WO2024009212A1

WO2024009212A1 - Method for cosmetic analysis of a face

Info

Publication number: WO2024009212A1
Application number: PCT/IB2023/056910
Authority: WO
Inventors: Joanna IFERGAN; Frédéric IFERGAN; Dan BENAMARA
Original assignee: Abbi
Priority date: 2022-07-04
Filing date: 2023-07-03
Publication date: 2024-01-11
Also published as: FR3137481A1

Abstract

The invention relates to a method for cosmetic analysis of a face, the method comprising the following steps: - an acquisition step (200) of acquiring at least one image of the face, referred to as the raw image (201); - a pre-processing step of pre-processing the raw image to obtain a calibrated image (203) from the raw image (201); - an analysis step (260) of analysing at least one of the areas of interest (1, 2, 3, 4, 5, 6, 7, 8) of the calibrated image (203), referred to as the analysed area of interest, the analysis step (260) comprising a step (261) of measuring at least one measurand (301, 302, 303, 304, 305, 306, 307) in the analysed area of interest (1, 2, 3, 4, 5, 6, 7, 8); - a step (263) of determining at least one intermediate score (351, 352, 353, 354, 355, 356, 357) for each analysed area of interest (1, 2, 3, 4, 5, 6, 7, 8).

Description

Cosmetic analysis process for a face

The technical context of the present invention is that of products intended to be applied to a person's skin. More particularly, the invention relates to cosmetic products intended to be applied to the skin of a person's face.

The definition of a cosmetic product best suited to a person's skin involves a precise and complete examination of that person's skin in the region of their body to which the cosmetic product is intended to be applied, in order to deduce a certain number of parameters characteristic of this skin. Generally, a visual examination, possibly supplemented by an oral interview with the person concerned, allows the aforementioned parameters to be identified and known: as non-exhaustive examples, visual defects of the skin, feelings of discomfort, etc.

Once this examination has been carried out, the professional offers one or more products and/or mixtures of products that he has defined as being able to meet the needs of the person.

The present invention aims to propose a simple and effective solution making it possible to define an optimal composition of one or more cosmetic products to be applied to the skin of a person's face based on a precise and complete examination of this skin.

To achieve its aim, the subject of the invention is, according to a first aspect, a method of cosmetic analysis of a face, the method comprising the following steps:
- a step of acquiring at least one image of the face, called raw image,

- a step of storing the raw image in a database kept on a memory,

- a step of preprocessing the raw image, the preprocessing step comprising a step of excluding at least one exclusion zone of the raw image and a step of delimiting several zones of interest of the raw image, the preprocessing step making it possible to obtain a calibrated image from the raw image,

- a step of analyzing at least one of the zones of interest of the calibrated image, called analyzed zone of interest, the analysis step comprising a step of measuring at least one measurand on the zone d interest analyzed,

- a step of determining at least one intermediate score for each area of interest analyzed.

It should be understood that the invention provides that the acquisition step is carried out by an optical sensor, for example a photographic sensor or a video sensor. More precisely, in the case of a video sensor, the raw image is obtained from at least one image coming from a video stream recorded by the video sensor.

Advantageously, the optical sensor used to carry out the acquisition step is associated with a mirror allowing a user to view their face. Advantageously, the optical sensor, whether it is a photographic sensor or a video sensor, is incorporated in such a mirror, preferably at the level of an upper edge located beyond, with reference to a vertical axis , from a line of the user's eyes.

Furthermore, it must also be understood here that by measurand we mean a characteristic parameter of the facial skin of the user, that is to say a quantified indicator representative of a visual characteristic of the facial skin of this user: as a non-exclusive example, a measurand can be defined to represent a presence or absence of wrinkles on the skin of the user's face.

According to an example of implementation of the invention, the database in which the raw image is stored is located locally, near the optical sensor. According to another example of implementation of the invention, this database is located remotely, for example on a remote server. In this case, the invention provides that the optical sensor is connected to such a remote server by any means of communication, wired or non-wired.

According to a characteristic of the method according to the invention, the preprocessing step comprises a step of calibrating the raw image, the raw image being resized so that the face contained in said raw image, delimited by the contours of the face, is contained in a rectangle of predetermined dimensions, preferably equal to 900 pixels on each side.

At the end of this calibration step, the recorded raw images therefore have identical digital dimensions, thus making the measurands determined on each area of interest during the analysis step comparable.

According to the invention, the resizing of the raw image is determined by measuring a spacing between the pupils of the face contained in the raw image.

Advantageously, the preprocessing step includes a step of rotating the raw image, so that an axis between two pupils of the face contained in the raw image is aligned with a row of pixels of the raw image. Preferably, the rotation of the raw image is carried out in such a way that an axis between the two pupils of the face contained in the raw image is aligned with a horizontal row of pixels of the raw image.

The preprocessing step also advantageously includes a step of cutting the raw image, in order to delimit the raw image near the contours of the face contained therein.

In other words, the preprocessing step, as it has just been described, makes it possible to standardize the raw images recorded in the database.

According to another characteristic of the invention, the step of delimiting the zones of interest comprises an identification of several points of interest on the face contained in the standardized raw image at the end of the operations previously described.

More precisely, the points of interest comprise first points of interest whose position on the standardized image is determined, for example, via a neural network previously trained to identify such first points of interest on reference images. As non-exhaustive examples, the first points of interest are located at the level of an eye contour, a bridge of the nose, the nostrils, the contour of the lips and the contour of the face.

The aforementioned points of interest also include second points of interest whose position is calculated relative to the position of the first points of interest. First points of interest and second points of interest thus make it possible, together, to define the zones of interest, previously mentioned, of the standardized image, and it follows from the above that the zones of interest include:

- a first zone of interest located at the level of the forehead of the face contained in the standardized image, the first zone of interest being delimited by the points of interest located around the forehead,

- a second zone of interest located at the level of each eye of the face contained in the standardized image, the second zone of interest being delimited by the points of interest located around each eye,

- a third zone of interest located at the level of the nose of the face contained in the standardized image, the third zone of interest being delimited by the points of interest located around the nose,

- a fourth zone of interest located at the level of the cheeks of the face contained in the standardized image, the fourth zone of interest being delimited by the points of interest located around each cheek of the face,

- a fifth zone of interest located between the upper lip and the nose of the face contained in the standardized image, the fifth zone of interest being delimited by the points of interest located between the upper lip and the nose,

- a sixth zone of interest located at the level of the chin of the face contained in the standardized image, the sixth zone of interest being delimited by the points of interest located around the chin,

- a seventh zone of interest located at the level of the mouth of the face contained in the standardized image, the sixth zone of interest being delimited by the points of interest located between the cheeks and the chin, the seventh zone of interest being located in an intermediate position between, on the one hand, the sixth zone of interest and, on the other hand, the fourth zone of interest and the fifth zone of interest,

- an eighth zone of interest located between the contour of the void contained in the standardized image and each eye, the eighth zone of interest being located in an intermediate position between the contour of the face and the second zone of interest.

According to another characteristic of the invention, the preprocessing step comprises, for each zone of interest, a step of excluding parts excluded from the zone of interest of the standardized image, the exclusion step defining the parts excluded from a color deviation and/or a contrast deviation and/or a brightness deviation measured on the excluded part in relation to, respectively, an average color and/or contrast value and/or brightness determined on the corresponding area of interest or on several areas of interest.

In other words, the exclusion step comprises a step of comparison between, on the one hand, the color and/or the contrast and/or the brightness of the excluded part and, on the other hand, respectively, the average value of color and/or contrast and/or brightness determined on the corresponding area of interest or on several areas of interest. If the difference observed is greater than a reference value, called the exclusion threshold, then the excluded part is removed from the zone of interest. It should be noted that different reference values, or threshold values, are advantageously defined beforehand for the various aforementioned optical parameters: color, contrast, brightness.

In other words, the exclusion step makes it possible to exclude, for the measurements carried out subsequently during the analysis step, one or more regions of the face which present, for example for reasons of exposure during the shooting, optical information which can induce measurement artifacts which in turn induce one or more biases in the analysis step. It must therefore be clearly understood here that these exclusion zones are defined solely on the basis of optical parameters of the shooting and recording of the image, and not on a form of averaging of possible imperfections or characteristics of the skin this image may contain.

We therefore have, at the end of the preprocessing step, a standardized image of a user's face, a standardized image in which are defined, on the one hand, several areas of interest intended to be analyzed to define the desired cosmetic profile, and, on the other hand, several exclusion zones defined so that the analysis of the standardized image is as relevant as possible, without possibly biasing the shot itself. This standardized image in which areas of interest and exclusion zones are defined will be referred to in the following as a calibrated image.

Advantageously, the analysis step comprises a step of comparing each area of interest analyzed with reference images stored in a reference database, the comparison step making it possible to quantify differences between each at least one measurand of the area of interest analyzed with a reference quantity determined from the reference images.

According to other characteristics of the invention:

– the analysis step comprises a step of analyzing wrinkles on at least one of the zones of interest, the step of analyzing wrinkles comprising a step of detecting wrinkles present on each at least one zone of interest by measuring point contrast variations on the area of interest considered, in order to determine at least one continuous trace of said contrast variations greater than a contrast threshold. According to one example, the continuous line characterizing each wrinkle detected in the area of interest considered is determined on the basis of a Frangi filter and implements Hessian type matrices. Advantageously, the zones of interest more particularly concerned by this wrinkle analysis step include at least the first zone of interest and/or the eighth zone of interest previously defined.

– the measuring step comprises a step of measuring wrinkles comprising a step of measuring a length and/or a width and/or a depth of each at least one continuous line previously determined and representative of at least one least one wrinkle detected in one of the areas of interest.

– the step of determining at least one intermediate score comprises a step of determining an intermediate score of wrinkles comprising a step of counting a number of wrinkles having a thickness greater than a second threshold, preferably equal to 0, 9mm.

More generally, the invention advantageously provides that the step of determining the intermediate wrinkle score comprises:

– a step of counting a number of wrinkles having a thickness less than a first threshold, preferably equal to 0.5 mm, a number of wrinkles having a thickness greater than a second threshold, preferably equal to 0.9 mm , and a number of wrinkles having a thickness between the first threshold and the second threshold,

- a step of determining an average length of the wrinkles,

- a step of determining a maximum length of the wrinkles,

- a step of determining an average surface area of the wrinkles,

- a step of determining a maximum surface area of the wrinkles.

The method according to the invention thus makes it possible to carry out, on the face of a user, a complete and precise analysis of the skin of this user, in connection with the presence of wrinkles.

The method according to the invention also makes it possible to carry out such a complete and precise analysis of the face of such a user, in relation to other measurands of the user's skin.

Thus, according to other characteristics of the invention:

– the analysis step comprises a brightness analysis step on at least one of the zones of interest, the brightness analysis step comprising a step of determining a chroma of a region of analysis over the entire at least one area of interest. The analysis region is here preferably a circular region whose diameter is less than 10 pixels, preferably equal to 3 pixels. The zones of interest more particularly concerned by this brightness analysis step include at least the first zone of interest and/or the third zone of interest and/or the fourth zone of interest and/or the fifth zone of interest and/or the sixth zone of interest and/or the seventh zone of interest and/or the eighth zone of interest previously defined.

– the measurement step comprises a brightness measurement step comprising a step of determining a standard deviation of the chroma values determined for each analysis region covering the entire at least one area of interest.

– the step of determining at least one intermediate score comprises a step of determining an intermediate brightness score determined by a percentage of the analysis regions of each zone of interest having a standard deviation greater than a threshold value .

According to other characteristics of the process according to the invention:

– the analysis step comprises a roughness analysis step on at least one of the zones of interest, the roughness analysis step comprising a step of detecting the skin grains present on each at least one of the zones of interest by measurement, the skin grains being detected by measuring the variations in point contrast on the area of interest considered in order to determine at least one closed contour of said variations in contrast greater than a contrast threshold. Advantageously, the contrast threshold is preferably between 0.15 and 0.30. The areas of interest more particularly concerned by this roughness analysis step include all or part of the areas of interest of the calibrated image, with the exception of the second area of interest.

– the measurement step includes a roughness measurement step comprising a step of counting the number of closed contours determined previously.

– the step of determining at least one intermediate score comprises a step of determining an intermediate roughness score comprising a step of determining a ratio between a total surface of the closed contours and a total surface of the face contained in the calibrated image.

According to other characteristics of the invention for another measurand:

– the analysis step comprises a redness analysis step on at least one of the areas of interest, the redness analysis step comprising a step of detecting a shade of each pixel of each at least one areas of interest by measurement. The areas of interest more particularly concerned by this redness analysis step preferably include the fourth area of interest and/or the seventh area of interest, previously defined, of the calibrated image.

– the measurement step includes a redness measurement step comprising a step of measuring an angular deviation of the hue determined for each pixel relative to a reference color. According to the invention, the selected reference color is pure red, and the angular deviation is determined in the colorimetric space designated by the acronym HSV (Hue, Saturation, Value) of the reference image. More precisely, the invention provides that the shade determined during the redness analysis step corresponds to the dimension H of the aforementioned HSV colorimetric space.

– the step of determining at least one intermediate score comprises a step of determining an intermediate redness score comprising, for each area of interest, a step of counting a number of pixels having a hue located above beyond a threshold angular distance from the reference color, divided by the total number of pixels in the area of interest considered.

According to other characteristics of the invention for another measurand:

– the analysis step comprises a step of analyzing buttons on at least one of the zones of interest, the step of analyzing buttons comprising a step of detecting the buttons present on each at least one of the zones of interest interest via a neural network configured to calculate, for each pixel of the area of interest, a probability of presence of a button, the neural network having been trained to recognize such buttons on reference images presenting buttons. The areas of interest more particularly concerned by this button analysis step include at least the first area of interest and/or the third area of interest and/or the fourth area of interest.

– the measurement step includes a button measurement step comprising a step of determining a percentage of pixels presenting, for a given area of interest, a probability of button presence greater than a reference threshold.

– the step of determining at least one intermediate score comprises a step of determining an intermediate score of buttons comprising a step of calculating an average value of the probability of button presence greater than the reference threshold.

According to other characteristics of the invention for another measurand:

– the analysis step comprises a step of analyzing tasks on at least one of the zones of interest, the step of analyzing tasks comprising a step of detecting the tasks present on each at least one of the zones of interest interest by filtering. The areas of interest more particularly concerned by this task analysis step include at least the first area of interest and/or the third area of interest and/or the fourth area of interest and/or the sixth area of interest. interest and/or the seventh area of interest and/or the eighth area of interest.

– the measurement step includes a task measurement step comprising a step of measuring the number of tasks analyzed and a step of measuring a surface area of each analyzed task.

– the step of determining at least one intermediate score comprises a step of determining an intermediate score of tasks comprising a step of normalizing a total surface of all the tasks analyzed by a surface of all the areas of interest in which at least one task was analyzed.

According to other characteristics of the method according to the invention for other measurands, taken separately or in combination:

- the analysis step comprises a gloss analysis step on at least one of the areas of interest, the gloss analysis step comprising a step of detecting a hue of each pixel of each at least one areas of interest per measurement. The areas of interest more particularly concerned by this gloss analysis step include the fourth area of interest, the sixth area of interest and the seventh area of interest of the calibrated image.

– the measurement step includes a brightness measurement step comprising a step of measuring a distance between the shade determined for each pixel and a reference color. Advantageously, the reference color chosen is a reference white, and the distance is determined in the LAB colorimetric space of the reference image.

– the step of determining at least one intermediate score comprises a step of determining an intermediate gloss score comprising, for each area of interest, a step of counting a number of pixels having a hue located above beyond a threshold distance from the reference color, divided by the total number of pixels in the area of interest considered.

– the method comprises a step of determining an overall score resulting from a weighted combination of each at least one intermediate score.

– the method includes a step of ranking all the intermediate scores calculated previously.

– the analysis method comprises a step of entering information relating to a user of the method, the entry step being carried out via a man-machine interface making it possible to record a profile in the database user including an age and gender of said user.

– the analysis method comprises, during the step of acquiring at least one image, a step of detecting a skin color of the user, the skin color being recorded in the database, in the user profile.

– the method according to the invention comprises a cosmetic formulation step comprising a step of selecting at least one active ingredient and at least one base, the at least one active ingredient and the at least one base together forming the cosmetic formulation, the selection step being carried out with regard to the 3 most important intermediate scores determined previously, the at least one active and the at least one base being selected via a reference table establishing an effectiveness of each base and each asset for each type of analysis carried out.

– the at least one asset and the at least one base selected during the selection step is determined as a function of the user profile, a quantity of each at least one asset and of each at least one base selected being determined as a function of the user profile and for each of the three most important intermediate scores determined previously.

By the process as just described, the invention achieves the goal it set for itself, by proposing a complete and precise solution, on the one hand, by carrying out a precise analysis of the skin of the face of a user and, on the other hand, by defining a cosmetic product formulation specifically adapted according to the results of the aforementioned analysis.

According to a second aspect, the invention also relates to a cosmetic recommendation system, comprising at least one mirror, an optical image taking device, and a control unit comprising a microcontroller configured to implement the method as described. previously.

According to different examples, the optical camera can be a photographic device or a camera.

Other characteristics and advantages of the invention will appear further through the description which follows on the one hand, and several examples of embodiment given for informational and non-limiting purposes with reference to the appended schematic drawings on the other hand, in which :

schematically illustrates a cosmetic recommendation system according to the invention,

schematically illustrates the different stages of the process according to the invention,

schematically illustrates the different areas of interest defined from the first points of interest as defined as illustrated by the ,

schematically illustrates the locations of the first points of interest, previously described, defined on a standardized image obtained from a cosmetic recommendation system such as that illustrated by the .

Of course, the characteristics, variants and different embodiments of the invention can be associated with each other, in various combinations, to the extent that they are not incompatible or exclusive of each other. It will be possible in particular to imagine variants of the invention comprising only a selection of characteristics described subsequently in isolation from the other characteristics described, if this selection of characteristics is sufficient to confer a technical advantage or to differentiate the invention from to the prior art.

In particular, all the variants and all the embodiments described can be combined with each other if nothing opposes this combination on a technical level.

On the FIGURES, elements common to several FIGURES retain the same reference.

In reference to the , a cosmetic recommendation system 100 according to the invention comprises a mirror 101 equipped with a camera 102 configured to capture one or more images of the face of a user 150 placed in front of the mirror 101. According to different implementation examples implemented, the camera 102 can be a photographic device or a camera configured to capture a video stream of the face of the user 150.

Whatever the mode of implementation of the cosmetic recommendation system 100, the invention provides that the camera device 102 is advantageously placed beyond, in a vertical direction V, a line connecting the pupils together. eyes of the user 150. In other words, the invention provides that the camera device 102 is advantageously placed substantially facing a region covering the eyebrows and/or the forehead of the user 150.

In the cosmetic recommendation system according to the invention, the camera device 102 is connected to a control unit 103 comprising, on the one hand, a unit 104 configured to record and store the raw images captured by the camera device. views 102, and, on the other hand, a calculation unit 105 configured to implement the different stages of the method according to the invention.

According to different examples of implementation of the invention, the control unit 103 can be placed in the vicinity of the mirror 101 carrying the camera device 102, or it can be placed remotely, for example on a remote server. In all cases, the connection between the camera device 102 and the control unit 103 can be made by wired or wireless means.

There schematically illustrates the different stages of the process according to the invention as described above.

As indicated previously, in a first step 200 of the method according to the invention, a raw image 201 of the face of a user 150 is captured by the camera device 102. According to different examples, the raw image 201 can be a photograph or an image extracted from a video stream recorded by the camera device 102.

In a second step 210 of the method according to the invention, the raw image 201 is recorded in the storage unit 104 of the control unit 103, storage unit 104 from which the raw image 201 is then processed by the calculation unit 105.

In a third step 220 of the method according to the invention, the raw image 201, recorded in the storage unit 104, is first resized so that the face of the user 150 is contained in a digital frame whose dimensions are predefined. Preferably, such a digital frame is square, and its sides have a dimension equal to 900 pixels.

More precisely, the invention provides that the resizing operation 220 comprises, on the one hand, a measurement of a spacing between the pupils of the eyes, and, on the other hand, a rotation of the raw image 201 of such so that the axis for measuring the spacing between the pupils of the user's face 150 is aligned with a row, preferably horizontal, of pixels of the digital frame in which the resized image must fit.

By resizing each raw image 201 captured by the camera 102 so that it is contained in such a square, the invention achieves a form of normalization of these raw images. The image obtained at the end of this operation is therefore a standardized image 202.

The next step 230 of the method according to the invention is an operation of defining a set of points and areas of interest on the standardized image 202 obtained at the end of the resizing step 220.

More precisely, according to the invention, these points and zones of interest comprise first points of interest 10a, ..., 10i, ..., 10n, illustrated on the . Advantageously, these first points of interest are identified on the standardized image 202 by means of a device of the type of a neural network embedded in the control unit 103 previously described and previously trained to identify such first points of interest. interest on reference images. The positions of the first points of interest 10a, …, 10i, …, 10n, are therefore defined by optical identification on the standardized image 202.

According to the example more particularly illustrated by the , these first points of interest are 68 in number, and together they define a form of "silhouette-model" of the face contained in the standardized image 202, this form of "silhouette-model" comprising, on the one hand, an evocation of an outline of the face, and, on the other hand, an evocation of the eyes, the nose and the mouth of this face.

The invention further provides that the operation of defining the points and zones of interest comprises the definition of a set of second points of interest 11a, ..., 11i, ..., 11n, not shown on the . According to the invention, the positions of the second points of interest 11a, ..., 11i, ..., 11n, are defined by calculation, from the positions of the first points of interest 10a, ..., 10i, ..., 10n.

The invention further provides an operation of defining a set of zones of interest on the face of the user 150. These zones of interest are, according to the embodiment of the invention more specifically described in the present document, non-exclusive, 8 in number, and they are schematically represented on the :

- a first zone of interest 1 located at the level of the forehead of the face contained in the calibrated image, the first zone of interest being delimited by the points of interest located around the forehead,

- a second zone of interest 2 located at the level of each eye of the face contained in the calibrated image, the second zone of interest being delimited by the points of interest located around each eye,

- a third zone of interest 3 located at the level of the nose of the face contained in the calibrated image, the third zone of interest being delimited by the points of interest located around the nose,

- a fourth zone of interest 4 located at the level of the cheeks of the face contained in the calibrated image, the fourth zone of interest being delimited by the points of interest located around each cheek of the face,

- a fifth zone of interest 5 located between the upper lip and the nose of the face contained in the calibrated image, the fifth zone of interest being delimited by the points of interest located between the upper lip and the nose,

- a sixth zone of interest 6 located at the level of the chin of the face contained in the calibrated image, the sixth zone of interest being delimited by the points of interest located around the chin,

- a seventh zone of interest 7 located at the level of the mouth of the face contained in the calibrated image, the sixth zone of interest being delimited by the points of interest located between the cheeks and the chin, the seventh zone of interest being located in an intermediate position between, on the one hand, the sixth zone of interest and, on the other hand, the fourth zone of interest and the fifth zone of interest,

- an eighth zone of interest 8 located between the contour of the void contained in the calibrated image and each eye, the eighth zone of interest being located in an intermediate position between the contour of the face and the second zone of interest.

At the end of the resizing step 220 and the step 230 of defining the points and zones of interest, the face of the user 150 is thus modeled by a precise geometric structure, defined on the basis of the different points of interest 10a, …, 10n, and 11a, …, 11n, determined both by measurement on the recorded image of the face and by calculation from this image.

To further improve the precision and representativeness of the image thus obtained and reworked in relation to the raw image 201 captured by the camera 102, the method according to the invention comprises a step 240 of excluding certain areas of the standardized image 202. The main purpose of this operation is to exclude from the standardized image 202, in which the different zones of interest 1, ..., 8, have been defined, regions whose in-depth optical analysis could lead to an error or bias in diagnosing the condition of the skin of the user's face 150. To do this, the invention provides, on the one hand, an analysis of the color and/or contrast and /or the brightness of the image in the different areas of interest 1, ..., 8, in order to deduce average values of color and/or contrast and/or brightness, and, on the other hand, the exclusion of regions from the zones of interest 1, ..., 8, in which a color and/or a contrast and/or a brightness present a deviation from the aforementioned average values greater than a previously defined threshold. It must be understood here that the excluded areas are defined on the basis of the parameters of the shooting, and not on the basis of the detection of possible imperfections which the process would like to overcome in order to carry out the desired cosmetic analysis. In other words, if an excluded zone can result from a flash effect generated on the surface of the mirror 101 by a parasitic reflection at the time of taking pictures, it cannot result from a characteristic of the skin such than, for example, an isolated mole.

Step 220 of resizing the raw image 201 to deduce the standardized image 202, step 230 of defining the points and zones of interest 1, ..., 8, and step 240 of defining the zones d exclusion, constitute the calibration step 250, previously described, of the method according to the invention, at the end of which a calibrated image 203 of the face of the user 150 is obtained. As a reminder, the calibrated image 203 is the standardized image 202 on which the zones of interest 1, … 8, as well as the exclusion zones have been defined.

The analysis of the calibrated image 203 comprises, according to the invention, the analysis of a certain number of measurands in the different zones of interest 1, ..., 8. As indicated previously, each measurand is a parameter whose value is representative of a characteristic of the user's facial skin taken into account for the desired cosmetic recommendations.

In accordance with the method according to the invention as described in this document, several measurands are advantageously taken into account:

– a first measurand 301 relates to the presence of wrinkles on the skin of the user's face,

– a second measurand 302 relates to skin radiance,

– a third measurand 303 is representative of roughness of the skin,

– a fourth measurand 304 relates to redness of the skin,

– a fifth measurand 305 is representative of the presence of pimples on the skin,

– a sixth measurand 306 is representative of the presence of spots on the skin,

– a seventh measurand is representative of skin shine.

And, for each aforementioned measurand, the invention provides that the analysis step 260 comprises:

– a step 261 of detecting the measurand considered,

– a step 262 of measuring the measurand considered,

– a step 263 of calculating and assigning an intermediate score to the measurand considered.

Depending on the measurand considered, the step 261 of detecting the measurand can include different operations and relate to different zones of interest 1, ..., 8 previously defined. So :

– for the first measurand 301 representative of the presence of wrinkles, the detection step 261 implements a measurement of contrast variations and aims to determine at least one continuous line along which the contrast variations are greater than one previously defined threshold, such a continuous trace being representative of a wrinkle. According to the invention, such a continuous trace determined on the basis of a Frangi filter and implements Hessian matrices, and the zones of interest more particularly concerned by the detection step 261 comprise at least the first zone d interest 1, located at the level of the forehead of the user 150, and/or the third zone of interest 3 located on the nose of the user 150 (in particular the upper area of the nose, between the eyebrows), and/or the eighth zone of interest 8, located on the outer contour of the eyes.

– for the second measurand 302 representative of a shine of the skin, the detection step 261 comprises an operation of defining a chroma of a preferably circular analysis region, the diameter of which is less than 10 pixels, preferably of the order of 3 pixels, and the areas of interest more particularly concerned by the detection step 261 advantageously include all the areas of interest with the exception of the second area of interest 2 located around the eyes.

– for the third measurand 303 representing a roughness of the skin, the detection step 261 implements a contrast variation measurement in order to determine at least one closed contour within which the contrast variations are greater than a threshold previously defined, for example between 0.15 and 0.30, and the zones of interest more particularly concerned by the detection step 261 advantageously include all the zones of interest with the exception of the second zone of interest 2 located around the eyes.

– for the fourth measurand 304 representative of a redness of the skin and for the seventh measurand 307 representative of a shine of the skin, the detection step 261 comprises an operation of detecting the shade of each pixel of the zone of interest considered. For the fourth measurand 304, the zones of interest more particularly concerned by the detection step 261 include at least the fourth zone of interest 4, located at the level of the cheeks of the user 150, and/or the sixth zone of interest 6 located at the level of the chin of the user 150, and/or the seventh zone of interest 7, located at the level of the mouth of the user 150. For the seventh measurand 307, the zones of interest more particularly concerned by the detection step 261 include at least the fourth zone of interest 4, located at the level of the cheeks of the user 150, and/or the sixth zone of interest 6 located on the chin of the user 150 ( in particular the upper area of the nose, between the eyebrows), and/or the seventh area of interest 7, located on and around the mouth of the user 150.

– for the fifth measurand 305 representative of the presence of buttons, the detection step 261 advantageously implements a neural network previously trained to detect buttons on a set of reference images, and the areas of interest more particularly concerned by the detection step 261 comprise at least the first zone of interest 1, located at the level of the forehead of the user 150, and/or the third zone of interest 3 located on the nose of the user 150 (in particular the upper area of the nose, between the eyebrows), and/or the fourth area of interest 4, located at the level of the user's cheeks 150.

– for the sixth measurand 306 representative of the presence of spots on the skin, the detection step 261 implements a filtering operation of the calibrated image 203 to deduce the presence of spots, and the areas of interest more particularly concerned by the detection step 261 advantageously include all the zones of interest with the exception of the second zone of interest 2, located around the eyes, and the fifth zone of interest 5, located between the bottom of the nose and the upper lip of the user 150.

Likewise, depending on the measurand considered, the measurement step 262 varies. So :

- for the first measurand 301, the measurement step 262 comprises an operation of measuring a length and/or a width and/or a depth of each at least one continuous trace previously described, detected during the 'preliminary detection step 261,

– for the second measurand 302, the measurement step 262 includes a determination of a standard deviation of the chroma values determined for each analysis region on the zone(s) of interest considered,

– for the third measurand 303, the measurement step 262 includes an operation of counting the closed contours detected during the detection step 261,

– for the fourth measurand 304, the measurement step 262 includes an operation of measuring an angular deviation, in the HSV colorimetric space of the calibrated image 203, of the hue determined for each pixel during the step of detection 261 compared to a pure red.

– for the fifth measurand 305, the measurement step 262 includes a determination of a percentage of pixels presenting, for a given area of interest, a probability of presence of a button greater than a previously defined threshold.

– for the sixth measurand 306, the measurement step 262 includes an operation of counting the tasks detected during the detection step 261, as well as an operation of measuring a surface of the detected tasks.

– for the seventh measurand 307, the measurement step 262 includes an operation of measuring a distance, in the LAB colorimetric space of the calibrated image 203, between the hue determined for each pixel during the detection step 261 and a reference blank.

Finally, step 263 of calculating and assigning an intermediate score to each measurand takes on various aspects depending on the measurand considered:

– for the first measurand 301, the step 263 of calculating and assigning an intermediate score 351 advantageously comprises an operation of counting wrinkles whose width is greater than a first threshold, preferably equal to 0.9 millimeters, an operation of counting wrinkles whose width is less than a second threshold, preferably equal to 0.5 millimeters, an operation for counting wrinkles whose width is between the first threshold and the second threshold mentioned above, as well as various operations for calculating a length maximum, an average length, a maximum surface and an average surface of the wrinkles detected during the detection step 261 and measured during the measurement step 262,

– for the second measurand 302, the step 263 of calculating and assigning an intermediate score 352 includes a determination of a percentage of the regions analyzed in each area of interest concerned having a standard deviation greater than a previously defined threshold ,

- for the third measurand 303, the step 263 of calculating and assigning an intermediate score 353 comprises a determination of a ratio between a total surface area of the closed contours measured during the measurement step 261 and a total surface area of the face contained in the calibrated image 203,

- for the fourth measurand 304 and for the seventh measurand 307, the step 263 of calculating and assigning an intermediate score 354, 357, includes an operation of counting the pixels whose hue is located beyond a threshold distance previously defined in relation, respectively, to the reference pure red or pure white, as well as an operation of calculating a ratio between the number of pixels thus determined and the total number of pixels in the zone of interest considered ,

- for the fifth measurand 305, step 263 of calculating and assigning an intermediate score 355 includes an operation of calculating an average value of probability of presence of a button greater than a previously defined reference threshold,

- for the sixth measurand 306, the step 263 of calculating and assigning an intermediate score 356 includes an operation of calculating a total surface area of the detected and measured tasks, as well as an operation of calculating a ratio between said total surface and a total surface of all areas of interest in which at least one task has been detected and measured.

At the end of the analysis step 260 as it has just been described, we therefore have, for each measurand 301, ..., 307, an intermediate score 351, ..., 357, representative of a number of occurrences, a frequency and a location, on the face of the user 150, of the skin characteristics represented by these measurands.

The next step of the method according to the invention is a step 270 of determining an overall score 400 of the facial skin of the user 150. More precisely, the overall score 400 is a weighted combination of the intermediate scores 351, ... , 357, assigned to each measurand 301, …, 307. Advantageously, different general information relating to the user 150 can be taken into account in such weighting: as non-exhaustive examples, we can cite an age of the user 150, a color of his skin, the gender of the user 150, a known skin disease, the existence of skin allergies, etc. Advantageously, this information is entered by the user himself or by a third party authorized person in a database bringing together different user profiles of the system according to the invention.

In a following step 280, the method according to the invention provides a ranking of the intermediate scores 351, ..., 357, previously calculated, in order to define the three highest intermediate scores.

Finally, in a final step 290, the method according to the invention comprises a step of formulating a cosmetic 500 comprising at least one active ingredient and at least one base chosen, on the basis of the three highest intermediate scores mentioned above, in a reference table establishing the effectiveness of each base and each asset for each measurand 301, …, 307 analyzed. It should be noted that the at least one base and the at least one asset are also advantageously selected on the basis of the general information contained in the user profile previously mentioned.

The invention, as it has just been described, therefore makes it possible to define, for each user, a personalized cosmetic formulation adapted to the characteristics and specific problems of their facial skin.

Of course, the invention is not limited to the examples which have just been described and numerous adjustments can be made to these examples without departing from the scope of the invention. In particular, the different characteristics, variants and embodiments of the invention can be associated with each other in various combinations to the extent that they are not incompatible or exclusive of each other. In particular, all the variants and embodiments described above can be combined with each other. Furthermore, the measurands mentioned in this document are not cited in a limiting or exhaustive manner, and other measurands can be taken into account, without harming the invention.

Claims

Method for cosmetic analysis of a face, the method comprising the following steps:
- a step (200) of acquiring at least one image of the face, called raw image (201);
- a step of storing the raw image (201) in a database stored on a memory (104);
- a step of preprocessing the raw image, the preprocessing step comprising a step of excluding at least one exclusion zone of the raw image (201) and a step (230) of delimiting several zones of interest (1, 2, 3, 4, 5, 6, 7, 8) of the raw image (201), the preprocessing step making it possible to obtain a calibrated image (203) from the image raw (201);
- a step (260) of analyzing at least one of the zones of interest (1, 2, 3, 4, 5, 6, 7, 8) of the calibrated image (203), called zone of interest analyzed, the analysis step (260) comprising a step (261) of measuring at least one measurand (301, 302, 303, 304, 305, 306, 307) on the zone of interest (1, 2 , 3, 4, 5, 6, 7, 8) analyzed;
- a step (263) of determining at least one intermediate score (351, 352, 353, 354, 355, 356, 357) for each zone of interest (1, 2, 3, 4, 5, 6, 7 , 8) analyzed;
characterized in that the preprocessing step comprises a step of calibrating the raw image (201), the raw image (201) being resized so that the face contained in said raw image (201), delimited by contours of the face, is contained in a rectangle of predetermined dimensions, preferably equal to 900 pixels on each side.
Method according to any one of the preceding claims, in which the analysis step (260) comprises, for each zone of interest (1, 2, 3, 4, 5, 6, 7, 8), a step ( 240) of excluding parts excluded from the zone of interest (1, 2, 3, 4, 5, 6, 7, 8) of the calibrated image (203), the exclusion step (240) defining the parts excluded from a color difference and/or a contrast difference and/or a brightness difference measured on the excluded part in relation to respectively an average color and/or contrast value and/ or brightness determined on the corresponding zone of interest (1, 2, 3, 4, 5, 6, 7, 8) or on several zones of interest (1, 2, 3, 4, 5, 6, 7, 8).
Method according to any one of the preceding claims, in which the analysis step (260) comprises a step of analyzing wrinkles on at least one of the areas of interest (1, 2, 3, 4, 5, 6 , 7, 8), the wrinkle analysis step comprising a step (261) of detecting wrinkles (301) present on each at least one of the areas of interest (1, 2, 3, 4, 5, 6 , 7, 8) by measuring variations in point contrast on the area of interest considered in order to determine at least one continuous trace of said variations in contrast greater than a contrast threshold.
Method according to any one of the preceding claims, in which the analysis step (260) comprises a step (261) of brightness analysis (302) on at least one of the zones of interest (1, 2, 3, 4, 5, 6, 7, 8), the brightness analysis step (261) comprising a step of determining a chroma of an analysis region over the entire at least one zone d interest (1, 2, 3, 4, 5, 6, 7, 8).
Method according to any one of the preceding claims, in which the analysis step (260) comprises a step (261) of roughness analysis (303) on at least one of the zones of interest (1, 2, 3 , 4, 5, 6, 7, 8), the roughness analysis step comprising a step of detecting the skin grains present on each at least one of the areas of interest by measurement, the skin grains being detected by measurement of point contrast variations on the area of interest considered in order to determine at least one closed contour of said contrast variations greater than a contrast threshold.
Method according to any one of the preceding claims, in which the analysis step (260) comprises a step (261) of analyzing redness (304) on at least one of the areas of interest, the step of redness analysis comprising a step of detecting a shade of each pixel of each at least one of the areas of interest per measurement
Method according to any one of the preceding claims, in which the analysis step (260) comprises a step (261) of analyzing buttons (305) on at least one of the areas of interest, the step of analysis of buttons comprising a step of detecting the buttons present on each at least one of the zones of interest (1, 2, 3, 4, 5, 6, 7, 8) via a neural network configured to calculate, for each pixel of the area of interest, a probability of the presence of a button, the neural network having been trained to recognize such buttons on reference images presenting buttons.
Method according to any one of the preceding claims, in which the analysis step (260) comprises a step (261) of analyzing tasks (306) on at least one of the areas of interest (1, 2, 3 , 4, 5, 6, 7, 8), the task analysis step (261) comprising a step of detecting the tasks present on each at least one of the areas of interest (1, 2, 3, 4, 5, 6, 7, 8) by filtering.
Method according to any one of the preceding claims, in which the analysis step (260) comprises a step (261) of gloss analysis (307) on at least one of the areas of interest (1, 2, 3 , 4, 5, 6, 7, 8), the brightness analysis step (261) comprising a step of detecting a hue of each pixel of each at least one of the areas of interest (1, 2, 3, 4, 5, 6, 7, 8) per measure.
Method according to any one of the preceding claims, in which the method comprises a step (270) of determining an overall score (400) resulting from a weighted combination of each at least one intermediate score (351, 352, 353, 354, 355, 356, 357).
Method according to the preceding claim, in which the method comprises a step (280) of ranking all the intermediate scores (351, 352, 353, 354, 355, 356, 357) calculated previously.
Method according to the preceding claim, in which the method comprises a step (290) of formulating a cosmetic (500) comprising a step of selecting at least one active ingredient and at least one base, the at least one active ingredient and the at least one base together forming the cosmetic formulation, the selection step being carried out with regard to the 3 most important intermediate scores (351, 352, 353, 354, 355, 356, 357) previously determined, the at least an active and the at least one base being selected via a reference table establishing an effectiveness of each base and each active for each type of analyzes carried out, that is to say for each measurand ( 301, 302, 303, 304, 305, 306, 307).