A method and device for estimating skin roughness using imaging method

Abstract

The present invention discloses a method and device for estimating skin roughness by imaging, comprising the following steps: S1, obtaining a set of low-resolution original grayscale images of the skin; S2, reconstructing a high-resolution grayscale image H using a set of low-resolution original grayscale images {O(t), t=1,2,3…N}; S3, calculating texture width based on the high-resolution grayscale image; S4, establishing a corresponding relationship model based on texture width data and skin roughness; S5, automatically estimating and realizing skin roughness distribution in a large field of view for any user. Oblique incident light is combined with a reconstruction algorithm to obtain a skin image with high spatial resolution, texture structure is analyzed and texture data is calculated based on a high-resolution skin map, and a texture and roughness correspondence model is constructed based on empirical data, and finally, skin roughness estimation can be realized automatically for any user.

Description

Technical Field

[0001] The present invention relates to the field of medical cosmetology technology, and in particular to a method and device for estimating skin roughness using an imaging method. Background Art

[0002] With the development of modern society, people pay more and more attention to the management of body skin. It is particularly important to detect the skin condition while maintaining the skin. At present, most people in China use subjective consciousness to judge the quality of skin, which has a large deviation. Therefore, in recent years, how to quantitatively detect and analyze skin roughness has attracted more and more attention, especially in the medical and beauty fields.

[0003] The commonly used skin roughness technology is the photoelectric scanning method to directly test the test skin, but it is a bit expensive; there are also studies that use optical imaging methods to estimate skin roughness, but their accuracy and skin observation field are limited. Since the imaging method is basically passive, it will not cause harm to the skin. At present, domestic manufacturers have not yet developed imaging skin roughness detection instruments and devices.

[0004] In the existing imaging image roughness estimation methods, the design of imaging systems and devices is generally not involved. All work is done based on images after imaging. As described in the research paper "Quantitative Evaluation of Face Skin Roughness and Its Application in Medical Cosmetology" (Dissertation: Chen Jin. Quantitative Evaluation of Face Skin Roughness and Its Application in Medical Cosmetology. University of Electronic Science and Technology, 2009), it studies the texture roughness based on the co-occurrence matrix, the roughness based on the Tamura texture feature, etc., which has great reference value. As described in the patent document "Skin Surface Roughness Detection Method Based on Image RGB Color Space" (Patent: Liu Ying, Qiu Xianrong. Skin Surface Roughness Detection Method Based on Image RGB Color Space, CN107157447A[P]. 2017.), it processes the macro skin digital image with the same resolution, directly uses the color space pixel value of the skin image to calculate the mean of the absolute value of the deviation, and uses it as the roughness feature value of the skin image to identify the skin roughness. According to the research paper "Texture Feature Analysis and Classification of Human Arm Skin Images" (Dissertation: Hu Liangjun. Texture Feature Analysis and Classification of Human Arm Skin Images. Fujian Normal University, 2017.), we seek to use a method combining Tamura feature method and wavelet transform to obtain the roughness of skin images. From the perspective of existing imaging technology, the current image resolution is not high enough, resulting in insufficient accuracy of skin roughness. However, if a microscope or other methods are used, the accuracy is sufficient, but the field of view is not enough, and the skin roughness distribution in a large field of view cannot be given at one time.

[0005] The current imaging method detects and analyzes local areas of skin images. Therefore, how to accurately estimate the roughness of the skin while ensuring the observation of a large field of view of the skin image has become an urgent problem to be solved. Summary of the invention

[0006] The purpose of the present invention is to propose a method and device for estimating skin roughness using an imaging method, which uses oblique incident light combined with a reconstruction algorithm to obtain a skin image with high spatial resolution, analyzes the texture structure and calculates texture data based on the high-resolution skin image, and constructs a texture and roughness correspondence model based on empirical data. Ultimately, the skin roughness can be automatically estimated for any user.

[0007] In order to solve the above technical problems, the technical solution of the present invention is:

[0008] A method for estimating skin roughness using an imaging method comprises the following steps:

[0009] S1. Get a set of low-resolution original grayscale images of the skin

[0010] By receiving the incident light and the reflected light from the surface of the skin to be tested, a low-resolution original grayscale image I is obtained, and by controlling the angle of the incident light, a series of low-resolution original grayscale image sets {O(t), t=1,2,3…N} are obtained respectively;

[0011] S2, reconstructing a high-resolution grayscale image H using a set of low-resolution original grayscale images {O(t), t=1, 2, 3…N};

[0012] S3, calculating texture width based on high-resolution grayscale image;

[0013] S4. Establishing a corresponding relationship model based on texture width data and skin roughness

[0014] Using a roughness measuring instrument, on the one hand, the roughness of a certain area of ​​the skin is measured, and on the other hand, the texture width data of the corresponding skin image area to be measured is calculated to form at least 100 sets of {roughness, texture width} data pairs, and then a corresponding relationship model between skin roughness and texture width is constructed;

[0015] S5. Automatically estimate the skin roughness distribution of any user in a large field of view

[0016] For any user, by repeating steps S1-S4, the roughness distribution of the skin to be measured can be obtained, corresponding to the skin in different areas under a large field of view, so that the high-precision roughness distribution of the skin to be measured under a large field of view can be obtained at one time.

[0017] Preferably, the step S2 includes the following sub-steps:

[0018] S2-1, Spectrum of the initial solution of the high-resolution grayscale image H

[0019] Perform linear interpolation H = L(I) on the low-resolution original grayscale image I, and obtain its frequency-domain expression ξ{H}, where L{·} is the linear interpolation operation and ξ{·} is the Fourier transform operation, F H = ξ{H} is the spectrum of the initial solution of the high-resolution image, and the initial t = 1;

[0020] S2-2 Calculate the amplitude distribution map of the skin corresponding to incident light at a certain angle

[0021] For incident light at a certain angle, the resolution original grayscale image corresponding to this angle is O(t), and the corresponding spectral domain aperture function is P(t). Then the corresponding amplitude distribution image of the skin is:

[0022]

[0023] where ξ -1 {·} is the inverse Fourier transform operation, and P(t) is the spectral domain aperture function;

[0024] S2-3 Update the spectrum of the high-resolution image

[0025] For this angle in step S2-2, use the corresponding original grayscale image O(t) to update the gray intensity part of its amplitude image

[0026]

[0027] where |·| is the modulus operation,

[0028] Calculate the solution of the updated high-resolution image, and its spectrum F' H :

[0029]

[0030] In the formula, λ is the convergence step size, which is generally a constant;

[0031] S2-4 Iterative update

[0032] (1) If t < N, let t = t + 1, and jump to step S2-2 to continue executing steps S2-2 to S2-4;

[0033] (2) If t = N, then judge convergence, that is, evaluate the closeness of |F' H - F H |, and judge ∑|F' H - F H|<0.1 is true, if true, it is very close, it is considered to have converged, and the high spatial resolution skin amplitude distribution image H' is calculated, that is, H'=ξ -1 {F' H};∑|F' H -F H |<0.1 is not true, then continue to jump to step S2-2 and continue to execute steps S2-2 to S2-4;

[0034] S2-5. Obtain the final updated high-resolution grayscale image H through a modulo operation.

[0035] Preferably, the method for calculating the texture width in step S3 is as follows:

[0036] S3-1. Divide the high-resolution grayscale image H into N H Sub-areas {H i ,i=1,2,…N H}, for different regional subgraphs H i , calculate their texture width respectively;

[0037] S3-2. Design gradient operator:

[0038] Horizontal (image width direction) operator

[0039] Vertical (image height direction) operator

[0040] S3-3. Use G x Convolution H i , combined with the threshold T H , obtain the texture image E;

[0041] S3-4, for any texture point E in E k , find the extreme point in the same row nearby in H corresponding to the position;

[0042] S3-5, according to the pixel distance between the extreme points in step S3-4, it is defined as the texture point E k Width(k);

[0043] S3-6, continue with steps S3-4 and S3-5, find the widths corresponding to all texture points, and obtain Width(k), k = 1, 2...M x , M x is the number of texture points;

[0044] S3-7, the average texture width in the horizontal direction is:

[0045] S3-8, Similarly, replace G in S3-3x Replace with G y , the same row in S3-4 is replaced by the same column, and M in S3-6 x Replace with M y , along steps S3-3 to S3-6, the average texture width in the vertical direction is obtained

[0046] S3-9. The average width of the texture that measures the skin texture structure is defined as:

[0047] Width=(Width x +Width y ) / 2

[0048] By analyzing N H sub-regions, for a high-resolution grayscale image H, we can obtain N H The average width of the texture i ,i=1,2,…N H}.

[0049] Preferably, in step S4, the corresponding relationship model is:

[0050] Ra=a 1 (Width) 2 +a 2 (Width)+a 3

[0051] where a 1 ,a 2 ,a 3 is an undetermined parameter. According to the data pair of {roughness, texture width}, the parameter a is obtained by least squares training. 1 ,a 2 ,a 3 .

[0052] Preferably, the spectral domain aperture function P(t) is determined by the position of the incident light relative to the center.

[0053] The present invention also provides a device for estimating skin roughness using an imaging method, comprising an acquisition device and a computer controller, wherein the computer controller comprises a memory, a processor, and a computer program stored in the memory and capable of executing the above-mentioned method for estimating skin roughness using an imaging method on the processor.

[0054] Preferably, the acquisition device includes an objective lens disk and a grayscale camera, a plurality of LED particles are provided on one side of the objective lens disk, the processor outputs a control signal to control the opening and closing of the LED particles, a light hole is provided at the center of the objective lens disk, an imaging objective lens is provided at the light hole, the imaging objective lens and the grayscale camera are on the same imaging optical axis, and the image acquired by the grayscale camera is transmitted to the processor.

[0055] Preferably, the plurality of LED particles are evenly distributed around the light-through hole and arranged in a divergent shape.

[0056] Preferably, a plurality of LED light rings are arranged outwardly along the edge of the light-through hole on one side of the objective lens disk, and a plurality of LED particles are equally divided on each of the LED light rings.

[0057] Preferably, the processor controls the opening and closing of LED particles on the LED light ring in sequence from the inside to the outside.

[0058] The present invention has the following characteristics and beneficial effects:

[0059] By adopting the above technical scheme, a high-resolution skin image is reconstructed by multiple imaging with oblique incident light, and the skin roughness is measured by texture width, which is accurately matched with the actual roughness. The skin roughness estimation can be accurately given over a large range at one time to form a roughness distribution, which solves the problems of small field of view and unstable model of traditional imaging methods, overcomes the shortcomings of insufficient precision of imaging-based methods, and effectively improves the accuracy of skin detection, providing convenience for users to conduct skin aging research, beauty status and effect analysis, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0061] Figure 1 The figure is a flow chart of a method for estimating skin roughness using imaging method.

[0062] Figure 2 The figure is a schematic diagram of the structure of a skin roughness estimation device using an imaging method.

[0063] Figure 3 for Figure 1 Schematic diagram of the structure of the objective lens disk.

[0064] In the figure, 1-skin to be tested, 2-objective lens disk, 201-LED particles, 202-imaging objective lens, 203-light hole, 3-grayscale camera, 4-computer controller. DETAILED DESCRIPTION

[0065] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0066] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0067] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood by specific circumstances.

[0068] The present invention provides a method for estimating skin roughness using imaging method, such as Figure 1 As shown, the following steps are included:

[0069] S1. Get a set of low-resolution original grayscale images of the skin

[0070] By receiving the incident light and the reflected light from the surface of the skin to be tested, a low-resolution original grayscale image I is obtained, and by controlling the angle of the incident light, a series of low-resolution original grayscale image sets {O(t), t=1,2,3…N} are obtained respectively;

[0071] It is understandable that by controlling the angle of the incident light, the angle of the irradiated skin position is different. According to the principles of physical optics and Fourier optics, the physical frequency of the skin acquired by imaging is related to the angle of the illuminating incident light here. Within the imaging field of view, the larger the oblique incident angle, the higher the acquired spatial frequency. The present invention can acquire low-resolution original grayscale images corresponding to different frequencies by irradiating light at different angles, thereby laying the foundation for achieving the acquisition of high-resolution images.

[0072] S2, reconstructing a high-resolution grayscale image H using a set of low-resolution original grayscale images {O(t), t=1, 2, 3…N};

[0073] S3, calculating texture width based on high-resolution grayscale image;

[0074] S4. Establishing a corresponding relationship model based on texture width data and skin roughness

[0075] Using a roughness measuring instrument, on the one hand, the roughness of a certain area of ​​the skin is measured, and on the other hand, the texture width data of the corresponding skin image area to be measured is calculated to form at least 100 sets of {roughness, texture width} data pairs, and then a corresponding relationship model between skin roughness and texture width is constructed;

[0076] S5. Automatically estimate the skin roughness distribution of any user in a large field of view

[0077] For any user, by repeating steps S1-S4, the roughness distribution of the skin to be measured can be obtained, corresponding to the skin in different areas under a large field of view, so that the high-precision roughness distribution of the skin to be measured under a large field of view can be obtained at one time.

[0078] Specifically, step S2 includes the following sub-steps:

[0079] S2-1. Spectrum of the initial solution of high-resolution grayscale image H

[0080] Perform linear interpolation H = L(I) on the low-resolution original grayscale image I and obtain its frequency domain expression ξ{H}, where L{·} is the linear interpolation operation, ξ{·} is the Fourier transform operation, and F H =ξ{H}, i.e., the spectrum of the initial solution of the high-resolution image, with initial t = 1;

[0081] S2-2 Calculate the amplitude distribution of the skin corresponding to the incident light at a certain angle

[0082] For a certain angle of incident light, the resolution of the original grayscale image corresponding to this angle is O(t), and the corresponding spectral domain aperture function is P(t), then the corresponding amplitude distribution image of the skin is:

[0083]

[0084] where ξ -1 {·} is the inverse Fourier transform operation, and P(t) is the spectral domain aperture function. It can be understood that the spectral domain aperture function P(t) is determined by the position of the incident light relative to the center position.

[0085] Furthermore, it should be noted that P(t) is a matrix, and the pixel size is the same as FH (otherwise, the two cannot be multiplied pixel by pixel, and in the above formula, it is the corresponding pixel multiplication). The matrix P(t) is an aperture function, which is actually the effect of the light passing hole on the incident light at different angles, equivalent to the operation in the frequency spectrum (i.e., the operation of F ×P(t) in the formula H ). That is, at the position of the frequency spectrum, it is all 1, and elsewhere it is 0.

[0086] The spectral domain aperture function P(t) is determined by the position of the incident light relative to the center position.

[0087] The position of the frequency spectrum can be changed, which is related to the incident light angle, that is, the position of the incident light relative to the center position.

[0088] S2-3 Update the frequency spectrum of the high-resolution image

[0089] For this angle in step S2-2, use the corresponding original grayscale image O(t) to update the grayscale intensity part of its amplitude image

[0090]

[0091] where |·| is the modulus operation

[0092] Calculate the solution of the updated high-resolution image, and its frequency spectrum F' H :

[0093]

[0094] In the formula, λ is the convergence step size, which is generally a constant;

[0095] S2-4 Iterative update

[0096] (1) If t < N, let t = t + 1, and jump to step S2-2 to continue executing steps S2-2 to S2-4;

[0097] (2) If t = N, then judge whether it converges, that is, evaluate the closeness of |F' H -F H |, and judge ∑|F' H -F H|<0.1 is true, if true, it is very close, it is considered to have converged, and the high spatial resolution skin amplitude distribution image H' is calculated, that is, H'=ξ -1 {F' H};∑|F' H -F H |<0.1 is not true, then continue to jump to step S2-2 and continue to execute steps S2-2 to S2-4;

[0098] S2-5. The final updated high-resolution grayscale image H is obtained through the modulo operation. It can be understood that a high spatial resolution skin amplitude distribution image H' is obtained, half of which is a complex number including intensity and phase. Then, the intensity map, i.e. the final updated high-resolution grayscale image H, can be obtained through the modulo operation.

[0099] The above technical solution and the designed reconstruction method take into account both the skin field of view and high resolution.

[0100] Specifically, the method for calculating the texture width in step S3 is as follows:

[0101] S3-1. Divide the high-resolution grayscale image H into N H Sub-areas {H i ,i=1,2,…N H}, for different regional subgraphs H i , calculate their texture width respectively;

[0102] S3-2. Design gradient operator:

[0103] Horizontal (image width direction) operator

[0104] Vertical (image height direction) operator

[0105] S3-3. Use G x Convolution H i , combined with the threshold T H , obtain the texture image E;

[0106] S3-4, for any texture point E in E k , find the extreme point in the same row nearby in H corresponding to the position;

[0107] S3-5, according to the pixel distance between the extreme points in step S3-4, it is defined as the texture point E k Width(k);

[0108] S3-6, continue with steps S3-4 and S3-5, find the widths corresponding to all texture points, and obtain Width(k), k = 1, 2...Mx , M x is the number of texture points;

[0109] S3-7, the average texture width in the horizontal direction is:

[0110] S3-8, Similarly, replace G in S3-3 x Replace with G y , the same row in S3-4 is replaced by the same column, and M in S3-6 x Replace with M y , along steps S3-3 to S3-6, the average texture width in the vertical direction is obtained

[0111] S3-9. The average width of the texture that measures the skin texture structure is defined as:

[0112] Width=(Width x +Width y ) / 2

[0113] By analyzing N H sub-regions, for a high-resolution grayscale image H, we can obtain N H The average texture width i ,i=1,2,…N H}.

[0114] Furthermore, in step S4, the corresponding relationship model is:

[0115] Ra=a 1 (Width) 2 +a 2 (Width)+a 3

[0116] where a 1 ,a 2 ,a 3 is an undetermined parameter. According to the data pair of {roughness, texture width}, the parameter a is obtained by least squares training. 1 ,a 2 ,a 3 .

[0117] The present invention also provides a skin roughness estimation device using imaging method, such as Figure 2 and Figure 3 As shown, it includes an acquisition device and a computer controller, and the computer controller 4 includes a memory, a processor, and a computer program stored in the memory and capable of executing the above-mentioned imaging method for estimating skin roughness on the processor.

[0118] According to a further configuration of the present invention, the acquisition device includes an objective lens disk 2 and a grayscale camera 3, a plurality of LED particles 201 are provided on one side of the objective lens disk 2, the processor outputs a control signal to control the opening and closing of the LED particles 201, a light hole 203 is provided at the center of the objective lens disk 2, an imaging objective lens 202 is provided at the light hole 203, the imaging objective lens 202 and the grayscale camera 3 are on the same imaging optical axis, and the image acquired by the grayscale camera 3 is transmitted to the processor.

[0119] Among them, the imaging objective lens generally adopts an imaging microscope objective lens.

[0120] Furthermore, the plurality of LED particles 201 are evenly distributed around the light-through hole 203 and are arranged in a divergent shape.

[0121] Furthermore, a plurality of LED light rings are arranged outwardly along the edge of the light-through hole 203 on one side of the objective lens disk 2 , and a plurality of LED particles 201 are equally divided on each of the LED light rings.

[0122] Specifically, five LED light rings are arranged on the objective lens disk 2 , and the LED particles on the five LED light rings are numbered 1, 2, 3, 4, and 5 from the inside to the outside, and there are 20 LED particles in total.

[0123] It can be understood that the processor controls the opening and closing of the LED particles 201 on the LED light ring from the inside to the outside in sequence, thereby achieving different incident angles such as oblique and straight.

[0124] It should be noted that the processor's control over the on and off of LED particles is a conventional technical means and will not be specifically explained or illustrated in the embodiments.

[0125] When the device is working, the processor executes the computer program as follows:

[0126] S1. Get a set of low-resolution original grayscale images of the skin

[0127] The tester places the skin 1 to be tested at an appropriate distance in front of the objective lens disk. The LED particles on the objective lens disk emit illumination light, which illuminates the skin and then reflects. The light is collected by the imaging objective lens in the center of the objective lens disk and transmitted into the grayscale camera to obtain a low-resolution original grayscale image of the skin.

[0128] By receiving the incident light and the reflected light from the surface of the skin to be tested, a low-resolution original grayscale image I is obtained, and by controlling the LED particles to light up clockwise, a set of 20 low-resolution original grayscale images {O(t), t=1,2,3…20} is obtained;

[0129] Among them, the LED particle numbered 1 is closest to the light-through hole and can achieve vertical incidence.

[0130] It is understandable that the processor can light up / extinguish any combination of LED particles through a program. For different LED particles in the objective lens disk, the angle of irradiation of the skin position is different. According to the principles of physical optics and Fourier optics, the physical frequency of the skin acquired by imaging is related to the angle of the incident light of the illumination here. Within the imaging field angle, the larger the oblique incident angle, the higher the spatial frequency acquired. The present invention can acquire low-resolution original grayscale images corresponding to different frequencies by irradiating light at different angles. Then, a reconstruction algorithm is designed to achieve the acquisition of high-resolution images.

[0131] S2, reconstructing a high-resolution grayscale image H using a set of low-resolution original grayscale images {O(t), t = 1, 2, 3 ... 20}

[0132] S2-1. Spectrum of the initial solution of high-resolution grayscale image H

[0133] Perform linear interpolation H = L(I) on the low-resolution original grayscale image I and obtain its frequency domain expression ξ{H}, where L{·} is the linear interpolation operation, ξ{·} is the Fourier transform operation, and F H =ξ{H}, i.e., the spectrum of the initial solution of the high-resolution image, with initial t = 1;

[0134] S2-2 Calculate the amplitude distribution of the skin corresponding to the incident light at a certain angle

[0135] For a certain angle of incident light, the resolution of the original grayscale image corresponding to this angle is O(t), and the corresponding spectral domain aperture function is P(t), then the corresponding amplitude distribution image of the skin is:

[0136]

[0137] Among them, ξ -1 {·} is the inverse Fourier transform operation, and P(t) is the spectral domain aperture function. It can be understood that the spectral domain aperture function P(t) is determined by the position of the incident light relative to the center.

[0138] S2-3 Update the spectrum of high-resolution image

[0139] For this angle in step S2-2, use the corresponding original grayscale image O(t) to update the grayscale intensity part of its amplitude image

[0140]

[0141] Among them, |·| is the modulus operation,

[0142] Calculate the solution of the updated high-resolution graph, whose spectrum F' H :

[0143]

[0144] In the formula, λ is the convergence step size, which is generally a constant;

[0145] S2 - 4 Iterative update

[0146] (1) If t < N, let t = t + 1, and jump to step S2 - 2 to continue executing steps S2 - 2 to S2 - 4;

[0147] (2) If t = N, then judge convergence, that is, evaluate the proximity of |F' H - F H |, judge whether ∑|F' H - F H | < 0.1 holds. If it holds, it is already very close and considered to have converged. Calculate the high - spatial - resolution skin amplitude distribution image H', that is, H' = ξ -1 {F' H}; If ∑|F' H - F H | < 0.1 does not hold, then continue to jump to step S2 - 2 to continue executing steps S2 - 2 to S2 - 4;

[0148] S2 - 5. Obtain the finally updated high - resolution grayscale image H through modulus operation. It can be understood that to obtain the high - spatial - resolution skin amplitude distribution image H', half of it is complex, including intensity and phase. Then, through modulus operation, the intensity image, that is, the finally updated high - resolution grayscale image H, can be obtained.

[0149] S3. Calculate the texture width based on the high - resolution grayscale image

[0150] S3 - 1. Divide the high - resolution grayscale image H into N H sub - regions {H i , i = 1, 2, … N H}. For different sub - region images H i , calculate their texture widths respectively;

[0151] S3 - 2. Design the gradient operator:

[0152] Horizontal (image width direction) operator

[0153] Vertical (image height direction) operator

[0154] S3 - 3. Use G x to convolve H i , combined with the threshold T H , to obtain the texture image E;

[0155] S3-4, for any texture point E in E k , find the extreme point in the same row nearby in H corresponding to the position;

[0156] S3-5, according to the pixel distance between the extreme points in step S3-4, it is defined as the texture point E k Width(k);

[0157] S3-6, continue with steps S3-4 and S3-5, find the widths corresponding to all texture points, and obtain Width(k), k = 1, 2...M x , M x is the number of texture points;

[0158] S3-7, the average texture width in the horizontal direction is:

[0159] S3-8, Similarly, replace G in S3-3 x Replace with G y , the same row in S3-4 is replaced by the same column, and M in S3-6 x Replace with M y , along steps S3-3 to S3-6, the average texture width in the vertical direction is obtained

[0160] S3-9. The average width of the texture that measures the skin texture structure is defined as:

[0161] Width=(Width x +Width y ) / 2

[0162] By analyzing N H sub-regions, for a high-resolution grayscale image H, we can obtain N H The average width of the texture i ,i=1,2,…N H}, at this time N H =50.

[0163] S4. Establishing a corresponding relationship model based on texture width data and skin roughness

[0164] Data training was performed on the facial skin of at least 50 testers to find the best 1 ,a 2 ,a 3 .

[0165] For any tester, the final N facial skin H =50 texture average width values ​​{Width i ,i=1,2,…N HAt the same time, the roughness standard value {Ra i ,i=1,2,…N H}, forming N H For {roughness Ra, texture width Width} data.

[0166] For all testers, the operations in the previous paragraph were performed to obtain a large number of {roughness Ra, texture width Width} data pairs.

[0167] These data are substituted into the corresponding formula and equation of skin roughness Ra and texture width Width:

[0168] Ra=a 1 (Width) 2 +a 2 (Width)+a 3 , where a 1 ,a 2 ,a 3 is the unknown parameter. Use the least squares method to obtain the best a 1 ,a 2 ,a 3 Parameter value.

[0169] S5. Automatically estimate the skin roughness distribution of any user in a large field of view

[0170] For any user, repeat steps S1-S4 to obtain a high-resolution image of the user's face and N H =50 average texture width values, and then the facial skin roughness distribution can be estimated, that is, for the facial skin N H = Estimates of the roughness of 50 regions.

[0171] At this point, the imaging method for estimating skin roughness is realized.

[0172] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions and variations of these embodiments including components are made without departing from the principles and spirit of the present invention, and still fall within the scope of protection of the present invention.

Claims

1. An imaging method for skin roughness estimation, characterized in that, it includes the following steps: S1. Obtain a set of low-resolution original grayscale images of the skin By receiving the reflected light of the incident light on the surface of the skin to be measured, obtain a low-resolution original grayscale image I, and obtain a series of sets of low-resolution original grayscale images {O(t), t = 1, 2, 3... N} by controlling the angle of the incident light; S2. Reconstruct a high-resolution grayscale image H using the set of low-resolution original grayscale images {O(t), t = 1, 2, 3... N}; S3. Calculate the texture width based on the high-resolution grayscale image; The calculation method of the texture width is as follows: S3-1. Divide the high-resolution grayscale image H into N H Sub-areas {H i ,i=1,2,…N H }, for different regional subgraphs H i , calculate their texture width respectively; S3-2. Design a gradient operator: Horizontal Operator Vertical Operator S3-3. Use G x Convolution H i , combined with the threshold T H , obtain the texture image E; S3-4, for any texture point E in E k , find the extreme point in the same row nearby in H corresponding to the position; S3-5, according to the pixel distance between the extreme points in step S3-4, it is defined as the texture point E k Width(k); S3-6, continue with steps S3-4 and S3-5, find the widths corresponding to all texture points, and obtain Width(k), k = 1, 2...M x , M x is the number of texture points; S3-7, the average texture width in the horizontal direction is: S3-8, Similarly, replace G in S3-3 x Replace with G y , the same row in S3-4 is replaced by the same column, and M in S3-6 x Replace with M y , along steps S3-3 to S3-6, the average texture width in the vertical direction is obtained S3-9. The average texture width measuring the skin texture structure is defined as: Width=(Width x +Width y ) / 2 By analyzing N H sub-regions, for a high-resolution grayscale image H, we can obtain N H The average width of the texture i ,i=1,2,…N H }; S4. Establish a corresponding relationship model between the texture width data and the skin roughness Using a roughness measuring instrument, on the one hand, measure the roughness of a certain area of the skin, and on the other hand, calculate the texture width data obtained for the corresponding area of the skin image to be measured, form at least 100 pairs of {roughness, texture width} data, and then construct a corresponding relationship model between the skin roughness and the texture width; S5. Automatically estimate the skin roughness distribution with a large field of view for any user For any user, repeat steps S1-S4, and the skin roughness distribution to be measured can be obtained, corresponding to the skin in different areas under a large field of view, so that the high-precision skin roughness distribution to be measured under a large field of view can be obtained at one time.

2. The imaging method for skin roughness estimation according to claim 1, characterized in that, the step S2 includes the following sub-steps: S2-1. The spectrum of the initial solution of the high-resolution grayscale image H Perform linear interpolation H = L(I) on the low-resolution original grayscale image I and obtain its frequency domain expression ξ{H}, where L{·} is the linear interpolation operation, ξ{·} is the Fourier transform operation, and F H =ξ{H}, i.e., the spectrum of the initial solution of the high-resolution image, with initial t = 1; S2-2. Calculate the amplitude distribution image of the skin corresponding to the incident light at a certain angle For the incident light at a certain angle, the resolution original grayscale image corresponding to this angle is O(t), and the corresponding spectral domain aperture function is P(t), then the corresponding amplitude distribution image of the skin is: Among them, ξ -1 {·} is the inverse Fourier transform operation, and P(t) is the spectral domain aperture function; S2-3. Update the spectrum of the high-resolution image For this angle in step S2-2, use the corresponding original grayscale image O(t) to update the grayscale intensity part of its amplitude image where |·| is the modulus operation, Calculate the solution for updating the high-resolution graph, whose spectrum is F′ H : in the formula, λ is the convergence step size, usually a constant; S2-4. Iterative update (1) When t < N, let t = t + 1, and jump to step S2-2 to continue executing steps S2-2 to S2-4; (2) t = N, then determine whether it converges or not, that is, evaluate |F′ H -F H |, judge ∑|F′ H -F H |<0.1 is true, if true, it is very close, it is considered to have converged, and the high spatial resolution skin amplitude distribution image H' is calculated, that is, H'=ξ -1 {F′ H };∑|F′ H -F H |<0.1 is not true, then continue to jump to step S2-2 and continue to execute steps S2-2 to S2-4; S2-5. Obtain the finally updated high-resolution grayscale image H through the modulus operation.

3. The imaging method for skin roughness estimation according to claim 2, characterized in that, in the step S4, the corresponding relationship model is: Ra=a 1 (Width) 2 +a 2 (Width)+a 3 where a 1 ,a 2 ,a 3 is an undetermined parameter. According to the data pair of {roughness, texture width}, the parameter a is obtained by least squares training. 1 ,a 2 ,a 3 .

4. The imaging method for skin roughness estimation according to claim 2, characterized in that, the spectral domain aperture function P(t) is determined by the position of the incident light relative to the center position.

5. An imaging device for skin roughness estimation, characterized in that, it includes a collection device and a computer controller, and the computer controller (4) includes a memory, a processor, and a computer program stored in the memory and executable on the processor for the imaging method for skin roughness estimation according to any one of claims 1-4.

6. The imaging device for skin roughness estimation according to claim 5, characterized in that, The acquisition device comprises an objective lens disk (2) and a grayscale camera (3); a plurality of LED particles (201) are arranged on one side of the objective lens disk (2); the processor outputs a control signal to control the opening and closing of the LED particles (201); a light through hole (203) is arranged at the center of the objective lens disk (2); an imaging objective lens (202) is arranged at the light through hole (203); the imaging objective lens (202) and the grayscale camera (3) are located on the same imaging optical axis; and an image acquired by the grayscale camera (3) is transmitted to the processor.

7. The imaging method skin roughness estimation device according to claim 6, It is characterized in that The plurality of LED particles (201) are evenly distributed around the light-through hole (203) and are arranged in a divergent shape.

8. The imaging method skin roughness estimation device according to claim 7, It is characterized in that A plurality of LED light rings are arranged outwardly along the edge of the light-through hole (203) on one side of the objective lens disk (2), and a plurality of LED particles (201) are equally spaced on each of the LED light rings.

9. The imaging method skin roughness estimation device according to claim 8, It is characterized in that The processor controls the on and off of the LED particles (201) on the LED light ring in sequence from the inside to the outside.

Images