SYSTEM AND METHOD FOR THE ANALYSIS OF IMAGES OF TATTOOS ON THE HUMAN BODY

Abstract

The invention relates to a system and method for analyzing images of tattoos on the human body. The technical problem concerns the need for in-person consultations or costly and potentially inaccurate manual assessments to determine the cost of tattoo removal. The claimed invention makes the process of estimating the cost of tattoo removal on the human body faster and more efficient, and avoids the need for in-person consultations or costly and potentially inaccurate manual assessments. Figure to be published with the abstract: Figure 1

Description

Title of the invention: SYSTEM AND METHOD FOR THE ANALYSIS OF IMAGES OF TATTOOS ON THE BODY HUMAN Technical field

[0001] The invention relates to the field of information and communication technologies, and more specifically to the field of image processing and machine learning. Prior art

[0002] Tattoo removal is a common practice, but the process of determining the cost of this service can present significant challenges.

[0003] Currently, determining the cost of tattoo removal often requires in-person consultations or manual assessments, which can be expensive and potentially inaccurate.

[0004] In current practice, tattoo removal professionals must examine the tattoo in person, assessing its size, color, density, as well as the part of the body on which it is located, among other factors.

[0005] Based on these observations, they provide an estimate of the cost of tattoo removal.

[0006] Despite these techniques, problems can arise if the assessment is inaccurate or if the estimated cost differs significantly from the actual cost.

[0007] If these inaccuracies are discovered during the removal process, this can lead to delays and additional costs for the patient.

[0008] Additionally, this method also requires the patient to come in person for a consultation, which can be a barrier for those who are located far from the tattoo removal professional's office or have time constraints.

[0009] Therefore, a key objective of the present application is to introduce a novel process that eliminates the need for in-person consultations or manual assessments for determining the cost of tattoo removal. Summary of the invention

[0010] The invention aims to resolve, at least partially, this need.

[0011] The invention relates in particular to a system for generating a segmented image of a tattoo to be removed in a non-therapeutic manner.

[0012] In particular, the system comprises a user device and one or more server devices.

[0013] In practice, the user device comprises an application and a first unit of communication.

[0014] The application presents a user interface that is designed to run on the user device so as to, - receiving one or more original two-dimensional images and data associated with the original two-dimensional image, the original dimensional image comprising a tattoo and one or more reference objects, the tattoo covering a surface of a human body part, the reference object comprising one or more scale markings which have a known actual size in one or more dimensions, the scale marking being configured to visually indicate, directly or indirectly, at least the known actual size of the scale marking, and - display, on the user interface, the original two-dimensional image and the data associated with the original two-dimensional image.

[0015] The first communication unit is configured to transmit the original two-dimensional image to a server device via a network.

[0016] The server device comprises a unit for detecting scale in an image, an image segmentation unit, a calculation unit and a second communication unit.

[0017] The scale detection unit in an image is designed to, - detect the presence of a scale marking in the original two-dimensional image, - determine, from the detected scale marking, directly or indirectly, the known actual size of the scale marking, - measure, in the original two-dimensional image, an apparent size of the scale marking, and - calculating a scale ratio, at least from the known actual size of the detected scale marking and the apparent size of the detected scale marking.

[0018] The image segmentation unit is designed to - running a first machine learning model which receives as input the original two-dimensional image and which has been previously trained to separate the tattoo from the rest of the original two-dimensional image and the surface of the human body part, so as to obtain as output a segmented tattoo, - create a first segmented image which includes the segmented tattoo, and - determine, in the first segmented image, one or more visual characteristics of the segmented tattoo.

[0019] The calculation unit is designed to automatically calculate a tattoo removal price, at least on the basis of the scale ratio and the visual characteristic,

[0020] The second communication unit is designed to transmit to the user device, via the network, the data associated with the two-dimensional image original sional and which include the first segmented image and the tattoo removal price.

[0021] In a first embodiment, the scale marking is a visual indicator that is machine readable and encodes the known actual size of the scale marking.

[0022] In a second embodiment, the scale marking is a visual indicator that is directly human readable.

[0023] In a first example of the first embodiment or the second embodiment, the visual characteristic of the segmented tattoo is selected from a number of contours, a number of edges, a number of color areas, a distribution of color areas, an area, a perimeter, or any combination thereof.

[0024] In a second example of the first embodiment or the second embodiment, when the original two-dimensional image is encoded in the Red-Green-Blue, RGB, color space, the visual characteristic of the segmented tattoo is chosen from, - a sum of a blue channel value of all pixels within an outline of the segmented tattoo, divided by a sum of a green channel value and a red channel value of all pixels within said outline of the segmented tattoo, - a sum of the green channel value of all pixels within the segmented tattoo outline, divided by a sum of the blue channel value and the red channel value of all pixels within said segmented tattoo outline, - a sum of the red channel value of all pixels within the segmented tattoo outline, divided by a sum of the blue channel value and the green channel value of all pixels within said segmented tattoo outline, or any combination thereof.

[0025] In a third example of the first embodiment or the second embodiment, when the original two-dimensional image is encoded in the hue-saturation-brightness, HSL, color space, the visual characteristic of the segmented tattoo is chosen from, - a sum of the Saturation channel value of all pixels within a segmented tattoo outline, divided by a sum of the Hue channel value and the Brightness channel value of all pixels within the segmented tattoo outline, - a sum of the Brightness channel value of all pixels within the segmented tattoo outline, divided by a sum of the Saturation channel value and the Hue channel value of all pixels within the segmented tattoo outline, - a sum of the Hue channel value of all pixels within the segmented tattoo outline, divided by a sum of the Brightness channel value and the Saturation channel value of all pixels within the segmented tattoo outline, or any combination thereof.

[0026] In a third embodiment, the image segmentation unit is further configured to, - running a second machine learning model that receives as input the original two-dimensional image and the first segmented image, and that has been previously trained to separate the human body part surface from the rest of the original two-dimensional image and the segmented tattoo, so as to obtain as output a segmented human body part surface, - create a second segmented image that includes the segmented human body part surface, and - executing a third machine learning model which receives as input the second segmented image and which has been previously trained to classify the segmented human body part surface, so as to obtain as output one class from a plurality of classes, each class corresponding to a human body element, - the calculation unit being further designed to automatically calculate the tattoo removal price on the basis of the class.

[0027] In a fourth embodiment, - the application is further designed to run on the user device so as to allow a user to select, in the original two-dimensional image, one or more pixels which have characteristics corresponding to the characteristics of the surface of the human body part not covered by the tattoo, so as to generate colorimetric information which is representative of the colorimetry of the selected pixel, and - the first machine learning model and / or the second machine learning model are further designed to receive colorimetric information as input.

[0028] In a fifth embodiment, - the server device further comprises an image perspective correction unit which is configured to perform perspective correction to the original two-dimensional image so as to output a corrected two-dimensional image, and - the first machine learning model and / or the second machine learning model are further designed to use the corrected two-dimensional image instead of the original two-dimensional image.

[0029] In a sixth embodiment, - the server device further comprises a multi-level thresholding unit which is further configured to apply multi-level thresholding to the original two-dimensional image or to the first segmented image or to the second image segmented so as to obtain as output a thresholded two-dimensional image, and - the first machine learning model, the second machine learning model and the second machine learning model are further configured to use the thresholded two-dimensional image, respectively, in place of the original two-dimensional image, the first segmented image and the second segmented image. Brief description of the drawings

[0030] Other characteristics and advantages of the invention will be better understood on reading the description which follows and with reference to the appended drawings, given for illustrative purposes and in no way limiting.

[0031] [Fig-1] [Fig.l] represents a system according to the invention.

[0032] [Fig.2] [Fig.2] represents an image according to the invention.

[0033] [Fig.3] [Fig.3] represents a server device according to the invention. Description of the embodiments

[0034] Preliminary remarks

[0035] In order not to obscure the description and distract the reader from understanding the teachings of the invention, our explanations will not go beyond what we consider necessary for the understanding and appreciation of the underlying concepts of the invention by a person in the technical field of handling devices. Indeed, the embodiments illustrated in the description are, for the most part, composed of elements known to the person skilled in the art.

[0036] Problem solved by the invention

[0037] The technical problem concerns the need for in-person consultations or expensive and potentially inaccurate manual assessments to determine the cost of tattoo removal.

[0038] Technical effect of the invention

[0039] The claimed invention aims to provide, in an objective and reproducible manner, an estimation of the cost and to make the process more practical and accessible for patients.

[0040] The invention: a system

[0041] A system is provided that is configured to generate a segmented image of a tattoo to be removed non-therapeutically.

[0042] This means that the system according to the invention is programmed to produce a segmented image of a tattoo. This tattoo is intended to be removed in a non-therapeutic manner, which means that it is not removed for medical reasons, but rather for aesthetic or personal reasons.

[0043] In practice, as illustrated in [Fig.l], the system 100 comprises a user device 110 and one or more server devices 120.

[0044] This means that the system 100 consists of at least two main components: a user device 110 and one or more server devices 120.

[0045] In other words, the user device 110 could be a device that the user uses to interact with the system 100, such as a computer, a smartphone or a tablet.

[0046] In practice, the user device 110 may be used to capture or receive an image, to send it to the server device 120 for processing, and to receive and display the results of the processing.

[0047] On the other hand, the server device 120 is where most of the image processing operations are performed. This device may include image processing units, machine learning models, and other components necessary to analyze the image acquired or received by the user device 110 and produce the desired results.

[0048] For example, once the image has been received by or taken with the user device 110, it may be sent to the server device 120 for analysis. The server device 120 may then process the image and return the results to the user device 110 for display to the user.

[0049] In particular, the user device 110 comprises an application and a first communication unit. Furthermore, the application, which has a user interface, is designed to run on the user device 110.

[0050] This means that the user device 110 is equipped with an application and a first communication unit.

[0051] In other words, the application is a computer program that is designed to run on the user device 110. The application provides a user interface, which is a way for the user to interact with the system 100.

[0052] For example, the user interface could allow the user to take a photo of a tattoo, select an image of a tattoo from a file, or start and follow the processing process of the acquired or received image.

[0053] Furthermore, the first communication unit is the component of the user device 110 which allows communication with other components of the system 100, such as the server device 120. This unit can use various communication technologies, such as Wi-Fi, Bluetooth, or the mobile network, to transmit the image of the tattoo 210 and receive the results of the processing of the original two-dimensional image 200.

[0054] For example, once the user has taken or received a photo of a tattoo with the application, the first communication unit may send this image to the server device 120 to be analyzed. Once the analysis is complete, the communication unit may receive the results from the server device 120 and transmit them to the application to be displayed to the user.

[0055] In practice, as illustrated in [Fig. 1], the application is designed to receive one or more original two-dimensional images 200 and data 300 associated with the original two-dimensional image 200.

[0056] In particular, as illustrated in [Fig.2], the original two-dimensional image 200 comprises a tattoo 210 and one or more reference objects 220. The tattoo 210 covers a surface of a human body part. The reference object 220 comprises a scale marking 221 which has a known actual size in one or more dimensions. The scale marking 221 visually indicates, directly or indirectly, the known actual size of the scale marking 221.

[0057] In other words, the application is configured to receive one or more original two-dimensional images 200 and data 300 linked to these images.

[0058] This means that the application is designed to process images that have two dimensions: width and height. These images are often called "2D" because they have length and width, but no depth.

[0059] In other words, a two-dimensional image is a flat representation of a scene. It can represent a three-dimensional object, but it does not contain information about the depth of that object. This is the type of image we typically see on a computer or television screen, or in a printed photograph.

[0060] For example, a photo taken with a digital camera is a two-dimensional image. It shows a scene as it appeared from the location where the photo was taken, but it does not provide information about the distance between the camera and the various objects in the scene. The application is designed to work with this type of image, using additional data 300 to help analyze and interpret the image.

[0061] In the invention, the original two-dimensional images 200 comprise a tattoo 210, which is located on a part of the human body, in particular the integuments, the appendages, the appendages or the mucous membranes and one or more reference objects 220.

[0062] The tattoo 210 is the primary target of the analysis of the original two-dimensional image 200. Its presence on a surface of the human body means that the image may contain different shades of color, texture, and brightness that reflect characteristics of both the tattoo 210 and the surrounding skin.

[0063] The reference object 220, in turn, comprises a scale marking 221 which has a known size in one or more dimensions, as well as a specific shape. These directions are generally the standard spatial dimensions: width, height and depth.

[0064] In other words, the scale marking 221 could have a known size in only one dimension, such as width. Or, it could have known sizes in two dimensions, such as width and height. It could even have known sizes in three dimensions, such as width, height, and depth.

[0065] Further, the scale marking provides a visual indication, either directly or indirectly, of its actual size. This may help establish a size scale in the original two-dimensional image 200, which is useful in determining the actual dimensions of the tattoo 210.

[0066] Furthermore, the scale marking 221 also provides a visual indication of its shape, either directly or indirectly. This information can help establish a size scale in the original two-dimensional image 200, which is useful for determining the actual dimensions of the tattoo 210.

[0067] In other words, knowing the shape of the scale marking 221, we can better interpret its size, in particular in images where it is distorted or oriented in a non-standard manner.

[0068] In a first implementation, the scale marking 221 is a machine-readable visual indicator. This visual indicator encodes the known actual size of the scale marking 221.

[0069] In practice, in the first implementation, the scale marking 221 is designed as a machine-readable visual indicator, which can take several forms, including a barcode, a QR code, a data matrix code, an Aztec code, a PDF417 code, a MaxiCode, a Code 128 or a Code 39. This indicator encodes the known actual size of the scale marking 221.

[0070] This means that the scale marking 221 is not simply an object of known size, it also includes a code that can be deciphered by a computer program to reveal the actual size of the object.

[0071] This may allow the machine to instantly understand the scale of the original two-dimensional image 200 without the need for additional analysis or calculations.

[0072] For example, if the scale marking 221 is a 5 cm long barcode which, when scanned by the machine, reveals that it measures 5 cm in reality, then the machine can instantly understand that each unit of measurement in the original two-dimensional image 200 corresponds to a certain number of pixels. This information can be used to determine the actual dimensions of the tattoo 210, based on its size in pixels in the original two-dimensional image 200.

[0073] Furthermore, the use of more complex codes such as the data matrix code or Code 128 can allow more detailed information on the scale to be encoded.

[0074] In a second implementation, the scale marking 221 is a visual indicator that is directly human readable.

[0075] This means that, in the second implementation, the scale marking 221 acts as a visual indicator that can be directly read by a human observer.

[0076] In other words, the scale marking 221 could be an object or sign of known size that is easily recognizable and whose size can be immediately understood by a human observer. This may allow a human user to understand the scale of the original two-dimensional image 200 without the need for specialized machines or software to decipher a code.

[0077] For example, the scale marking 221 could be a 5 cm long ruler that is clearly marked with centimeter graduations. If this ruler appears in the original two-dimensional image 200 next to the tattoo 210, a human observer could immediately understand that each centimeter on the ruler corresponds to a certain number of pixels in the original two-dimensional image 200. This information could then be used to estimate the actual dimensions of the tattoo 210, based on its pixel size in the original two-dimensional image 200.

[0078] In another example, a familiar object of publicly known size may be used as scale marking 221.

[0079] Thus, a coin could serve as a visual indicator. Take a euro coin, for example. One euro has a precise diameter of 23.25 mm, a dimension that is publicly available and widely known.

[0080] If this euro coin is placed next to the tattoo 210 in the original two-dimensional image 200, a human observer could immediately understand that the diameter of the coin corresponds to a certain number of pixels in the original two-dimensional image 200. This information could then be used to estimate the actual dimensions of the tattoo 210, based on its size in pixels in the original two-dimensional image 200. For example, if the tattoo 210 appears to be twice as large as the euro coin in the original two-dimensional image 200, then it is known that the tattoo 210 has an actual size of approximately 46.5 mm.

[0081] In the invention, the application is also designed to display the original two-dimensional image 200 and the data 300 associated with the original two-dimensional image 200 on its user interface.

[0082] This means that the application is programmed to display both the original two-dimensional image 200 and the data 300 associated with this image on its user interface.

[0083] In other words, the application may show the image of the tattoo 210 and the reference object 220 to the user. In addition to this, it may also present the data 300 that is related to this image. This data could include a variety of information, such as the estimated dimensions of the tattoo 210, the method recommended tattoo removal, or any other relevant information.

[0084] For example, after receiving the image of the tattoo 210 and the reference object from the user, the application could display this image on its user interface. Below the original two-dimensional image 200, it could display the estimated dimensions of the tattoo 210, which are calculated from the size of the reference object in the original two-dimensional image 200. It could also display recommendations for the removal of the tattoo 210, such as the type of laser to use and the number of sessions required.

[0085] Further, as illustrated in [Fig.l], the first communication unit is configured to transmit the original two-dimensional image 200 to the server device 120 via a network 400.

[0086] This means that the first communication unit has the capability to transmit the original two-dimensional image 200 to the server device 120 using a network 400.

[0087] In other words, after the tattoo image 210 and the reference object have been captured by the user and received by the application, the communication unit can send them to the server device 120 for processing. This transmission is done via a network 400, which could be a local area network, a wide area network, a wireless network, or even the Internet.

[0088] For example, once the user has taken a picture of the tattoo 210 with the application, this image is sent to the server device 120 by the first communication unit. This transfer can be done via a local Wi-Fi network, a 5G mobile network, or any other type of accessible network. Once the server device 120 has received the image, it can begin to perform the analysis necessary to determine the dimensions of the tattoo 210 and other relevant information.

[0089] Furthermore, as illustrated in [Fig.3], the server device 120 comprises a scale detection unit in an image 121, an image segmentation unit 122, a calculation unit 123 and a second communication unit 124.

[0090] This means that the server device 120 is equipped with several specialized units for processing the original two-dimensional image 200 received from the communication unit 112 of the user device 110.

[0091] In particular, the image scale detection unit 121 is designed to locate and interpret the scale marking 221 in the original two-dimensional image 200. It uses this information to establish a size scale for the original two-dimensional image 200.

[0092] The image segmentation unit 122 is capable of dividing the image into distinct segments. This process makes it possible to separate the watermark 210 from the other elements of the original two-dimensional image 200 for more precise analysis.

[0093] The calculation unit 123 then uses the size scale established by the scale detection unit in an image 121 and the segmentation information provided by the image segmentation unit 122 to calculate the actual dimensions of the tattoo 210.

[0094] Finally, the second communication unit 124 is responsible for transmitting the results calculated by the calculation unit 123 to the user device 110 via the network 400.

[0095] For example, once the image of the tattoo 210 has been received by the server device 120, the scale detection unit in an image 121 identifies the scale marking 221 and establishes a size scale for the original two-dimensional image 200. The image segmentation unit 122 then isolates the tattoo 210 from the other elements of the original two-dimensional image 200. The calculation unit 123 uses this information to determine the actual size of the tattoo 210. These results are then transmitted to the user device 110 by the second communication unit 124.

[0096] In the invention, the image scale detection unit 121 is configured to detect the presence of the scale marking 221 in the original two-dimensional image 200, to determine the known actual size of the scale marking 221 from the detected scale marking, to measure the apparent size of the scale marking 221 in the original two-dimensional image 200, and to calculate a scale ratio from the known actual size and the detected apparent size of the scale marking 221.

[0097] This means that the image scale detection unit 121 performs a specific set of tasks to process the scale marking 221 in the original two-dimensional image 200.

[0098] In practice, this unit begins by identifying the scale marking 221 in the original two-dimensional image 200. This identification could be achieved using image recognition techniques to locate the specific scale marking 221 in the image 200.

[0099] In other words, the scale detection unit in an image 121 could look for specific shapes or patterns that indicate the presence of the scale marking 221. For example, if the scale marking 221 is a ruler with clearly marked graduations, the scale detection unit in an image 121 could look for this type of shape and pattern in the image 200.

[0100] Once the scale marking is detected, it is able to determine the actual size of the scale marking 221 thanks to its coding or its visual representation.

[0101] The next step for the scale detection unit 121 is to measure the apparent size of the scale marking 221 in the original two-dimensional image 200. The apparent size is the size of the object as displayed in the original two-dimensional image 200, which may be different from its actual size due to factors such as perspective or distance of the object from the camera.

[0102] Finally, the image scale detection unit 121 calculates a scale ratio from the known actual size and the apparent size of the scale marking 221. This scale ratio can then be used to convert the dimensions of other objects in the original two-dimensional image 200, such as the tattoo 210, from their apparent size in the original two-dimensional image 200 to their actual size.

[0103] For example, if the scale marking 221 is a 5 cm long object that appears in the original two-dimensional image 200 as being 700 pixels long, then the scale ratio would be 100 pixels per centimeter. If the tattoo 210 appears in the original two-dimensional image 200 as being 1000 pixels long, then its actual length would be 10 cm, according to this scale ratio.

[0104] In a particular embodiment, the image scale detection unit 121 calculates the scale ratio not only from the known actual size and apparent size of the scale marking 221, but also from its shape. This scale ratio can then be used to convert the dimensions of other elements in the original two-dimensional image 200, such as the tattoo 210, from their apparent size in the original two-dimensional image 200 to their actual size. In other words, the shape of the scale marking 221 plays a role in determining the scale ratio, which can help provide a more accurate estimate of the actual dimensions of the objects in the image.

[0105] In the invention, the image segmentation unit 122 is configured to execute a first machine learning model that receives as input the original two-dimensional image 200 and has been trained to separate the tattoo 210 from the rest of the original two-dimensional image 200 and the human body part surface. The model outputs a segmented tattoo 210. As illustrated in [Fig. 3], the image segmentation unit 122 creates a first segmented image 500 that includes the segmented tattoo 210, and determines one or more visual features 600 of the segmented tattoo 210 in the first segmented image 500.

[0106] This means that the image segmentation unit 122 is programmed to execute a first machine learning model. This model is trained to separate the tattoo 210 from the rest of the original two-dimensional image 200 and from the surface of the part of the human body on which it is located.

[0107] Thus, the image segmentation unit 122 uses a first machine learning algorithm to identify the tattoo 210 in the original two-dimensional image 200 and separate it from other elements present. The result is a segmented tattoo 210, which is essentially a version of the original two-dimensional image 200 where only the tattoo 210 is visible.

[0108] Once this segmentation process is completed, the image segmentation unit 122 creates a first segmented image 500 which includes only the segmented tattoo 210.

[0109] Additionally, the image segmentation unit 122 is capable of determining one or more visual characteristics 600 of the segmented tattoo 210 in the first segmented image 500. These visual characteristics 600 could include elements such as the shape, size, color, pattern, or any other relevant visual characteristic 600 of the tattoo 210.

[0110] For example, if tattoo 210 is a 10 cm long black tribal pattern on a person's arm, the first machine learning model would separate this tattoo 210 from the rest of the original two-dimensional image 200, including the surrounding arm skin. It would then create a first segmented image 500 that includes only tattoo 210, and determine visual characteristics 600 of tattoo 210, such as its length, black color, and tribal pattern.

[0111] In particular, various types of machine learning algorithms could be employed to train the first machine learning model. For example, convolutional neural networks (CNNs) are often used for image processing because they are good at recognizing patterns in visual data. A CNN could be trained to recognize the tattoo 210 based on examples of tattoo images and non-tattoo images.

[0112] Furthermore, another type of machine learning algorithm that could be used is deep learning, which is a subcategory of machine learning. Deep neural networks (DNNs), and in particular convolutional neural networks (CNNs), are particularly effective at detecting complex features in images, which could be useful for distinguishing the tattoo 210 from the rest of the original two-dimensional image 200 and from the surface of the human body part. Furthermore, “Vision Transformers” (VITs) represent another interesting possibility. Using the Transformer architecture, generally used in natural language processing, they could constitute an alternative to CNNs for detecting complex features in images such as the tattoo 210.

[0113] In practice, the image segmentation unit 122 could run multiple machine learning models in parallel or in series, using a method called ensemble learning. This could help improve the accuracy of image segmentation by combining the strengths of multiple different models.

[0114] In a first implementation of the image segmentation unit 122, the visual characteristic 600 of the segmented tattoo 210 may be a number of contours. It can also be a number of edges. Another possibility is that it is a number of color areas. It could also be a distribution of color areas. Area is another option for visual characteristic 600. Perimeter is also a possibility. Finally, visual characteristic 600 can be a combination of these elements.

[0115] This means that the visual characteristic 600 of the segmented tattoo 210, as determined by the image segmentation unit 122, can take a variety of different forms depending on the specific tattoo and the information that one wishes to gather about it.

[0116] In other words, the visual characteristic 600 could be the number of outlines of the tattoo 210, which would be the number of distinct lines or shapes that compose it.

[0117] It could also be the number of edges, which would be the number of points where two contour lines meet.

[0118] The visual characteristic 600 could also be the number of color areas, which would be the number of distinct areas in the tattoo 210 that are colored differently.

[0119] Additionally, it could be the distribution of the color areas, which would be how these color areas are distributed in the tattoo 210.

[0120] Another option for visual feature 600 is the area of ​​the tattoo 210, which would be the total area it occupies.

[0121] The perimeter, which would be the total length of the contours of the tattoo 210, is also a possibility.

[0122] Finally, the visual characteristic 600 may be a combination of these elements. For example, it could be a data set comprising both the number of contours and the area of ​​the tattoo 210.

[0123] For example, if tattoo 210 is a 10 cm long black tribal pattern that has three different color areas (black, white, and red), three distinct outlines, and an area of ​​15 cm2, then this information could all be recorded as visual features 600 of the segmented tattoo 210.

[0124] In a second implementation of the image segmentation unit 122, the original two-dimensional image 200 is encoded in the Red-Green-Blue (RGB) color space. In this case, the visual characteristic 600 of the segmented tattoo 210 is selected from several options.

[0125] This means that in this second implementation of the image segmentation unit 122, the original two-dimensional image 200 is processed in the Red-Green-Blue (RGB) color space.

[0126] The RGB color space is a color representation system used commonly in digital technologies, where each color is represented by a mixture of three primary colors: red, green and blue.

[0127] In other words, the original two-dimensional image 200 is analyzed in terms of these three color components. Each pixel of the original two-dimensional image 200 has a value for red, green, and blue, which, when combined, yield the final color of the pixel.

[0128] The visual characteristic 600 of the segmented tattoo 210 in this context can be selected from several options.

[0129] An option of the visual characteristic 600 of the segmented tattoo 210 is the sum of a blue channel value of all pixels within an outline of the segmented tattoo 210, divided by the sum of a green channel value and a red channel value of all pixels within said outline of the segmented tattoo 210.

[0130] This means that one of the possible visual characteristics 600 of the segmented tattoo 210 could be calculated from the values ​​of the color channels of the pixels that make up the tattoo 210.

[0131] In other words, this characteristic is obtained by summing the value of the blue channel for all the pixels within the contour of the segmented tattoo 210, then dividing this sum by the sum of the values ​​of the green and red channels for all the pixels within the same contour.

[0132] This characteristic could be used to represent the relative dominance of blue compared to green and red in the 210 tattoo. The higher this value, the more dominant blue is in the 210 tattoo compared to green and red.

[0133] For example, if tattoo 210 is an image of a blue wave, this characteristic would likely be quite high, because the majority of pixels within the outline of tattoo 210 would have a high blue channel value and lower green and red channel values. On the other hand, if tattoo 210 is an image of a green leaf, this characteristic would likely be quite low, because the majority of pixels would have a high green channel value and lower blue and red channel values.

[0134] Another option of the visual characteristic 600 of the segmented tattoo 210 is the sum of the green channel value of all pixels within the outline of the segmented tattoo 210, divided by the sum of the blue channel value and the red channel value of all pixels within said outline of the segmented tattoo 210.

[0135] This means that, similarly to the previous option, another possible visual characteristic 600 of the segmented tattoo 210 can be calculated from the values ​​of the color channels of the pixels within the outline of the tattoo 210.

[0136] In other words, this characteristic is obtained by summing the value of the green channel for all the pixels inside the contour of the segmented tattoo 210, then dividing this sum by the sum of the values ​​of the blue and red channels for all the pixels inside the contour of the segmented tattoo 210. inside the same outline.

[0137] This characteristic could be used to represent the relative dominance of green compared to blue and red in the 210 tattoo. The higher this value, the more dominant green is in the 210 tattoo compared to blue and red.

[0138] For example, if tattoo 210 is an image of a green leaf, this characteristic would likely be quite high, as the majority of pixels within the outline of tattoo 210 would have a high green channel value and lower blue and red channel values. Conversely, if tattoo 210 is an image of a blue wave, this characteristic would likely be quite low, as the majority of pixels would have a high blue channel value and lower green and red channel values.

[0139] Another possibility of the visual characteristic 600 of the segmented tattoo 210 is to take the sum of the red channel value of all the pixels within the outline of the segmented tattoo 210, divided by the sum of the blue channel value and the green channel value of all the pixels within said outline of the segmented tattoo 210.

[0140] This means that another visual characteristic 600 of the segmented tattoo 210 can be derived from analyzing the values ​​of the color channels of the pixels that lie within the outline of the tattoo 210.

[0141] In other words, this characteristic is obtained by summing the value of the red channel for all the pixels within the contour of the segmented tattoo 210, then dividing this sum by the sum of the values ​​of the blue and green channels for all the pixels within the same contour.

[0142] This characteristic could be used to represent the relative dominance of red compared to blue and green in the 210 tattoo. The higher this value, the more dominant red is in the 210 tattoo compared to blue and green.

[0143] For example, if tattoo 210 is an image of a red rose, this characteristic would likely be quite high, as the majority of pixels within the outline of tattoo 210 would have a high red channel value and lower blue and green channel values. Conversely, if tattoo 210 is an image of a blue wave or a green leaf, this characteristic would likely be quite low, as the majority of pixels would have lower red channel values ​​relative to the blue or green channel values.

[0144] Finally, the visual characteristic 600 of the visual characteristic 600 of the segmented tattoo 210 may be a combination of these options, which would allow a more detailed and nuanced description of the colors and their distribution in the segmented tattoo 210.

[0145] This means that instead of using only one of the above options to describe the visual characteristic 600 of the segmented tattoo 210, it would also be possible to combine several of them to obtain a more complete and more nuanced colors and their distribution in the tattoo 210.

[0146] In other words, instead of being limited to a single measurement, such as red, green, or blue dominance, the image segmentation unit 122 could calculate several of these visual characteristics 600 and use them together to describe the tattoo 210.

[0147] This type of combination could allow for better representation of complex colors and color distribution patterns in the tattoo 210.

[0148] For example, if the watermark 210 is an image of a sunset scene with a glowing red sky over a blue sea, the visual characteristic 600 could include both the dominance of red in the sky and the dominance of blue in the sea, as well as how these colors are distributed throughout the watermark 210. By combining this information, it would be possible to create a more detailed and accurate description of the segmented watermark 210.

[0149] In a third implementation of the image segmentation unit 122, the original two-dimensional image 200 is encoded in the hue-saturation-brightness (HSL) color space. In this case, the visual characteristic 600 of the segmented watermark 210 can be chosen from several options.

[0150] This means that in this third implementation of the image segmentation unit 122, the original two-dimensional image 200 is transformed into the hue-saturation-brightness (HSL) color space.

[0151] HSL space is a color representation model that describes colors in terms of their hue (the pure color), saturation (the intensity of the color), and luminosity (the lightness or darkness of the color).

[0152] In other words, each pixel of the original two-dimensional image 200 is analyzed according to these three color characteristics. Hue determines the base color, saturation determines how intense or pale that color is, and brightness determines how light or dark that color is.

[0153] The visual characteristic 600 of the segmented tattoo 210 can be chosen from several options in this context.

[0154] An option of the visual characteristic 600 of the segmented tattoo 210 is a sum of the Saturation channel value of all pixels within an outline of the segmented tattoo 210. This sum is divided by a sum of the Hue channel value and the Brightness channel value of all pixels within the outline of the segmented tattoo 210.

[0155] This means that one of the possible visual characteristics 600 of the segmented tattoo 210 can be calculated from the values ​​of the color channels of the pixels that are inside the outline of the tattoo 210.

[0156] In other words, this characteristic is obtained by summing the value of the channel Saturation for all pixels inside the 210 segmented tattoo outline, then dividing that sum by the sum of the Hue and Brightness channel values ​​for all pixels inside the same outline.

[0157] This characteristic could be used to represent the relative dominance of saturation versus hue and brightness in the tattoo 210. The higher this value, the more dominant saturation is in the tattoo versus hue and brightness.

[0158] For example, if tattoo 210 is an image of a bright red rose, this characteristic would likely be quite high, as the majority of pixels within the outline of tattoo 210 would have a high Saturation channel value (signifying an intense color) and lower Hue and Brightness channel values. Conversely, if tattoo 210 is an image of a pale red rose, this characteristic would likely be quite low, as the majority of pixels would have lower Saturation channel values ​​(signifying a pale color) relative to the Hue and Brightness channel values.

[0159] Another option of the visual characteristic 600 of the segmented tattoo 210 is to take the sum of the Brightness channel value of all pixels within the outline of the segmented tattoo 210. This is then divided by a sum of the Saturation channel value and the Hue channel value of all pixels within the outline of the segmented tattoo 210.

[0160] This means that, similarly to the previous option, another possible visual characteristic 600 of the segmented tattoo 210 can be calculated from the values ​​of the color channels of the pixels within the outline of the tattoo 210.

[0161] In other words, this characteristic is obtained by summing the value of the Brightness channel for all the pixels within the contour of the segmented tattoo 210, then dividing this sum by the sum of the values ​​of the Saturation and Hue channels for all the pixels within the same contour.

[0162] This characteristic could be used to represent the relative dominance of brightness versus saturation and hue in the tattoo 210. The higher this value, the more dominant brightness is in the tattoo 210 versus saturation and hue.

[0163] For example, if tattoo 210 is an image of a bright lighthouse in the night, this characteristic would likely be quite high, as the majority of pixels within the outline of tattoo 210 would have a high Brightness channel value (signifying a light color) and lower Saturation and Hue channel values. Conversely, if tattoo 210 is an image of a dark starry night, this characteristic would likely be quite low, as the majority of pixels would have lower Brightness channel values ​​(signifying a dark color) compared to Saturation and Hue channel values.

[0164] Yet another possibility of the visual characteristic 600 of the segmented tattoo 210 is to calculate the sum of the Hue channel value of all pixels within the outline of the segmented tattoo 210. This sum is then divided by a sum of the Brightness channel value and the Saturation channel value of all pixels within the outline of the segmented tattoo 210.

[0165] This means that another visual characteristic 600 of the segmented tattoo 210 can be derived from analyzing the values ​​of the color channels of the pixels that lie within the outline of the tattoo 210.

[0166] In other words, this characteristic is obtained by summing the value of the Hue channel for all the pixels within the contour of the segmented tattoo 210, then dividing this sum by the sum of the values ​​of the Brightness and Saturation channels for all the pixels within the same contour.

[0167] This characteristic could be used to represent the relative dominance of hue relative to brightness and saturation in the tattoo 210. The higher this value, the more dominant the hue is in the tattoo 210 relative to brightness and saturation.

[0168] For example, if tattoo 210 is an image of a bright red rose, this characteristic would likely be quite high, as the majority of pixels within the outline of tattoo 210 would have a high Hue channel value (signifying a red color), and lower Brightness and Saturation channel values. Conversely, if tattoo 210 is an image of a gray rose, this characteristic would likely be quite low, as the majority of pixels would have lower Hue channel values ​​(signifying an absence of color) relative to the Brightness and Saturation channel values.

[0169] Finally, the visual characteristic 600 of the visual characteristic 600 of the segmented tattoo 210 may be a combination of these options, which would allow a more detailed description of the colors and their distribution in the segmented tattoo 210 using the HSL color space.

[0170] This means that instead of using only one of these options to describe the visual characteristic 600 of the segmented tattoo 210, it would also be possible to combine several of them to obtain a more complete description of the colors and their distribution in the tattoo 210.

[0171] In other words, instead of being limited to a single measurement, such as hue, brightness, or saturation dominance, the image segmentation unit 122 could calculate several of these features and use them together to describe the watermark 210. This type of combination could allow for better representation of complex colors and color distribution patterns in the tattoo 210.

[0172] For example, if the watermark 210 is an image of a sunset scene with a glowing red sky over a blue sea, the visual characteristic 600 could include both the dominance of the red hue in the sky and the dominance of the saturation in the sea, as well as how these colors are distributed across the watermark 210. By combining this information, it would be possible to create a more detailed and accurate description of the segmented watermark 210.

[0173] In the invention, as illustrated in [Fig.3], the calculation unit 123 is designed to automatically calculate a tattoo removal price 700, at least based on the scale ratio and the visual characteristic 600.

[0174] This means that the calculation unit 123 has the function of automatically determining a price for tattoo removal, using at least two types of information: the scale ratio and the visual characteristic 600 of the segmented tattoo 210. In particular, this price can be determined per session or globally.

[0175] In other words, the scale ratio could refer to the size of the tattoo 210 relative to some reference, such as the patient's body size. A larger size could result in a higher tattoo removal price 700 per session, as it would require more time and resources to remove the tattoo 210.

[0176] The visual characteristic 600, in turn, could refer to different aspects of the tattoo 210, such as color, saturation, and brightness, as discussed previously. These visual characteristics 600 could also influence the tattoo removal price 700, as some types of tattoos could be more difficult to remove than others. For example, a tattoo 210 with bright, saturated colors could be more difficult to remove than a tattoo 210 with pale, less saturated colors, and therefore cost more overall.

[0177] Thus, by combining this information, the calculation unit 123 is capable of automatically calculating a price for tattoo removal which reflects both the size of the tattoo 210, its visual complexity and the number of sessions required for tattoo removal.

[0178] In the invention, as illustrated in [Fig.3], the second communication unit 124 is designed to transmit to the user device 110, via the network 400, the data 300. This data 300 comprises the first segmented image 500 and the tattoo removal price 700, which means that they are directly associated with the original two-dimensional image 200.

[0179] Thus, the second communication unit 124 has the role of transmitting certain information to the user device 110 via the network 400. This information includes the tattoo removal price 700 and the first segmented image 500, which correspond to the data 300 associated with the original two-dimensional image 200.

[0180] In other words, once the calculation unit 123 has determined the price of die tattoo 700 using the scale ratio and visual characteristic 600 of the tattoo 210, this information is sent to the user device 110. The first segmented image 500 is also sent, which allows the user to see precisely which parts of the original two-dimensional image 200 have been identified as part of the tattoo 210.

[0181] For example, if a user submitted an image of a part of their body with a tattoo 210 to obtain an estimate of the tattoo removal cost, the second communication unit 124 would send a segmented version of this image, where the tattoo 210 is clearly identified, as well as the estimate of the tattoo removal cost. This allows the user to understand both the estimated cost and how the system 100 identified and analyzed their tattoo 210.

[0182] First embodiment of the invention: performing a detailed and accurate analysis of the original two-dimensional image 200, which can be used to obtain useful information about the tattoo 210 and to determine the cost of its removal. Thus, this makes it possible to provide a more accurate and personalized service.

[0183] In practice, in the first embodiment of the invention, the image segmentation unit 122 executes a second machine learning model. This second machine learning model receives as input the original two-dimensional image 200 and the first segmented image 500. It has been previously trained to separate the human body part surface from the rest of the original two-dimensional image 200 and the segmented tattoo 210. The result of this operation is a segmented human body part surface.

[0184] This means that the image segmentation unit 122 uses a second machine learning model to identify and segment the surface of the human body in the original two-dimensional image 200. The second machine learning model receives two types of inputs: the original two-dimensional image 200 and the first segmented image 500, which contains the segmented tattoo 210.

[0185] In practice, the second machine learning model has been trained in advance to distinguish the surface of the human body from the rest of the original two-dimensional image 200 and the segmented tattoo 210. This means that the model has been trained with example images containing human body parts, tattoos and other types of content, and has learned to distinguish between them.

[0186] The result of this process is what is called a segmented human body part surface. This is a version of the original two-dimensional image 200 where the human body part has been clearly identified and separated from the rest of the original two-dimensional image 200.

[0187] For example, if the original two-dimensional image 200 is a photo of a person with a tattoo 210 on the arm, the second machine learning model would be able to identify and segment the portion of the arm in the original two-dimensional image 200, distinguishing it from the tattoo 210 and the rest of the original two-dimensional image 200.

[0188] In practice, several types of machine learning algorithms can be used to train the second machine learning model.

[0189] An example is the convolutional neural network (CNN), which is often used for image processing. It is particularly effective in identifying visual patterns in images, which would be useful in distinguishing the surface of the human body from the rest of the image 200 and the segmented tattoo 210.

[0190] Thus, the image segmentation unit 122 could use a combination of machine learning algorithms to perform its task. This means that the second machine learning model is able to effectively distinguish the surface of the human body from the rest of the original two-dimensional image 200 and the segmented tattoo 210.

[0191] Next, in the first embodiment of the invention, the image segmentation unit 122 creates a second segmented image. This image comprises the segmented human body part surface.

[0192] This means that, once the surface of the human body part has been segmented by the second machine learning model, the image segmentation unit 122 creates a second segmented image. This second segmented image contains the surface of the human body part that has been identified and separated from the rest of the original two-dimensional image 200.

[0193] In other words, the second segmented image is a version of the original two-dimensional image 200 where only the human body part is highlighted, the rest of the original two-dimensional image 200 being presumably masked or made less visible.

[0194] For example, if the original two-dimensional image 200 is a photo of a person with a tattoo 210 on their arm, the second segmented image would be a version of that photo where only the person's arm is clearly visible, while the rest of the original two-dimensional image 200, including the tattoo 210, is made less visible or completely obscured. This second segmented image would make it easier to analyze the human body part without being distracted by the rest of the original two-dimensional image 200.

[0195] Then, in the first embodiment of the invention, the image segmentation unit 122 executes a third machine learning model. This model receives as input the second segmented image. It has been previously trained to classify the segmented human body part surface. As illustrated in [Fig.3], the result is a class 800 among a plurality of classes, each class corresponding to a human body element.

[0196] In other words, the image segmentation unit 122 uses a third machine learning model that functions as a classifier to categorize the surface of the human body part that has been segmented. This classifier, based on the third machine learning model, receives as input the second segmented image, which contains this surface of the human body part.

[0197] In practice, this classifier, which is the third machine learning model, has been previously trained to identify different elements of the human body. This means that the model has been trained with sample images of different parts of the human body and has learned to recognize and classify these different parts.

[0198] The result of this process is a class 800 which is assigned to the surface of the segmented human body part. There are several of these classes, each corresponding to a different element of the human body.

[0199] For example, if the second segmented image is an image of an arm, the classifier, based on the third machine learning model, would be able to recognize and classify this image as an image of an arm. This classification provides additional information about the image, which could be used in further analyses.

[0200] In practice, several types of machine learning algorithms can be used to train the third machine learning model. For example, convolutional neural networks (CNNs) can be used to classify different parts of the human body. CNNs are effective for this type of task because they are able to recognize visual patterns in images, which would be useful for distinguishing different parts of the human body.

[0201] In addition, deep neural networks (DNNs) can also be used. DNNs are capable of learning complex features from large amounts of data, which could be useful for recognizing and classifying different parts of the human body in the image 200.

[0202] In practice, another type of machine learning model that could be used is the recurrent neural network (RNN). RNNs are particularly effective at processing sequences of data, which could help analyze the image 200 in sequence to detect the presence and outline of the human body part.

[0203] In particular, the image segmentation unit 122 could use a combination of these machine learning algorithms to accomplish its task. This means that the third machine learning model is able to effectively classify the surface of the segmented human body part in the second segmented image.

[0204] Thus, the classifier, based on the third machine learning model, uses machine learning techniques to identify and classify the different parts of the human body present in the image 200. Each of these body parts is assigned to a specific class, allowing precise identification of the different parts of the body in the image 200.

[0205] Finally, as illustrated in [Fig.3], in the first embodiment of the invention, the calculation unit 123 is designed to automatically calculate the tattoo removal price 700. This calculation is also based on the class 800 obtained.

[0206] This means that the calculation unit 123, in addition to using the scale ratio and the visual characteristic 600 of the tattoo 210 to calculate the tattoo removal price 700, also uses the class 800 obtained from the third machine learning model.

[0207] In other words, this class 800, which corresponds to a specific element of the human body, also plays a role in determining the price of tattoo removal 700. This could be due to the fact that different elements of the human body may require different tattoo removal techniques, which may have different costs.

[0208] For example, removing a 210 tattoo from an arm may cost less than removing a 210 tattoo from a more sensitive part of the body, such as the neck or face, due to the difference in size and complexity of the skin in those areas.

[0209] Thus, the calculation unit 123 takes into account the class 800 obtained to provide a more precise estimate of the cost of tattoo removal.

[0210] This approach allows for a more accurate and detailed estimate of the cost of tattoo removal.

[0211] Second embodiment of the invention: improving the accuracy of image analysis, which makes it possible to obtain more accurate information about the tattoo 210 and the part of the human body on which it is located. Therefore, this makes it possible to provide a more accurate and personalized service to the user.

[0212] In practice, in the second embodiment of the invention, the application allows a user to select, in the original two-dimensional image 200, one or more pixels. This pixel has characteristics corresponding to the characteristics of the surface of the human body part not covered by the tattoo 210. This selection makes it possible to generate colorimetric information. This information are representative of the colorimetry of the selected pixel.

[0213] This means that the application offers the user the possibility of choosing one or more specific pixels in the original two-dimensional image 200. The selected pixel or pixels are those which have characteristics corresponding to the part of the human body which is not covered by the tattoo 210.

[0214] In other words, the user can choose a pixel or pixels that represent the color of the untattooed skin. This selection generates colorimetric information, which is data representative of the colorimetry of the selected pixel or pixels.

[0215] For example, if the user selects a pixel on the portion of the arm that is not tattooed, the generated colorimetric information would be representative of the skin color of the user's arm. This information can be used in various applications, such as adjusting the skin color in the segmented image, estimating the skin color after the tattoo 210 is removed, or even creating a simulation of what the skin might look like after tattoo removal.

[0216] Then, in the second embodiment of the invention, the machine learning models, the first and / or the second, are designed to receive the colorimetric information as input.

[0217] This means that the machine learning models, either the first, the second, or both, are able to take as input the colorimetric information generated by the selection of pixels in the application.

[0218] In other words, these machine learning models can use the information about the colorimetry of the untattooed skin to assist in the segmentation of the original two-dimensional image 200 and the classification of the body part. This information can be used to improve the accuracy of the segmentation and classification, taking into account the specific color of the user's skin.

[0219] For example, if the colorimetric information indicates that the user's skin has a particular shade, the machine learning models can use this information to better differentiate the skin from the tattoo 210 in the original two-dimensional image 200, or to more accurately classify the body part on which the tattoo 210 is located. This feature makes the machine learning models more adaptive and more accurate, by allowing them to account for the diversity of skin colors among users.

[0220] Third embodiment of the invention: improving the accuracy of image analysis performed by the machine learning models. This can contribute to better identification and classification of the tattoo 210 and the human body part in the image, which in turn can improve the accuracy of the cost estimate for tattoo removal 210.

[0221] In practice, in the third embodiment of the invention, as illustrated in [Fig. 3], the server device 120 further comprises an image perspective correction unit 125. The image perspective correction unit 125 is configured to perform perspective correction. This correction is applied to the original two-dimensional image 200. As illustrated in [Fig. 3], the result of this operation is a corrected two-dimensional image 900.

[0222] This means that the image perspective correction unit 125 is programmed to perform a specific operation on the original two-dimensional image 200. This operation is called perspective correction.

[0223] In other words, the image perspective correction unit 125 adjusts the original two-dimensional image 200 to correct distortions that may be caused by the viewpoint from which the original two-dimensional image 200 was taken. This adjustment may involve resizing, rotating, moving, or other transformations of the original two-dimensional image 200 to obtain a more frontal view of the objects depicted.

[0224] For example, if the original two-dimensional image 200 was taken from an angle that makes the watermark 210 distorted or difficult to distinguish, the image perspective correction unit 125 could adjust the image to make the watermark 210 clearer and easier to analyze. The result of this perspective correction operation is a corrected two-dimensional image 900, which is then used in subsequent steps of the original two-dimensional image 200 processing process.

[0225] In practice, the image perspective correction unit 125 may use several types of algorithms to perform the perspective correction.

[0226] A common example is the homographic transformation algorithm, which allows perspective distortions in the original two-dimensional image 200 to be modeled and corrected. This may include resizing, rotating, moving, or other transformations of the image.

[0227] Furthermore, techniques based on feature point detection may also be used. For example, a SIFT (Scale-Invariant Feature Transform) algorithm could be used to detect and match with the feature points in the original two-dimensional image 200. These matches can then be used to calculate the perspective transformation necessary to correct the image.

[0228] In particular, once the image perspective correction unit 125 corrects the perspective of the original two-dimensional image 200, the tattoo 210 and the human body part can be more clearly seen and analyzed by the machine learning models.

[0229] Thus, perspective correction algorithms can improve the accuracy of image analysis and, consequently, the quality of the services offered by the system.

[0230] Finally, in the third embodiment of the invention, as illustrated in [Fig.3], the first and / or second machine learning models are designed to use the corrected two-dimensional image 900. They use the corrected two-dimensional image 900 instead of the original two-dimensional image 200.

[0231] This means that the machine learning models, whether the first, second, or both, are programmed to work with the corrected two-dimensional image 900, rather than the original two-dimensional image 200.

[0232] In other words, once the original image 200 has been corrected, it is this corrected version that is used by the machine learning models.

[0233] Correcting the original two-dimensional image 200 could include various modifications to improve the quality of the original two-dimensional image 200, such as adjusting lighting, color correction, noise removal, among others.

[0234] These machine learning models can perform various tasks on the corrected image, such as segmentation of the tattoo 210, analysis of visual features 600, or other operations relevant to tattoo removal.

[0235] This helps ensure that models are working with the best possible data, which can improve the accuracy of their results and ultimately the quality of service provided to the user.

[0236] Fourth embodiment of the invention: improving the accuracy of image analysis performed by the machine learning models. This can contribute to better identification and classification of the tattoo 210 and the human body part in the image, which in turn can improve the accuracy of estimating the cost of removing the tattoo 210.

[0237] In practice, in the fourth embodiment of the invention, the server device 120 further comprises a multi-level thresholding unit (not shown). The multi-level thresholding unit is configured to apply multi-level thresholding. This thresholding may be applied to the original two-dimensional image 200, to the first segmented image 500, or to the second segmented image. Applying this thresholding produces a thresholded two-dimensional image.

[0238] This means that the multi-level thresholding unit has the capability of applying multi-level thresholding to an image. This image may be the original two-dimensional image 200, the first segmented image 500, or the second segmented image.

[0239] In other words, the multi-level thresholding unit transforms the image into a simplified version, where the pixels are reduced to a more limited color palette. This is done by setting multiple brightness or color thresholds, and assigning each pixel a value based on those thresholds. For example, pixels with a brightness above a certain threshold can be rendered white, while those with a brightness below that threshold can be rendered black.

[0240] Applying this thresholding produces a thresholded two-dimensional image. This thresholded image simplifies the information contained in the original two-dimensional image 200, which may make certain image processing tasks, such as segmentation or classification, easier or more accurate.

[0241] For example, if the goal is to distinguish a tattoo 210 from untattooed skin, applying thresholding can make this distinction clearer by reducing color and brightness variations that are not relevant to this task.

[0242] In practice, the multi-level thresholding unit may use various algorithms to perform the thresholding. Common examples include the global thresholding algorithm and the adaptive thresholding algorithm.

[0243] With global thresholding, a fixed threshold is used for all pixels in the image. If the value of a pixel exceeds this threshold, the pixel's value is set to one value (e.g., white), otherwise it is set to another value (e.g., black).

[0244] On the other hand, adaptive thresholding is a more sophisticated method that adjusts the threshold for each pixel based on its neighbors. This can be useful for images with varying lighting conditions, as it takes into account local variations in brightness.

[0245] In particular, once the multi-level thresholding unit has applied thresholding to the image 200, the tattoo 210 and the human body part can be more easily distinguished. This can improve the accuracy of machine learning models in segmenting and classifying these features in the image.

[0246] Thus, the use of thresholding techniques can improve the efficiency and accuracy of image processing.

[0247] Finally, in the fourth embodiment of the invention, the machine learning models, namely, the first model, the second model and the third model are designed to use the thresholded two-dimensional image. They use this thresholded image instead of the original two-dimensional image 200, the first segmented image 500 and the second segmented image respectively.

[0248] This means that the three machine learning models, the first, the second and the third, are designed to work with the thresholded two-dimensional image that was produced by the multi-level thresholding unit. These models use the thresholded image instead of the original two-dimensional image 200, the first segmented image 500 or the second segmented image.

[0249] In other words, once the original image or the segmented image has been thresholded, it is this thresholded version that is used by machine learning models. The thresholded image is a simplified version of the original two-dimensional image 200 or the segmented image, where each pixel is either fully on or fully off, depending on whether it is above or below a certain threshold.

[0250] For example, if the goal is to identify the tattoo 210 in the image, thresholding may make the tattoo 210 easier to distinguish from the rest of the original two-dimensional image 200 by simplifying color and brightness variations. Machine learning models may then use this thresholded image to perform their tasks, such as segmenting the tattoo 210, analyzing visual features 600, or other operations relevant to tattoo removal.

[0251] Conclusion

[0252] We have described and illustrated the invention. However, the invention is not limited to the embodiments that we have presented. Indeed, numerous combinations of variants, alternatives, embodiments and implementations can be envisaged without requiring substantial modifications of the invention. Thus, an expert in the field can deduce other variants, alternatives, embodiments and implementations, upon reading the description and the appended figures and depending on the economic, ergonomic and dimensional constraints to be respected.

[0253] In particular, when an element is "designed" to perform a particular function, this means that this element is created specifically for the purpose of performing this particular function.

[0254] However, depending on the needs and available resources, it may be possible to consider using an existing element, which will be modified or adapted to fulfill this particular function, without requiring substantial modifications to the invention.

[0255] The invention may be the subject of numerous variants and applications other than those described above. In particular, unless otherwise indicated, the different structural and functional features of each particular manner described above should not be considered as combined and / or closely and / or inextricably linked to each other, but, on the contrary, as simple juxtapositions. Furthermore, the structural and / or functional features of the different embodiments described above may be the subject in whole or in part of any different juxtaposition or any different combination.

[0256] In practice, the various units that make up the server device 120 are processing mechanisms. These mechanisms may take the form of processors, which may be standard or specialized, calculators, sub-calculators, or computer modules in the form of software or programs. These units can be interconnected by data buses.

[0257] For example, a standard processor within the server device 120 could be used to perform general tasks, while a specialized processor could be used for more specific tasks, such as image processing.

[0258] Similarly, a calculator or sub-calculator could be used to perform specific calculations, such as determining the price of tattoo removal.

[0259] The computer modules in the form of software or programs could include machine learning algorithms for segmenting and analyzing the original two-dimensional image 200.

[0260] And the data bus allows efficient communication between these different units.

Claims

Claims

1. A system (100) for generating a segmented image of a tattoo (210) to be removed non-therapeutically, the system (100) comprising a user device (110) and one or more server devices (120), - the user device (110) comprising an application and a first communication unit, - the application having a user interface and being adapted to run on the user device (110) so as to, - receive one or more original two-dimensional images (200) and data (300) associated with the original two-dimensional image (200), the original dimensional image (200) comprising a tattoo (210) and one or more reference objects (220), the tattoo (210) covering a surface of a human body part, the reference object (220) comprising one or more scale markings (221) which has a known actual size in one or more dimensions,the scale marking (221) being adapted to visually indicate, directly or indirectly, at least the known actual size of the scale marking (221), and - displaying, on the user interface, the original two-dimensional image (200) and the data (300) associated with the original two-dimensional image (200), - the first communication unit being adapted to transmit the original two-dimensional image (200) to a server device (120) via a network (400), - the server device (120) comprising an image scale detection unit (121), an image segmentation unit (122), a calculation unit (123) and a second communication unit (124), - the image scale detection unit (121) being adapted to, - detect the presence of a scale marking (221) in the original two-dimensional image (200), - determine, from the detected scale marking (221), directly or indirectly,the known actual size of the scale marking (221), - measuring, in the original two-dimensional image (200), an apparent size of the scale marking (221), and - calculating a scale ratio, at least from the known actual size of the detected scale marking (221) and the apparent size of the detected scale marking (221), - the image segmentation unit (122) being designed to, - execute a first machine learning model which receives as input the original two-dimensional image (200) and which has been previously trained to separate the tattoo (210) from the rest of the original two-dimensional image (200) and from the surface of the human body part, so as to obtain as output a segmented tattoo (210), - create a first segmented image (500) which comprises the segmented tattoo (210), and - determine, in the first segmented image (500), one or more visual characteristics (600) of the segmented tattoo (210), - the calculation unit (123) being designed to automatically calculate a tattoo removal price (700), at least on the basis of the scale ratio and the visual characteristic (600), - the second communication unit (124) being designed to transmit to the user device (110), by through the network (400),the data (300) associated with the original two-dimensional image (200) and which includes the first segmented image (500) and the tattoo removal price (700).,

2. The system (100) of claim 1, wherein the scale marking (221) is a visual indicator that is machine readable and encodes the known actual size of the scale marking (221).

3. The system (100) of claim 1, wherein the scale marking (221) is a visual indicator that is directly human readable.

4. The system (100) of any one of claims 1 to 3, wherein the visual characteristic (600) of the segmented tattoo (210) is selected from a number of contours, a number of edges, a number of color areas, a distribution of color areas, an area, a perimeter, or any combination thereof.

5. The system (100) of any one of claims 1 to 3, wherein, when the original two-dimensional image (200) is encoded in the Red-Green-Blue, RGB, color space, the visual characteristic (600) of the segmented tattoo (210) is selected from, - a sum of a blue channel value of all pixels within an outline of the segmented tattoo (210), divided by a sum of a green channel value and a red channel value of all pixels in said outline of the segmented tattoo (210), - a sum of the green channel value of all pixels within the

6.

7. segmented tattoo outline (210), divided by a sum of the blue channel value and the red channel value of all pixels in said segmented tattoo outline (210), - a sum of the red channel value of all pixels within the segmented tattoo outline (210), divided by a sum of the blue channel value and the green channel value of all pixels within said segmented tattoo outline (210), or any combination thereof. The system (100) of any one of claims 1 to 3, wherein when the original two-dimensional image (200) is encoded in the Hue-Saturation-Lightness, HSL, color space, the visual characteristic (600) of the segmented tattoo (210) is selected from, - a sum of the Saturation channel value of all pixels within an outline of the segmented tattoo (210), divided by a sum of the Hue channel value and the Lightness channel value of all pixels in the outline of the segmented tattoo (210), - a sum of the Lightness channel value of all pixels within the outline of the segmented tattoo (210), divided by a sum of the Saturation channel value and the Lightness channel value of all pixels in the outline of the segmented tattoo (210), - a sum of the Hue channel value of all pixels within the outline of the segmented tattoo (210),divided by a sum of the Brightness channel value and the Saturation channel value of all pixels within the segmented (210) tattoo outline, or any combination thereof., The system (100) of any one of claims 1 to 6, wherein, - the image segmentation unit (122) is further configured to, - execute a second machine learning model that receives as input the original two-dimensional image (200) and the first segmented image (500), and that has been previously trained to separate the human body part surface from the rest of the original two-dimensional image (200) and the segmented tattoo (210), so as to obtain as output a segmented human body part surface, - create a second segmented image that includes the segmented human body part surface, and - execute a third machine learning model that receives as input the second segmented image and that has been previously trained to classify the segmented human body part surface, so as to obtain as output a class (800) from a plurality of classes, each class corresponding to a human body element, - the calculation unit (123) being further adapted to automatically calculate the tattoo removal price (700) on the basis further of the class (800).

8. The system (100) of any one of claims 1 to 7, wherein, - the application is further configured to run on the user device (110) so as to allow a user to select, in the original two-dimensional image (200), one or more pixels that have characteristics corresponding to the characteristics of the surface of the human body part not covered by the tattoo (210), so as to generate colorimetric information that is representative of the colorimetry of the selected pixel, and - the first machine learning model and / or the second machine learning model are further configured to receive as input the colorimetric information.

9. The system (100) of any one of claims 1 to 8, wherein, - the server device (120) further comprises an image perspective correction unit (125) which is configured to perform a perspective correction to the original two-dimensional image (200) so as to output a corrected two-dimensional image (900), and - the first machine learning model and / or the second machine learning model are further configured to use the corrected two-dimensional image (900) instead of the original two-dimensional image (200).

10. The system (100) of any one of claims 1 to 9, wherein, - the server device (120) further comprises a multi-level thresholding unit which is further configured to apply multi-level thresholding to the original two-dimensional image (200) or to the first segmented image (500) or to the second segmented image so as to output a thresholded two-dimensional image, and - the first machine learning model, the second machine learning model and the second machine learning model automation are further configured to use the thresholded two-dimensional image, respectively, in place of the original two-dimensional image (200), the first segmented image (500) and the second segmented image.

Images