Method for selecting lens with special light transmissive filter

Abstract

The invention relates to a method for selecting a lens from predetermined lenses having different light-transmitting filters, according to the environment in which said lens is to be used, comprising the following steps:-providing at least one image of said environment,-identifying a predetermined number of main colors on said at least one image, -estimating, for each primary color, a first value of at least one color component,-determining a modified color corresponding to the primary color seen through a first of the predetermined lenses,-calculating, for each modified color, a second value of the color component,-calculating, for each of the other predetermined lenses, a second value of the color component corresponding to the primary color,-calculating, for each of the other predetermined lenses, a second value of the color component corresponding to the primary color. The determining step and the calculating step are repeated,-for each main color, comparing said first value with each second value, and-deriving the selected lens therefrom.

Description

[0001] – Field of the Invention

[0002] The present invention generally relates to the field of glasses.

[0003] More specifically, the present invention relates to a method for selecting a lens from a set of predetermined lenses having different light transmission filters according to the environment in which the lens is to be used. Background Art

[0004] When a customer wants to buy sunglasses, the tint of the lens is usually already defined. This is why most sunglasses are selected based on the aesthetics of the frame rather than the solar filter applied. As a result, most of the sunglass lenses sold are grey, brown and grey-green.

[0005] Document WO 2019002416 proposes a solution for selecting lenses based on the wearer's needs rather than on appearance.

[0006] To select the best color (i.e., the best filter) for the sunglass lens, this document suggests taking into account the wearer's usual environment (usual location, usual activities, etc.).

[0007] Thanks to this solution, since each wearer practices specific activities in a typical environment, it is possible to select a lens that has a positive impact on the wearer's color vision.

[0008] In order to provide the wearer with the best sunglass lens, it is necessary to know the wearer's habits and preferences in terms of their specific activities and locations in their daily life.

[0009] The main difficulty then lies in defining the optimal parameters in order to select a sunglass lens that perfectly matches the wearer's color vision needs related to each environment. Summary of the Invention

[0010] In this context, the object of the present invention is to define the most suitable and personalized filter according to personal criteria or habits and more generally according to the environment in which the lens is to be worn.

[0011] To this end, the present invention provides a method for selecting a light transmission filter for a lens from a set of predetermined light transmission filters according to the environment in which the lens is to be used, the method comprising the following steps:

[0012] - providing at least one image of the environment,

[0013] - identifying a predetermined number of main colors on the at least one image,

[0014] - estimating a first value of at least one color component for each main color,

[0015] - For each main color, determine the modified color corresponding to the main color as seen through the lens of the first light transmission filter among the predetermined light transmission filters,

[0016] - Calculate a second value of the at least one color component for each modified color,

[0017] - For each of the other predetermined light transmission filters, repeat the determination step and the calculation step,

[0018] - For each main color and each predetermined light transmission filter, compare the first value with each second value, and

[0019] - Derive the selected light transmission filter therefrom.

[0020] In other words, the present invention includes:

[0021] - Simplify the complex color content of a typical outdoor or indoor environment (the environment selected by the wearer), for example by selecting six dominant colors of the scene of the environment,

[0022] - Calculate how these dominant colors are modified when observing the scene through different solar lenses, and

[0023] - Derive the ideal solar lens based on the difference between these dominant colors and the corresponding modified colors.

[0024] For example, the selected lens is the lens that least modifies the hue of the image and / or more modifies the chroma of the image.

[0025] In a variant, if the lens is to be used in a dazzling environment (snowy, maritime...), the lens can be a lens that better protects the eyes.

[0026] Due to the present invention, the lens prescription can recommend a stable filter for a given environment (landscape, activity...), or recommend a variable filter that can change according to the environment, such as an active filter that takes into account the periodic update of the surrounding colors.

[0027] Other preferred features of the present invention are as follows:

[0028] - The providing step includes acquiring different images of the environment, and during the identifying step, identifying the main colors on the different images.

[0029] - Before the identifying step, associate the different images into a single final image, and during the identifying step, identify the main colors on the final image.

[0030] - During the recognition step, the intermediate dominant colors are recognized on each different image, and the dominant color is selected from the intermediate dominant colors according to the proportion of each intermediate dominant color on the different images.

[0031] - The predetermined number is greater than two, preferably equal to six.

[0032] - The recognition step includes dividing the image into a predetermined number of region groups, each region group being defined by a dominant color representing the color of the region.

[0033] - During the recognition step, the provided image is divided into region groups by using a K-means process or a process derived from the K-means process.

[0034] - The provided image is blurred, for example, by a low-pass Gaussian filter, and the process is performed on the blurred image.

[0035] - The recognition step includes assigning a dominant color selected from a color list to each region group, and if two region groups in the region group are assigned to the same dominant color, the region groups are merged.

[0036] - The color components are determined according to the CIELAB color space or a color space derived from the CIELAB color space.

[0037] - The color component is color hue.

[0038] - During the comparison step, for each dominant color and for each predetermined light-transmitting filter, a first difference between the hue of the dominant color and the hue of the corresponding modified color is determined, and during the derivation step, the light-transmitting filter is selected according to the first difference.

[0039] - The color component is color chroma.

[0040] - During the comparison step, for each dominant color and for each predetermined light-transmitting filter, a second difference between the chroma of the dominant color and the chroma of the corresponding modified color is determined, and during the derivation step, the light-transmitting filter is selected according to the second difference.

[0041] - The lens is selected in the following two sub-steps: selecting one or more predetermined light-transmitting filters from all the predetermined light-transmitting filters for which all the first differences are less than a predetermined threshold, and selecting at least one predetermined light-transmitting filter from the selected predetermined light-transmitting filters for which the second difference is greater than the predetermined threshold, preferably the largest.

[0042] - At least one of the provided images is provided by a future wearer of the lens.

[0043] - At least one of the provided images is read from a database storing images respectively associated with a given environment.

[0044] - Images of many different environments are provided, and a transmissive filter is selected for each environment.

[0045] - The acquired images are of the RGB type. During the determining step, the modified color is determined by applying the spectral transmittance of a predetermined transmissive filter to each primary color, thanks to a color book of spectral reflectances.

[0046] - During the determining step, the acquired images are of the hyperspectral type. The modified color is determined by applying the spectral transmittance of a predetermined transmissive filter to each pixel of the acquired image to obtain a filtered image, and performing the recognition step on this filtered image. Detailed Description

[0047] The following description, given with reference to the accompanying drawings and by way of non - limiting examples, makes clear what is included in the present invention and how the present invention can be practiced.

[0048] In the drawings:

[0049] - Figure 1 is a flowchart showing different steps of the process according to the present invention,

[0050] - Figure 2A is a first example of an original image and a blurred image for implementing the Figure 1 process,

[0051] - Figure 2B is a second example of an original image and a blurred image for implementing the Figure 1 process,

[0052] - Figure 3 shows another original image, the regions into which the original image is divided, and the primary colors of these regions,

[0053] - Figure 4 is a graph showing the hue difference imposed by two different lenses (or lens groups) on the primary colors,

[0054] - Figure 5 is a graph showing the chroma difference imposed by two different lenses (or lens groups) on the primary colors, and

[0055] - Figure 6 represents a color system.

[0056] The present invention generally relates to protective eyewear suitable for protecting an individual's eyes from the effects of light. Herein, the protective eyewear is more particularly suitable for sunglasses, but in variants, it can also be suitable for lenses commonly used in indoor environments (lenses with blue light filters, lenses with contrast enhancement filters...).

[0057] Such lenses include a light transmission filter that must be selected based on the preferences of the individual.

[0058] To this end, the process according to the present invention proposes a method for selecting lenses with a filter that best suits the individual's usual environment.

[0059] Thus, for example, it can be considered that an individual (also called the future wearer of sunglasses) wants to buy sunglasses and the salesperson tries to find the most suitable sunglasses lenses.

[0060] To this end, different methods can be considered.

[0061] Hereinafter, the following example will be considered, according to which the salesperson has a catalog in which different sunglasses lenses are associated with a given environment (activity, landscape), such that the individual can select his sunglasses lenses, for example, according to the activity he is accustomed to practicing. In this case, using the process described below, the association between the lens filter and the environment is made upstream.

[0062] But alternatively, it is also possible to provide the individual with custom-made lenses, the filter of which will be selected from several filters, by using the method described below.

[0063] This method is executed by a processing unit that includes a CPU, a memory, and different input and output interfaces.

[0064] The processing unit is suitable for receiving images due to its input interface. The processing unit is suitable for allocating lenses to each environment due to its output interface.

[0065] The processing unit stores a computer application program due to its memory, which consists of a computer program including instructions, and the execution of these instructions by the processor enables the processing unit to implement the process described below.

[0066] Now the process of allocating lenses (i.e., light transmission filters) to a given environment can be described in detail.

[0067] The process is executed in a number of successive steps as Figure 1 shown.

[0068] The first step S1 of the process includes providing at least one image of a specific environment.

[0069] The image is preferably characteristic of the environment in which the sunglasses are to be worn.

[0070] The concept of "environment" can be defined as a typical situation (landscape, habitual view...).

[0071] The environment can correspond to a given geographical location, or to a given object, or to a series of similar objects located at different places or viewed from different angles.

[0072] In a given environment, the same objects (trees, mountains, buildings...) and / or a similar average brightness (even if the colorimetry may be different) are usually found.

[0073] For example, the countryside (nature), forest, mountains, city (urban), roads, snow, water form different environments.

[0074] And in a more general way, the environment can be defined as a situation depending on landscape data or personal criteria.

[0075] By way of example, the environment can be associated with a specific activity (golf, car driving, tennis, skiing, etc.) or a specific habit (the usual route of a bus driver...).

[0076] The environment can be defined by an image and characterized by a set of colors.

[0077] Therefore, at this stage, the idea is to obtain at least one typical image of the environment.

[0078] Preferably, a series of several typical images of the environment are obtained.

[0079] Different embodiments can be envisaged to carry out this first step S1.

[0080] In a first embodiment, the image is provided by the customer who has taken a photo of the environment in which he will use his future sunglasses.

[0081] In a second embodiment, an image of a given environment is provided by the owner of the processing unit who has stored typical images of different environments in the memory of the processing unit.

[0082] In a third embodiment, the image is automatically read in a database which is connected to the processing unit via the Internet for example and stores images respectively associated with a given environment by metadata or tags.

[0083] In a fourth embodiment, the image is read in a database which is connected to the processing unit via the Internet for example, and the image is manually selected according to the environment chosen by the individual.

[0084] The first step S1 may also be performed by combining some of these embodiments.

[0085] In each embodiment, the acquired images may be of the hyperspectral type or of the RGB type (ideally without any post-processing).

[0086] If the image is of the hyperspectral type, it is first converted to RGB by a color appearance model, such as the color appearance model named iCAM06.

[0087] This model is described, for example, in the document “Kuang, J., Johnson, GM and Fairchild, M. (2007). iCAM06: A refined image appearance model for HDR image rendering. Journal of Visual Communication and Image Representation, 18(5). https: / / doi.org / 10.1016 / j.jvcir.2007.06.003”.

[0088] When the images are acquired, the processing unit determines the dominant colors of the images (here, the predominant colors of these images).

[0089] The concept of dominant color can be defined as the color that is present most in the image.

[0090] In order to identify a predetermined number of dominant colors on an image, the processing unit divides the image into a predetermined number of area groups (also called "clusters"), in each area group one of the dominant colors represents the color of an area of ​​the cluster.

[0091] First, it may be described in detail how this operation is performed when only a single image of the environment is acquired.

[0092] The second step S2 of the process consists in undoing any gamma correction of the acquired image.

[0093] This gamma correction is usually applied to the picture to be displayed faithfully on the screen. Here, this correction is removed in order to obtain the original image.

[0094] In the subsequent step S3, the RGB original image is transformed into a new image encoded in a determined color space. Due to this step, the color of each pixel of the image can be transformed into a color space that can more easily process the color.

[0095] The idea is to characterize the color of an image by means of parameters that are easy to find (hereinafter referred to as "components").

[0096] Here, the CIELAB color space is preferably selected over other color spaces, not only because the color distribution is simple, uniform and consistent, but also because of the perceptual effectiveness of CIELAB color differences.

[0097] As Figure 6 shown in the upper left corner of

[0098] In this color system, a color can be defined by three components: the lightness or brightness labeled L* (0 represents black and 10 represents white) and two coordinates labeled (a*, b*) or (H*, C*).

[0099] In this CIELAB system, the value difference ΔL*, the hue difference ΔH*, and the chroma difference ΔC* are normalized and easy to interpret.

[0100] However, the CIELAB color space requires knowledge of the white point of the illuminant, which can be a problem for images with an unknown illuminant.

[0101] Therefore, in the third step S3, the processing unit first estimates the illuminant.

[0102] For this purpose, there are various illuminant estimation methods.

[0103] The white-patch Retinex algorithm and the gray-world algorithm (Buchsbaum, 1980; Land, 1977) are commonly used algorithms, but they are prone to producing large estimation errors (Hordley 2006). Another method that exists identifies the bright and dark pixels of an image based on the distance that the pixels of the image are from the average color of the scene. Then, principal component analysis (PCA) is performed on the bright and dark pixels. Thus, the first component of the PCA is the estimated illuminant. This method is described in the literature "Cheng, D., Prasad, D. K., and Brown, M. S. (2014). Illuminant estimation for color constancy: Why spatial-domain methods work and the role of the color distribution. Journal of the Optical Society of America A, 31(5), 1049. https: / / doi.org / 10.1364 / JOSAA.31.001049".

[0104] Another method is described in the literature "Hordley, S. D. (2006). Scene illuminant estimation: Past, present, and future. Color Research & Application, 31(4), 303–314. https: / / doi.org / 10.1002 / col.20226".

[0105] Then, in the fourth step S4, due to the illuminant white point, the linear image is converted into an LAB image (i.e., an image characterized by three components L*, a*, b* or L*, H*, C*).

[0106] Then, during the fifth step S5, the LAB image is filtered in order to obtain a blurred image that is easier to process. The blur intensity is adjusted to reduce the saliency of the edges and local differences in the image while preserving the overall distribution of the colors in the image.

[0107] Here, as Figure 2A and Figure 2B shown, in order to visually explain the effect of blurring, for two images Img 0 , Img' 0Apply the discrete Fourier transform while gradually increasing the blurring intensity. Figure 2A and Figure 2B shows that as the blurring intensity of the filter increases, the high-frequency components of the images Img 0 , Img' 0 effectively decrease.

[0108] The selected filter is a low-pass Gaussian filter, and its blurring intensity is denoted as σ.

[0109] In Figure 2A and Figure 2B , the images Img 0.5 , Img' 0.5 correspond to the images blurred with a blurring intensity σ of 0.5, and the images Img 1 , Img' 1 correspond to the images blurred with a blurring intensity σ of 1...

[0110] Here, the blurring intensity σ is selected to be equal to 8 (which corresponds to the images Img Figure 2A and Figure 2B ) of 8 , Img' 8 .

[0111] Gaussian blurring is achieved by convolving the image with a Gaussian bell-shaped kernel, as described in Mordinstov&K, 2013. In this document, the blurring intensity σ corresponds to the standard deviation of the distribution. The standard deviation controls the variance around the mean of the Gaussian distribution.

[0112] This fifth step is essential for accelerating the convergence of the algorithm for finding the main color of the image. In fact, blurring with a relatively high blurring intensity σ reduces the significance of local differences, thus highlighting the overall color trend of the image.

[0113] Then, the blurred LAB image is ready for the clustering process, which aims to divide the image into a predetermined number of groups of regions (clusters), in each of which the main color dominates.

[0114] The idea of this process actually consists of distributing the colors present in the image among clusters of various similar colors and extracting the dominant color of the scene.

[0115] To this end, during the sixth step S6, the processing unit uses the K-means algorithm or an algorithm derived therefrom.

[0116] Here, the K-means++ algorithm is used to segment the image into a predetermined number of clusters (as described in Arthur&Vassilvitskii, 2007).

[0117] The predetermined quantity is at least equal to two. The predetermined quantity is preferably greater than three and is equal to six herein.

[0118] In other words, the algorithm includes dividing the image into different regions, each region being associated with one of the six colors.

[0119] The algorithm will not be described in detail herein, but the following explanations can be provided.

[0120] The K-means algorithm randomly locates six initial seed points and works in a loop to try to divide the image into clusters.

[0121] The K-means algorithm randomly determines all the seed points. This results in the initialization of center points that are far apart, leading to poor results and time-consuming clustering.

[0122] Herein, the K-means++ algorithm determines the first seed point by random assignment, but carefully determines the remaining seed points to maximize the distance between the center points. This method takes longer in initialization, but the clustering process has been proven to be faster than the original K-means clustering, thus reducing the overall time taken for convergence.

[0123] This algorithm is described in the literature "Aubaidan. (2014). Comparative study of k-means and k-means++ clustering algorithms on crime domain [Comparative study of k-means and k-means++ clustering algorithms on crime domain]. Journal of Computer Science [Journal of Computer Science], 10(7), 1197–1206. https: / / doi.org / 10.3844 / jcssp.2014.1197.1206".

[0124] The algorithm depends on the determination of six clusters. Each cluster corresponds to a group of regions where the dominant color prevails.

[0125] Therefore, the six resulting clusters enable the identification of the spatial positions of the color clusters in their original state and, for example, the calculation of the distribution of each cluster (in terms of percentage of pixels).

[0126] Given the distribution of each cluster on the blurred image, for each cluster, the median sRGB triple corresponding to the dominant color of the cluster is calculated on the RGB image (or on the sRGB image which is an image encoded in standard RGB). Then, this sRGB triple is converted into another triple formed by three components L*, a*, b* or L*, H*, C*.

[0127] During the seventh step S7, the processing unit may identify six "near primary colors", i.e., six colors that are very close to the six primary colors and are clearly defined in a dictionary (or "color palette") that records a limited number of colors.

[0128] To this end, the processing unit compares each found primary color with predefined colors recorded in the dictionary. Then, each found primary color is approximated to its closest predefined color (named "standard primary color").

[0129] More specifically, the triple of each found primary color is compared with the standard data of the color system. Here, the color system is the Munsell color system, as Figure 6 shown. This system is chosen because it fits perfectly into the CIELAB color space.

[0130] More precisely, the triple is compared with the ISCC-NBS color dictionary in order to try to find for each found primary color the closest color triple (and its name) as defined by the CIEDE2000 formula (Cobeldick, 2019; Judd & Kelly, 1939; Sharma et al., 2005).

[0131] Here, it can be noted that the dictionary uses 13 basic color names for the first level, and 29 intermediate color categories form a more refined second level, while 20 adjectives (such as vivid, dull, bright, moderate, etc.) form the most refined third level. Finally, the final dictionary contains 267 different color names.

[0132] At this stage, the names of the six standard primary colors of the image are identified.

[0133] Then, during the eighth step S8, the processing unit checks whether two or more clusters are associated with the same color. If two different found primary colors are approximated to the same standard color, redundancy may indeed occur.

[0134] If there is redundancy, the two clusters are merged to form a new cluster (step S81).

[0135] Then, during step S82, the percentage ratio of the color distribution is updated.

[0136] If no redundancy is detected, the process continues to the next step S9.

[0137] Before describing this step S9, reference can be made to Figure 3 .

[0138] Figure 3An example of the acquired RGB image 10 is shown. This image does not represent the environment but is shown to clearly demonstrate the clustering operation.

[0139] On the right side of this image 10, six clusters 11 to 16 obtained by the above process are shown. It seems that each cluster corresponds to a group of regions with similar colors.

[0140] Below this image 10, a bar chart 17 is represented, which shows the distribution percentages of each primary color. On the right side, the box 18 gives the names of six standard primary colors. Finally, the color wheel 19 shows the coordinates of these standard primary colors in the a*, b* coordinate system.

[0141] At this time, it can be noted that several images are submitted to a group of people, and these people are required to try to extract the six primary colors of each image. The results of this experiment prove that the above algorithm very accurately finds the standard primary colors of each image compared with the perception of these people. For example, see "Aiman Raza, Sophie Jost, Marie Dubail and Dominique Dumortier (2021). Automatic colour segmentation and colour palette identification of complex images, Journal of the International Color Association (2021): 26, 11 - 21".

[0142] The above algorithm is described in the case of a single image of the considered environment.

[0143] However, in a preferred embodiment where several different typical images of a given environment have been acquired, the process is carried out slightly differently.

[0144] These six standard primary colors are indeed identified based on all the images of the environment.

[0145] In this embodiment, the acquired images can be associated into a single large image (not by superimposing the images but by aggregating the images). The images can be associated side by side, or one above the other, or in a matrix shape. Then, steps S2 to S8 can be performed on this large image.

[0146] In a variant, steps S2 to S8 can be performed on each image to find six clusters associated with six "intermediate standard primary colors" for each image.

[0147] Then, the final six standard primary colors are calculated based on the proportions of each intermediate standard primary color on the different images. For example, the final six standard primary colors can be the most representative colors on the image.

[0148] Here, it can be noted that the analysis of a single image and the analysis of multiple images yield different dominant colors. To fully represent the environment and not focus on specific objects, it is preferable to process multiple images, each of which represents a part of the overall environment. Additionally, specific objects or elements (such as skin color or the main object of an activity) can be added as additional images or directly as dominant colors.

[0149] At this stage, the hue and chroma of each standard primary color are known.

[0150] Then, during the ninth step S9, the processing unit determines the influence of at least two different filters (i.e., in this example, at least two different pairs of sunglasses) on the image and / or the standard primary colors.

[0151] Sunglasses are different in the sense that their filters have different effects on the hue and / or chroma of at least one color when viewing through the lenses.

[0152] For example, the lens filter can be defined by the variation of transmittance with wavelength.

[0153] To better understand the method, first, a first pair of sunglasses can be considered.

[0154] To determine the influence of this pair of sunglasses, the processing unit can calculate the influence of the sunglasses on the standard primary colors in order to find six "modified colors", or calculate the influence of the sunglasses on the original RGB image (in this case, repeat steps S2 to S8 for this image filtered by the sunglasses in order to find six "modified colors").

[0155] These calculations can be detailed using two examples.

[0156] In the first example, each acquired image is of the hyperspectral type. In other words, each pixel is associated with the variation of radiance with wavelength.

[0157] To determine the influence of the sunglasses on the image, the spectral transmittance of the sunglasses is applied to each pixel to obtain a filtered image.

[0158] In other words, for each pixel, its radiance variation is combined with the transmittance variation of the sunglasses in order to obtain a filtered hyperspectral image.

[0159] Then, the filtered image is converted to RGB by the color appearance model iCAM06. Thereafter, steps S2 to S8 can be performed again on the RGB image to find six modified colors.

[0160] In the second example, the acquired image is of RGB type.

[0161] To find the six modified colors, the processing unit applies the spectral transmittance of the sunglasses to each of the six standard primary colors (instead of applying to the original image). The spectral reflectance of the dominant colors is extracted from a color book (such as the Munsell color book). Then the transmittance of the lens is applied, and the resulting colors are calculated.

[0162] This transformation is performed, for example, by using the standard XYZ to CIE-L*a*b* matrix or by the CIECAM02 process (as disclosed in "CIE159.(2004).A colour appearance model for colour management systems:CIECAMO2 [for color management systems: CIECAMO2].CIE Central Bureau [CIE Central Bureau]").

[0163] How this transformation is performed can be briefly explained.

[0164] Considering the triple (R, G, B) of the standard primary colors, the triple (R c G c B c ) of the modified colors can be calculated by using the following calculation:

[0165]

[0166] In this equation, D refers to the degree of chromatic adaptation and is between 0 (no adaptation) and 1 (full adaptation). The degree of chromatic adaptation is calculated by the following equation:

[0167]

[0168] α = 1 / R w

[0169] β = 1 / G w

[0170] λ = 1 / B w

[0171] Here, F is a constant selected according to the surrounding environment, and L A corresponds to the luminance of the adaptation field.

[0172] The subscript w refers to the color of the white dots (original light emitters).

[0173] At this time, the six standard primary colors (of the original image) and the six modified colors (of the image seen through the solar lens) are known.

[0174] Therefore, the processing unit knows the hue H* and chroma C* of all these colors.

[0175] Then, during the tenth step S10, the processing unit compares each primary color with the corresponding modified color by analyzing the differences in hue and chroma.

[0176] To this end, the processing unit calculates the hue difference ΔH* between the hue H* of the standard primary color of each cluster and the hue H c * of the modified color.

[0177] The processing unit also calculates the chroma difference ΔC* between the chroma C* of the standard primary color of each cluster and the chroma C c * of the modified color.

[0178] Steps S9 and S10 are repeated using at least one other solar lens in order to determine another hue difference ΔH*' and another chroma difference ΔC*' for each standard primary color.

[0179] In practice, these steps are repeated using a much larger number of other solar lenses (and filters).

[0180] Figure 4 The curves G1, G2 shown respectively show the hue differences ΔH*, ΔH*' calculated for the six primary colors for two different solar lenses. Figure 5 The curves G3, G4 shown in show the chroma differences ΔC*, ΔC*' calculated for these six primary colors for two different solar lenses.

[0181] Then, during the eleventh step S11, one of the solar lenses is selected.

[0182] To this end, the aim can be to find the solar lens that has the weakest effect on the color of the environment (due to hue calculation) and the best effect on the saturation of the scene viewed through the lens (due to chroma calculation).

[0183] Here, it can be noted that the process of selecting this solar lens can depend on the personal criteria of the future wearer. For example, this selection can take into account his preferences according to whether the future wearer prefers to see natural colors or prefers to see colors with saturation. If he is light-sensitive, this information can also be taken into account. Hereinafter, it will be assumed that he likes to view colors with high saturation.

[0184] In fact, a large hue shift indicates a large color distortion, and a large chroma shift indicates a large saturation.

[0185] This step S11 is preferably performed in two sub-steps.

[0186] During the first sub-step, a first lens group (or a first filter group) is selected from all the considered solar lenses (or filters).

[0187] Here, the selected lens is the lens that minimizes the hue shift or the lens whose hue shift does not exceed a predetermined threshold.

[0188] For this purpose, solar lenses can be selected for which the hue differences ΔH*, ΔH*' calculated for all primary colors are the smallest and ideally less than a predetermined threshold. The absolute value of this threshold can be equal to 10, or more preferably equal to 5. Thus, in Figure 4 the example, only one of the two solar lenses is selected (the solar lens corresponding to curve G2).

[0189] In a variant, a K-means analysis can be performed to cluster the lenses according to the hue differences ΔH*, ΔH*' calculated for all primary colors of the lenses. In another variant, the selected solar lenses will be those for which the sum (or maximum value) of the hue differences ΔH* for six colors is below the threshold. This threshold can be pre-determined or calculated based on all the calculated sums.

[0190] Then, the second sub-step includes selecting the lens from the selected lens group that will give the best visual experience.

[0191] For this purpose, the color attribute considered is chroma. The selected solar lens will be, for example, the lens that increases the chroma the most.

[0192] In other words, the selected solar lens is the lens that presents the highest chroma difference ΔC* for the largest number of dominant colors (e.g., by considering the sum of these differences for six colors).

[0193] In Figure 5 the example, the lens with the largest sum of chroma differences is the lens corresponding to curve G4. Thus, this lens can be recommended to future wearers for the considered environment.

[0194] In a variant, the lens can only be selected if all six chroma differences are positive.

[0195] In another variant, the selected lens can be a lens for which all the hue differences and chroma differences follow a mathematical law, and this mathematical law can be:

[0196] |ΔE*| < 6 & ΔC* > 0, or

[0197] |ΔE*| < 2 & ΔC* > 0.

[0198] When selecting lenses for one environment, all the processes can be repeated to find other sunglasses lenses suitable for other environments.

[0199] The present invention is in no way limited to the embodiments described and shown.

[0200] In the above main embodiment, the lenses are compared with each other to select only one lens. However, in a variant, if several lenses have similar filters, they can be combined into a lens cluster. Thus, at the end of the process, a lens cluster is selected. Therefore, several lenses can be recommended for a given environment. If too many lenses are recommended, a larger number of clusters can be used to reproduce the process.

[0201] In the above main embodiment, one lens is selected from several lenses according to the selected environment. However, in a variant, lenses can be manufactured with filters that are optimal for the environment under consideration. Such lenses are molded by using monomers or polymers doped with a tint. Thus, the spectral transmittance of the selected tint can be defined as the spectral transmittance that transmits more of the wavelengths that make up the six primary colors. Several dyes can be selected to obtain this tint. After a combination of specific ratios, these dyes will induce the spectral transmittance of the lens. By maximizing a function, the concentration of each dye can be selected to achieve the highest transmittance for each dominant color. This function can be selected to additionally match constraints such as the TV range (sunglasses categories 1, 2, or 3), the minimum transmittance for each band, and the maximum transmittance for each band (here, a band refers to a spectral characteristic that can be selected, such as short-wavelength filtering for retinal protection, minimum transmittance following the Q-signal norm for driving, etc.).

Claims

1. A method for selecting a light-transmitting filter for a lens from a predetermined set of light-transmitting filters according to the environment in which the lens is to be used, the selection method comprising the steps of: - providing at least one image (10) of the environment, - Identify a predetermined number of dominant colors (C 1 - C 6 ) on the at least one image - estimating a first value of at least one color component (H*, C*) for each primary color, - For each primary color (C 1 - C 6 ) determine the modified color corresponding to the primary color as seen through the lens with the first light transmissive filter of the predetermined light transmissive filters, - calculating a second value of the color component (H*, C*) for each modified color, - repeating the determination step and the calculation step for each of the other predetermined light-transmitting filters, - comparing the first value with each second value for each primary color and each predetermined light-transmitting filter, and - deriving the selected light-transmitting filter therefrom.

2. The selection method according to claim 1, wherein: - the providing step comprises obtaining different images (10) of the environment, and - during the recognition step, the primary colors are recognized on the different images (10).

3. The selection method according to claim 2, wherein: - before the recognition step, the different images (10) are associated into a single final image, and - During the recognition step, identify the main color (C 1 - C 6 ) on the final image.

4. The selection method according to claim 2, wherein, During the recognition step, an intermediate dominant color is recognized on each different image (10), and the dominant color (C 1 -C 6 ) is selected from the intermediate dominant colors according to the proportion of each intermediate dominant color on the different images (10).

5. The selection method according to any one of claims 1 to 4, wherein, the predetermined number is greater than two, preferably equal to six.

6. The selection method according to any one of claims 1 to 5, wherein, the recognition step comprises dividing the image into a predetermined number of region groups, each region group being defined by a primary color representative of the color of the region.

7. The selection method according to claim 6, wherein, during the recognition step, the provided image is divided into region groups by using a K-means process or a process derived from the K-means process.

8. The selection method according to claim 7, wherein, For example, the provided image (10) is blurred by a low-pass Gaussian filter, and the process is performed on the blurred image (Img 8 ).

9. The selection method according to any one of claims 6 to 8, wherein, The recognition step includes assigning a main color (C 1 -C 6 ) selected from the color list to each region group, and wherein, if two region groups in the region group are assigned to the same main color, the region groups are merged.

10. The selection method according to any one of claims 1 to 9, wherein, the color component (H*, C*) is determined according to the CIELAB color space or according to a color space derived from the CIELAB color space.

11. The selection method according to any one of claims 1 to 10, wherein, the color component (H*, C*) is color hue or color chroma.

12. The selection method according to claim 11, wherein, the color component (H*) is color hue, and wherein, During the comparison step, for each primary color (C 1 -C 6 ) and for each predetermined light-transmitting filter, a first difference (ΔH*) between the hue (H*) of the primary color (C 1 -C 6 ) and the hue of the corresponding modified color is determined, and during the derivation step, the light-transmitting filter is selected based on the first difference (ΔH*).

13. The selection method according to claim 11, wherein, The color component (C*) is the color chromaticity, and wherein, during the comparison step, for each primary color (C 1 -C 6 ) and for each predetermined light-transmitting filter, a second difference (ΔC*) between the chromaticity (C*) of the primary color (C 1 -C 6 ) and the chromaticity of the corresponding modified color is determined, and during the filter derivation step, the light-transmitting filter is selected based on the second difference (ΔC*).

14. The selection method according to claims 12 and 13, wherein, the lens is selected in the following two sub-steps: - selecting one or more predetermined light-transmitting filters from all the predetermined light-transmitting filters for which all the first differences (ΔH*) are less than a predetermined threshold, and - selecting from the selected predetermined light-transmitting filters at least one predetermined light-transmitting filter for which the second difference (ΔC*) is greater than a predetermined threshold, preferably the largest.

15. The selection method according to any one of claims 1 to 14, wherein, The at least one provided image (10) is provided by a future wearer of the lens or is read in a database storing images respectively associated with a given environment.

16. The selection method according to any one of claims 1 to 15, wherein, a plurality of images (10) of a plurality of different environments are provided, and wherein a light-transmitting filter is selected for each environment.

Images