, in 10% increments. Figure 7 is represented by a curve SP exhibiting a maximum M for a resampling rate d equal to 60%. The compressed image PI2s selected in step 170 will therefore be the image associated with said resampling rate Ti equal to 60%. The invention cannot be limited to this single example of an iteration criterion equal to a number of iterations of sequences of steps 120, 130, 140, 150, and 160 equal to eight. Four, ten, sixteen or any other number of iterations could be chosen instead.By choosing a limited number of sequence iterations, for example eight, as opposed to a full sweep of possible resampling rate values, it is thus possible to divide the time required to select a good quality compressed image by a factor of between twenty and fifty.

[0081] The invention provides for different embodiments allowing respectively to improve the selection criterion and / or to reduce the time (or material resources such as microprocessors or microprocessor cores) required to implement a process 100 in accordance with the invention.

[0082] To reduce the number of iterations of steps 120 to 160, without compromising the quality of the compressed image to be selected in step 170, one solution is to reduce the range of possible values ​​for the resampling rate TL. Thus, instead of selecting a wide range between 20% and 90% as previously mentioned, which risks varying the resampling rate TÏ in large increments (e.g., 10%) to avoid excessively increasing the number of iterations, the iteration criterion 120 can be arranged so that the resampling rate TÏ can vary in smaller increments (e.g., 5% versus 10%) within a narrower, or even median, range, for example, between 40% and 60%. However, there is no guarantee that such a range will remain relevant for all reference images PI0.Indeed, experience has shown that the resampling rate Ti selected to produce a compressed image with an excellent similarity measurement can be close to a value of 25%, 60%, or 85%, depending on the reference images PI0 involved. To select a narrow and relevant range, with the aim of drastically reducing the time required to select a very high-quality compressed image (experience has shown a reduction in the implementation time of the process by a factor of ten compared to a full traversal of the possible resampling rate values), the invention provides an embodiment consisting of determining a range of possible values ​​of said resampling rate Ti, referenced p in Figure 7, around a maximum M of a theoretical, interpolated, or even modeled curve, passing through several measurement points SSIMi for discrete and increasing resampling rate values.Such an SP curve is described in Figure 7. Thus, a range p of possible values ​​of the resampling rate TÏ can be determined as comprising the value of the rate associated with the maximum M of said SP similarity measure curve (in this case a rate equal to 60% for the example of the SP curve in Figure 7), with as a minimum bound said value of resampling rate associated with the maximum M of the SP curve reduced by a given step (for example 5%, i.e. in this case 55% for the example of the SP curve in Figure 7), and as a maximum bound said value (60%) associated with said maximum M increased by said given step (for example 5%, i.e. in this case 65% for the example of the SP curve in Figure 7).Such a range p could also include minimum and maximum bounds not evenly distributed on either side of the value of the resampling rate associated with the maximum M but determined differently so as to be off-center from said maximum while including the value of the resampling rate associated with said maximum M of the SP curve.

[0083] The SP curve illustrated in Figure 7 as an example shows a single maximum M. However, such an SP curve could describe several maxima. Indeed, the SSIMi measurement can generally exhibit highly variable amplitudes from one step to the next and does not describe a smooth curve as the SP example seems to suggest. The invention therefore provides for retaining as many ranges (centered on or) containing a maximum as there are maxima, such as the range p described previously, to vary the resampling rate within said ranges.

[0084] To optimize the compression step 140 by pre-selecting a promising quantization or by facilitating the convergence of the search for such a relevant quantization, or even by selecting a narrow and relevant range to vary said resampling rate and avoid having to implement a large number of iterations, the invention provides an alternative embodiment based on learning or feedback following multiple previous implementations of such a method 100 for compressing different reference images. To establish a history necessary for this feedback, the invention provides that the dataset {PI2i} associated with a compressed image PI2s selected in step 170 can be recorded in a persistent memory 31M of the information processing unit 31.Such a data record {PI2i} can be advantageously enriched by the chosen quantification and / or the data volume of the compressed image during the implementation of production step 140 of the compressed image PI2i, or even by other metadata characterizing said reference image PI0, or even the resampled images associated with it, such as the dimensions of images PI0, PI2s, PI11i, PI3i, etc. Thus, each implementation of a process 100 makes it possible to retain, after the selection of a relevant compressed image in 170, the characteristics of the images concerned and the implementation parameters of said process 100, thus constituting, as and when implementing a process 100 according to the invention, an incremental history {PI2s}*.

[0085] A method 100 according to the invention can then be arranged so that step 120 of satisfying the iteration criterion exploits, that is to say reads and traverses, said history {PI2s}* in the light of characteristics of the reference image PI0 (portrait, biometric fingerprint, iris, contrast, colors), subject of compression, and implementation parameters (target or maximum volume of the compressed image, compression algorithm, quantization, etc.) of said method 100. Said step 120 thus consists of selecting from the set of records forming the history {PI2s}*, the one which has the most similarities with the image PI0 concerned, the expected compression objective, the compression algorithm exploited, etc.Such a dataset extracted from said history includes the quantization associated with the compression algorithm that jointly produced a compressed image PI2i, whose oversampled image PI3i has a very good similarity measure SSIMi. The iterations required to complete step 140 to determine the appropriate quantization, according to a dichotomous algorithm for example, are thus reduced and facilitated by reading the previous quantization from the history {PI2s}*.

[0086] Furthermore, such a dataset extracted from said history includes the resampling rate that produced the most promising and high-quality compressed image. Step 120 can thus consist of determining the range of possible p-values ​​for varying the resampling rate Ti, including the value of said resampling rate Ti retained in said history.Thus, for example, if the historical data {PI2s}* indicates a previously retained resampling rate value ti of 63%, such a range of values ​​p' for the present implementation of said method 100 could advantageously be determined as being centered on said historical value, with the calculated minimum value being equal to said historical value 62% deduced by a determined step, for example set at 5%, i.e., in this case 57%, and the calculated maximum value being said historical value 62% increased by said determined step, i.e., in this case 67% for a step, for example, equal to 5%. A number of iterations can thus be determined to traverse said range of values ​​and produce the set {PI2i}* used by step 170 to select the compressed image associated with the best SSIMi similarity measure.Alternatively, the traversal of this range can be summarized as simply selecting the previous value (62% in the preceding example) of the resampling rate from the dataset, derived from the {PI2s}* history, considered closest to the specific case. This reduces the {PI2i}* data structure to a single {PI2i} dataset associated with a single PI2i compressed image, presumed to be the best candidate. Leveraging the {PI2s}* history of previous implementations of process 100 thus allows the effort of producing PI2i candidate compressed images (iterations of steps 120 to 160) to be concentrated on a restricted and relevant range of resampling rate values.

[0087] Regardless of the technique used to retain one or more ranges p, p', the invention provides that the iterations of steps 120 to 160 can cease, during a traversal of a range p, p', as soon as a similarity measure SSIMi describes a measurement lower than that SSIMi-1 produced in the previous iteration. The invention provides, alternatively or in addition, that step 120, which determines the value of the resampling rate TÏ, can determine a variable step size, for example, a smaller one, to produce a resampling rate Ti with respect to the value TÏ-1 of said rate produced in the previous iteration, particularly when the SSIMi-1 measurement produced in the previous iteration is lower than that produced during the penultimate iteration.Any other strategy allowing variation of said resampling rate could be considered as an alternative to limit the number of iterations of the sequence of steps 120 to 160 while maintaining a priori a range of relevant p values.

[0088] As Figure 6 shows, the information contained within the reference images PI0 can vary greatly, depending on the subject's hair, hairstyle, accessories, or even the background, which may describe "extraneous" details or structures, or at least secondary features with respect to one or more areas of interest (for example, a face in a portrait) defined by the intended use or primary purpose of the document that will carry the two-dimensional symbol encoding said image. The implementation of the method 100 according to the invention could be suboptimal or biased if the determination of the optimal resampling rate for producing a compressed image with a high degree of similarity were unduly influenced by the presence of such extraneous or secondary information with respect to said area(s) of interest in the reference image PI0.To overcome this drawback, the invention provides that a method 100 according to the invention may include a step 111 for determining a region of interest RIO' (possibly multiple or plural) within the reference image PIO, using any known techniques, such as, for example, the Viola-Jones method or any other equivalent methods adapted to the intrinsic characteristics of such a region of interest. In this case, the step 160 for producing a similarity measurement SSIMi between a second resampled image PI3i and the reference image PIO is arranged to focus solely on said region of interest PIO'.Thus, the said accessories, decorations and other extraneous or secondary information are ignored to determine the resampling rate producing promising compressed images and to ultimately select, in step 170, the compressed image associated with the best measure of similarity, with regard to the area of ​​interest PIO' within the reference image PIO, to be encoded in step 180 in the form of a two-dimensional symbol 20.

[0089] Figures 8 and 9 illustrate the contribution of the invention with regard to the problems illustrated by Figure 6. Figure 8 thus allows us to compare, even if only visually, the SCPI compressed images in a standardized and undifferentiated manner (for example by the implementation of a known lossy compression algorithm such as AVIF (Anglo-Saxon acronym for "AV1 Image File Format") or WebP whose quantification has been established, for example in a dichotomous way, to obtain an SCPI image whose data volume can be encoded in the form of a single two-dimensional symbol, such as a QR code) already illustrated in Figure 6, and the PI2s compressed images selected according to the invention. Figure 8 also mirrors said SCPI and PI2s images, the PIO reference images of subjects S1 and S2 (see figure 6) and more particularly EIA enlargements of certain IA areas of interest selected on said PIO reference images.It is evident that the right eye or the lip contour of subject S1 is more closely related and similar in the EIA enlargement derived from the PI2s images than in the EIA enlargement of the SCPI images. Similarly, the nose contour of subject S2 is much less altered in the EIA enlargement of the PI2s image than in the SCPI image, making the PI2s image more visually and structurally accurate.Similarly, Figure 9 allows for a comparison, even if only visually, between SCPI-compressed images produced in a standardized and undifferentiated manner (for example, by implementing a known lossy compression algorithm such as AVI F (the English acronym for "AV1 Image File Format") or WebP, the quantization of which has been established, for example, dichotomously, to obtain an SCPI image whose data volume can be encoded as a single two-dimensional symbol, such as a QR code), as illustrated in Figure 6, and the PI2s-compressed images produced and selected according to the invention. Like Figure 8, Figure 9 also mirrors said SCPI and PI2s images, the PIO reference images of subjects S3 and S4, and more specifically, EIA enlargements of certain IA areas of interest selected from said PIO reference images.For example, subject S4 wears an accessory garment rich in detail and information that could influence process 100 according to the invention, based solely on the person's face. The same applies to subject S2 in Figure 8, whose garment presents information similar to or comparable to that of the subject's hair. Thanks in particular to the variant of the invention (step 111 and adaptation of step 160) which allows focusing solely on the face as the primary area of ​​interest PIO', it is clear that the right eye or the lip contour of subjects S3 and S4 is closer and more similar in the EIA enlargement derived from the PI2s images than in the EIA enlargement of the SCPI images. The numerous artifacts present on the right eye of subject S3 in the EIA enlargement of the SCPI image disappear in the enlargement of the PI2s image, resulting in a very reasonable degree of blurring. The same is true for the lip contour of subject S4.

[0090] As indicated in connection with the previous description of Figures 1 and 4, an individual's identification data is generally not limited to a single image depicting biometric data (portrait, fingerprint(s), iris). Complementary identification data (possibly alphanumeric) to said biometric data, or even metadata characterizing the image representing said biometric data, may also be encoded together with said image in the form of a single two-dimensional symbol 20. It is indeed advantageous to encode all or part of complementary identification data IPD, such as that derived, for example, from information 12 and / or 13 of document 10 as shown in Figure 1, alongside a compressed image derived from a reference image PI0 within said two-dimensional symbol 20.Indeed, it becomes possible to compare the aforementioned supplementary identification data obtained after decoding the two-dimensional symbol to ensure that the identity document 10 bearing the symbol has not been subject to any attempt at deliberate or accidental alteration. To this end, a method 100 for encoding biometric data using a two-dimensional symbolism according to the invention includes a step 103 for collecting such supplementary identification data (SID).

[0091] Step 140 of said process 100 is then adapted to deduce from the predetermined target volume used to retain the quantification employed for implementing step 140 of compression of the reference image PI0, the volume of said supplementary identification data (IPD). Thus, said target data volume for the compression of the image PI0 is calculated to be less than or equal to the maximum encoding capacity of each symbol of the two-dimensional symbolism minus the volume of said supplementary identification data (IPD). When said supplementary IPD is extensive and likely to increase said remaining target volume for image compression, a process 100 according to the invention may include a step 112 for reducing said collected IPD to retain only identification data of interest (IPD'), which will be encoded in the two-dimensional symbol in place of said original IPD.Thus, a method 100 according to the invention may include a preprocessing step 110 to implement such a reduction 112 of the supplementary identification data IPD and / or a selection of a region of interest PI0' from a reference image PI0. The encoding step 180 of the compressed image PI2s in the form of a two-dimensional symbol 20 is also adapted so that said two-dimensional symbol 20 jointly encodes said original identification data IPD or reduced identification data IPD' and said compressed image PI2s selected in step 170.

[0092] As illustrated in Figures 8 and 9, the invention enables the inclusion of a compressed image in a graphic symbol, for example a two-dimensional one, while preserving its identifying power and improving its visual quality. This is achieved through automatic and dynamic optimization that is a function of the relevant reference image PI0, the compression quantization, and the resampling rate to lower dimensions of said reference image before compression. The resampling provided by the invention reduces image compression artifacts in exchange for an acceptable degree of blur when the image included in a two-dimensional symbol is rendered.

[0093] As shown in Figure 7, the invention further relates to a method 200 for producing a digital or physical identity document 10, said method 200 being designed to be implemented by an information processing unit 31 of an identity document production system 30, such as the system 30 shown in Figure 4, after adaptation of said unit 31 to implement a method 100 for encoding biometric data using two-dimensional symbolism, the maximum encoding capacity of each two-dimensional symbol 20 being predetermined and also conforming to the invention. Such a method 200 comprises:

[0094] - a step of digitizing biometric data and producing an image representing said biometric PI data, which may characterize a face, one or more fingerprints or an iris;

[0095] - a triggering step for the implementation of a 100 method for encoding biometric data represented by said PI image produced, under a two-dimensional symbolism, said method conforming to any of the embodiment variants according to the invention;

[0096] - a step 230 of integration of the two-dimensional symbol 20 generated by said implementation of said process 100 of biometric data encoding and production of said identity document 10.

[0097] As previously mentioned, in order to combine PI0 biometric data and complementary IPD identification data to produce a two-dimensional symbol 20, such a process 200 includes a digitization step 210 of identification data and production of a set of (possibly alphanumeric) IPD identification data.

[0098] To adapt the operation of an identity document production system 30, a program memory 31M of the information processing unit 31 of said system 30 (or any other storage medium readable by said information processing unit 31) stores program instructions executable by said information processing unit 31 of a suitable computer program product. Such program instructions are arranged so that, after being loaded or written to said memory 31M, their execution by said information processing unit 31 triggers the implementation of a method 100 for encoding biometric data using a symbology and / or producing an identity document 200 in accordance with the invention.

[0099] The invention has been described through various configurations of a physical or digital identity document and cannot be limited to this single example. The same applies to two-dimensional symbolism such as two-dimensional barcodes.

[0100] The invention would find full application in meeting the need to encode biometric data in the form of a graphic symbol without the said biometric data losing too much of its similarity with respect to a reference.

Claims

33 DEMANDS 1. A method (100) for encoding biometric (PI) data using two-dimensional symbolism, the maximum encoding capacity of each two-dimensional symbol (20) being predetermined, said method (100) being designed to be implemented by an information processing unit (31) of an identity document (10) production system (30), said method comprising: - a step (101, 102) of collecting biometric data (PI) in the form of a reference image (PI0) of predetermined dimensions; - a step (140) of producing a compressed image (PI2i) from said reference image (PI0); - a step (180) of encoding said compressed image (PI2i) in the form of a two-dimensional symbol (20); said process (100) being characterized in that it comprises: - a subsampling step of the reference image (PI0) according to a determined resampling rate (TÏ) to produce a first resampled image (PI1i) prior to the implementation of the step (140) of producing a compressed image (PI2i), said step (140) of producing a compressed image (PI2i) consisting of the compression of said first resampled image (PI1 i); - a step (150) of upsampling the compressed image (P I2i) to produce a second resampled image (P I3i) of the same dimensions as the reference image (PI0); - a step of producing a similarity measure (SSIMi) between the second resampled image (PI3i) and the reference image (PI0); 34 the encoding step (180) of the compressed image (PI2i) in the form of a two-dimensional symbol (20) is only implemented if (120y) the compressed image (PI2i, PI2s) satisfies (170) a determined selection criterion based on the similarity measure (SSIMi) of the second resampled image (PI3i) associated with it.

2. Method (100) according to claim 1, comprising a step of determining (120) the resampling rate (T1) used to produce a first resampled reference image (PI1 i).

3. A method according to claim 2, wherein the iterations are sequentially repeated until an iteration criterion is satisfied: - the step of determining (120) the resampling rate (TÏ); - the step of subsampling the reference image (PI0); - the step (140) of producing a compressed image (PI2i); - the step (150) of upsampling said compressed image (PI2i) to produce a second resampled image (PI3i); - the step (160) of similarity measurement (SSIMi) between the second resampled reference image (PI3i) and the reference image (PI0).

4. A method (100) according to claim 3, wherein the iteration criterion is satisfied as soon as a compressed image (PI2i) satisfies (170) a determined selection criterion which is based on the similarity measurement of the second resampled image (PI3i) associated with it, said method (100) therefore comprising a selection step (170) of said compressed image for the subsequent implementation of the encoding step (180) of the selected compressed image (PI2s) in the form of a single two-dimensional symbol (20).

5. Method (100) according to claim 3, wherein: - the compressed image (PI2i) and the similarity measure (SSIMi) between the second resampled image (PI3i) and the reference image (PIO) produced at each of said iterations are recorded as a data set ({Pi2i}) in a data memory (31 M) of the information processing unit (31 ); - the iteration criterion consists of a minimum number of iterations to be reached; - said process (100) comprises, to satisfy said iteration criterion, a selection step (170) of the compressed image (PI2s) for the subsequent implementation of the encoding step (180) of the selected compressed image (PI2s) in the form of a single two-dimensional symbol (20) consisting of: o reading from the data memory (31 M) of the data sets ({Pi2i}*) comprising respectively the compressed images (PI2i) and the similarity measures (SSIMi) of the second resampled reference images (PI3i) respectively associated; o selecting the compressed image (PI2i) associated with the best of the similarity measures (SSIMi) with regard to those respectively associated with the compressed images included in said data sets ({PI2i}*) read.

6. Method (100) according to any one of claims 2 to 5, wherein the similarity measurement step (160) (SSIMi) consists of calculating a measure (SSIMi) of the visual quality of the second resampled image (PI3i) relative to the reference image (PIO).

7. Method (100) according to any one of the preceding claims comprising a step (110, 111) of determining a region of interest (ROI') within the reference image (PI0) and for which the step of producing a similarity measure (SSIMi) between the second resampled image (PI 3i) and the reference image (PI0) relates only to said region of interest (ROI').

8. Method (100) according to any one of the preceding claims wherein the quantification retained for the implementation of the step (140) of producing a compressed image (PI2i) is determined according to a binary search algorithm such that the volume of compressed data is less than a predetermined target volume which is a function of the maximum encoding capacity of a two-dimensional symbol (20).

9. A method (100) according to any one of claims 3 to 5 wherein the iteration criterion (120) consists of a predetermined number of iterations to be performed and wherein the step of determining (120) the resampling rate (T1) used to produce a first resampled image (P11 i) consists of calculating said resampling rate (T1) so that its value changes at each iteration by a predetermined step within one or more predetermined possible value ranges.

10. Method (100) according to claim 9 wherein the possible ranges of values ​​of the resampling rate (T1) are chosen such that said rate induces an average similarity measure between the second resampled images (PI3i) produced during the iterations and the reference image (PI0) greater than a predetermined threshold. 37 11. Method (100) according to claim 9 or 10, wherein the quantification retained for the implementation of the step (140) of producing a compressed image (PI2i) is established from a history ({PI2s}*) of prior implementations of said method (100) of encoding biometric data under a two-dimensional symbolism by the information processing unit (31) of a system (30) for producing an identity document (10) producing respectively compressed images (PI2s) by an identical or similar compression algorithm from distinct reference images (PI0), having respective dimensions similar to those of the compressed image (PI2i) and the reference image (PI0).

12. A method (100) according to any one of claims 9 to 11, wherein a predetermined range of possible values ​​of the resampling rate (T1) is established from a history ({PI2s}*) of prior implementations of said method (100) of encoding biometric data under a two-dimensional symbolism by the information processing unit (31) of an identity document production system (30) (10) producing respectively compressed images (PI2s) by an identical or similar compression algorithm from distinct reference images (PI0), having respective dimensions similar to those of the compressed image (PI2s) and the reference image (PI0).

13. A method (100) according to claim 8, comprising a step (103) of collecting complementary identification data (CID) to biometric data (BID), and wherein: - the predetermined target volume to retain the quantization used for the implementation of step (140) of compressed image production (PI2i), is calculated to be less than or equal to the 38 maximum encoding capacity of each symbol of the two-dimensional symbolism reduced by the volume of said complementary identification data (CID); - the encoding step (180) of the selected compressed image (PI2s) in the form of a two-dimensional symbol (20) is arranged so that said two-dimensional symbol jointly encodes said complementary identification data (IPD) and said selected compressed image (PI2s).

14. Method (100) according to the preceding claim, comprising a step of reducing (110, 112) the supplementary identification data (SID) collected to retain only supplementary identification data of interest (SID').

15. Method (100) according to any one of the preceding claims, comprising a production step (102) of the reference image (PI0) of predetermined dimensions by lossless compression of a second image (PI) of larger dimensions, said second image (PI) representing said biometric data.

16. A method (200) for producing a digital or physical identity document (10), said method (200) being designed to be implemented by an information processing unit (31) of an identity document production system (30), said system further comprising a biometric data capture device (36, 37) and an output device (38, 34), said method comprising: - a step of digitizing (210) biometric data by the capture device (36, 37) and producing an image representing said biometric data (PI); 39 - a triggering step for the implementation of a method (100) for encoding biometric data represented by said image (PI) produced, under a two-dimensional symbolism, said method conforming to any one of claims 1 to 15; - a step (230) of integrating the two-dimensional symbol (20) generated by said implementation of said process (100) of encoding biometric data in said identity document (10) and / or of printing or displaying said document by the output device (38, 34).

17. Method (200) according to the preceding claim, comprising a step of digitizing (210) additional identification information by the capture device (36, 37) and producing a set of additional identification data (IPD), the method (100) of biometric data encoding the implementation of which is triggered being in accordance with claim 13 or 14.

18. Computer program product comprising one or more program instructions executable by an information processing unit (31) of a system (30) for producing an identity document (10), said program instructions being: - loadable into a memory (31 M) of said information processing unit (31); - designed so that their execution by said information processing unit (31) causes the implementation of a method (100) for encoding biometric data under a two-dimensional symbolism according to any one of claims 1 to 15. 40 19. Computer program product comprising one or more program instructions executable by an information processing unit (31) of a system (30) for producing an identity document (10), said program instructions being: - loadable into a memory (31 M) of said information processing unit (31); - designed so that their execution by said information processing unit (31) causes the implementation of a method for producing a digital or physical identity document (10) according to claim 16 or 17.

20. Computer-readable storage medium containing instructions for a computer program product according to claim 18 or 19.

21. System (30) for producing an identity document (10) comprising: - an information processing unit (31) comprising a memory (31 M) recording the program instructions of a computer program product according to claim 18 or 19; - a biometric data capture device (36, 37); - an output device (38, 34) for printing or displaying an identity document (10).

Images