Facial wrinkle detection method and system based on adaptive hybrid Hessian filter

Abstract

The present invention provides a facial wrinkle detection method and system based on an adaptive hybrid Hessian filter, relating to the technical field of facial wrinkle detection. The method comprises: obtaining facial feature points of a subject from a standardized facial image of the subject; adaptively adjusting the scale range of the hybrid Hessian filter based on the subject's age, filtering and processing the standardized facial image using the adjusted hybrid Hessian filter to obtain a processed filtered image; setting the grayscale values ​​of the eyebrows, eyes, nose, lips, and other areas outside the face in the processed filtered image to zero based on the subject's facial feature points, and performing pseudo-color processing to obtain a wrinkle detection color image; and superimposing the wrinkle detection color image on the standardized facial image to obtain a wrinkle detection result image of the subject. The present invention can achieve the purpose of accurately and quickly detecting wrinkles on the entire face.

Description

Technical Field

[0001] The present invention relates to the field of facial wrinkle detection, and in particular to a facial wrinkle detection method and system based on an adaptive hybrid Hessian filter. Background Art

[0002] Currently, there are many methods for facial wrinkle detection. For example, filter-based facial wrinkle detection methods convert facial images into grayscale images and filter the resulting grayscale images to obtain wrinkle images. However, this method uses a fixed filter scale, making it less adaptable to facial images of subjects of different ages. For younger subjects with fewer wrinkles, shadows on facial features are misidentified as coarse wrinkles, and the algorithm takes a long time to run. Another example is facial wrinkle detection methods based on edge detection algorithms. When detecting wrinkles, these methods first roughly detect some pixels that represent the wrinkle contour, then connect these pixels using some connection rules. Finally, they detect and connect previously unidentified boundary points, remove false pixels and boundary points, and form a coherent edge. However, the wrinkle contours detected by this method can easily be confused with the line segments of the wrinkles themselves. Another example is facial wrinkle detection methods based on deep learning algorithms. This method labels wrinkles in a portion of facial image data, constructs a deep learning model, and uses this model to learn the weights of the labeled wrinkle features. Based on these weights, wrinkles are detected in a given image. However, this method is very time-consuming and labor-intensive in wrinkle labeling, and is only applicable to wrinkle detection using a dermatoscope, and is not suitable for full-face wrinkle detection. Summary of the Invention

[0003] The present invention aims to provide a facial wrinkle detection method and system based on an adaptive hybrid Hessian filter, which can accurately and quickly detect wrinkles on the entire face.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] The present invention provides a facial wrinkle detection method based on an adaptive hybrid Hessian filter, comprising:

[0006] Performing image standardization processing on the facial image of the subject to be detected to obtain a facial standardized image;

[0007] Obtaining facial feature points of the subject on a standardized facial image;

[0008] Adaptively adjust the scale range of the hybrid Hessian filter based on the subject's age;

[0009] Based on the hybrid Hessian filter, the facial standardized image is filtered using the adjusted scale range to obtain a filtering result, and the filtering result is image processed to obtain a processed filtered image;

[0010] According to the facial feature points of the subject, the grayscale values ​​of the eyebrows, eyes, nose, lips and other areas outside the face in the processed filtered image are set to zero to obtain a wrinkle detection grayscale image of the skin area, and the wrinkle detection grayscale image of the skin area is pseudo-colored to obtain a wrinkle detection color image;

[0011] The wrinkle detection color map is superimposed on the standardized facial image to obtain the subject's wrinkle detection result map.

[0012] The present invention also provides a facial wrinkle detection system based on an adaptive hybrid Hessian filter, comprising:

[0013] A standardization processing module is used to perform image standardization processing on the facial image of the subject to be detected to obtain a facial standardized image;

[0014] A facial feature point acquisition module, used to acquire facial feature points of a subject on a standardized facial image;

[0015] A scale range adjustment module, used to adaptively adjust the scale range of the hybrid Hessian filter based on the subject's age;

[0016] A filtering processing module is used to filter the normalized facial image based on a hybrid Hessian filter using the adjusted scale range to obtain a filtering result, and perform image processing on the filtering result to obtain a processed filtered image;

[0017] a wrinkle detection color map determination module, configured to set the grayscale values ​​of the eyebrows, eyes, nose, lips, and other areas outside the face in the processed filtered image to zero based on the subject's facial feature points to obtain a wrinkle detection grayscale map of the skin area, and to perform pseudo-color processing on the wrinkle detection grayscale map of the skin area to obtain a wrinkle detection color map;

[0018] The wrinkle detection result image determination module is used to superimpose the wrinkle detection color image on the facial standardized image to obtain the wrinkle detection result image of the subject.

[0019] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0020] To address the problem of how to quickly and accurately detect wrinkles across the entire face, the present invention provides a facial wrinkle detection method and system based on an adaptive hybrid Hessian filter. This method uses an adaptive scale hybrid Hessian filter to adaptively adjust the scale range of the hybrid Hessian filter according to the subject's age, reducing the number of unnecessary convolutions of the filter to achieve the purpose of accurately and quickly detecting wrinkles across the entire face. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. 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 any creative work.

[0022] Figure 1 A schematic diagram of a flow chart of a facial wrinkle detection method based on an adaptive hybrid Hessian filter provided in an embodiment of the present invention;

[0023] Figure 2 An overall flow chart of a facial wrinkle detection method based on an adaptive hybrid Hessian filter provided in an embodiment of the present invention;

[0024] Figure 3 A facial standardized image provided by an embodiment of the present invention;

[0025] Figure 4 A grayscale image for wrinkle detection provided by an embodiment of the present invention;

[0026] Figure 5 This is a structural block diagram of a facial wrinkle detection system based on an adaptive hybrid Hessian filter provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0028] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0029] Example 1

[0030] like Figure 1 and Figure 2As shown, this embodiment provides a facial wrinkle detection method based on an adaptive hybrid Hessian filter, comprising:

[0031] Step 100: performing image normalization processing on the facial image of the subject to be detected to obtain a facial normalized image, specifically: scaling the facial image of the subject to be detected into an image with a height of H=1024 in equal proportion.

[0032] Step 200: Obtain facial feature points of the subject on the facial standardized image. Specifically, based on Dlib81 facial feature points, obtain 81 facial feature points of the subject on the facial standardized image to locate the contours of the face and eyebrows, eyes, nose, and lips.

[0033] Step 300: Adaptively adjust the scale range of the hybrid Hessian filter based on the subject's age. Specifically, define the scale range of the hybrid Hessian filter as an arithmetic array σ k , the difference is Δ, which is the scale step s.

[0034] First, according to Table 1 and the age of the subjects, the maximum value σ of the scale range of the mixed Hessian filter is determined K and the minimum value σ1 and the scaling step size.

[0035] Secondly, the scale range of the hybrid Hessian filter is adaptively calculated according to formula (1); where Δ = s, s represents the scale step; 1 ≤ k ≤ K, K is the maximum value σ of the scale range K and the minimum value σ1 and the scale step size, σ k Indicates the scale value of the mixed Hessian filter corresponding to the kth element.

[0036] σ k =σ1+(k-1)×Δ (1).

[0037] Table 1 Filter scale adaptation strategy based on subject age

[0038]

[0039]

[0040] Step 400: Based on the hybrid Hessian filter, the facial normalized image is filtered using the adjusted scale range to obtain a filtering result, and the filtering result is image processed to obtain a processed filtered image.

[0041] Let I(x, y) be a pixel point in the two-dimensional image I. xx , I xy , I yx , I yyis the second-order partial differential of point I(x, y) in four directions. According to scale space theory, the second-order partial differential of point (x, y) can be obtained by convolving the second-order partial differential of the Gaussian function with the two-dimensional image I, that is:

[0042]

[0043]

[0044]

[0045] Among them, G(x, y, σ) is a Gaussian function with scale σ, that is:

[0046]

[0047] The scale σ is determined according to the filter scale adaptation strategy based on the subject's age in Table 1. The Hessian matrix is ​​obtained as follows:

[0048]

[0049] As shown in formula (6), I xx , I xy , I yx , I yy The Hessian matrix composed of can describe the two-dimensional derivative of the pixel point (x, y) on the two-dimensional image I. The two eigenvectors of the matrix can be calculated by formula (7) and formula (8), that is:

[0050]

[0051]

[0052] Eigenvector λ A ,λ B Represents the curvature size and direction of the grayscale change of the pixel point (x, y) in the x and y directions respectively. The eigenvalue with a smaller absolute value is set to λ1, representing the direction with smaller grayscale change; the eigenvalue with a larger absolute value is set to λ2, representing the direction with larger grayscale change. Based on the eigenvectors λ1 and λ2, the measure of the hybrid Hessian filter is calculated. and S:

[0053]

[0054]

[0055] Then, the similarity function of the mixed Hessian filter to enhance wrinkle lines is:

[0056]

[0057] Among them, β1 and β2 are the threshold parameters of the hybrid Hessian filter, which are used to control and the sensitivity of S.

[0058] Finally, the scales of different wrinkles are matched using formula (12), namely:

[0059]

[0060] In this embodiment, image processing is performed on the filtering result to obtain a processed filtered image, specifically including:

[0061] The filtering result is converted into a binary filtered image, and then the connected domain area threshold is used to remove small area anomaly detection results in the binary filtered image to obtain the processed filtered image.

[0062] Step 500: According to the facial feature points of the subject, the grayscale values ​​of the eyebrows, eyes, nose, lips and other areas outside the face in the processed filtered image are set to zero to obtain a wrinkle detection grayscale image of the skin area, and the wrinkle detection grayscale image of the skin area is pseudo-colored to obtain a wrinkle detection color image. Figure 3 As shown, the wrinkle detection grayscale image is as follows Figure 4 shown.

[0063] In this embodiment, pseudo color processing is performed on the wrinkle detection grayscale image of the skin area to obtain the wrinkle detection color image, which specifically includes: performing pseudo color processing on the wrinkle detection grayscale image according to the grayscale value to obtain the wrinkle detection color image.

[0064] Step 600: superimpose the wrinkle detection color map on the standardized facial image to obtain a wrinkle detection result map of the subject.

[0065] This embodiment can accurately and quickly detect the distribution, depth, and width of wrinkles across the entire face. Distribution information is obtained by overlaying the wrinkle detection result image with the normalized original facial image; depth information is obtained from the grayscale value i of the wrinkles in the wrinkle detection result image; and width information is obtained from the width and area of ​​the wrinkles in the wrinkle detection result image.

[0066] In this embodiment, the method further includes: converting the wrinkle detection result image of the subject into a wrinkle detection result grayscale image, and inputting the wrinkle detection result grayscale image into a scoring model to determine the degree of aging of the subject's facial skin.

[0067] The scoring model is:

[0068]

[0069] Where P is the total number of pixels in the wrinkle area, and M is the product of the grayscale value and the number of pixels at that grayscale value, as shown in Equations (14) and (15). This reflects the impact of each pixel in the wrinkle area on the score, and the deeper the wrinkle (the larger the grayscale value), the greater the impact.

[0070]

[0071]

[0072] Where nim(i) is the number of pixels with grayscale value i. A larger grayscale value indicates a deeper wrinkle. P represents the wrinkle area, and M reflects the wrinkle depth.

[0073] Finally, the degree of facial skin aging is quantified with a score between 0 and 100, which can scientifically quantify the degree of facial skin aging of the subject. Figure 4 The face shown scored 75 points.

[0074] Example 2

[0075] In order to execute the method corresponding to the above embodiment 1 and achieve the corresponding functions and technical effects, a facial wrinkle detection system based on an adaptive hybrid Hessian filter is provided below.

[0076] like Figure 5 As shown, this embodiment provides a facial wrinkle detection system based on an adaptive hybrid Hessian filter, comprising:

[0077] The standardization processing module 1 is used to perform image standardization processing on the facial image of the subject to be detected to obtain a facial standardized image.

[0078] The facial feature point acquisition module 2 is used to acquire the facial feature points of the subject on the facial standardized image.

[0079] The scale range adjustment module 3 is used to adaptively adjust the scale range of the hybrid Hessian filter based on the age of the subject.

[0080] The filtering processing module 4 is used to filter the normalized facial image based on the hybrid Hessian filter using the adjusted scale range to obtain a filtering result, and perform image processing on the filtering result to obtain a processed filtered image.

[0081] The wrinkle detection color map determination module 5 is used to set the grayscale values ​​of the eyebrows, eyes, nose, lips and other areas outside the face in the processed filtered image to zero based on the facial feature points of the subject to obtain a wrinkle detection grayscale map of the skin area, and perform pseudo-color processing on the wrinkle detection grayscale map of the skin area to obtain a wrinkle detection color map.

[0082] The wrinkle detection result image determination module 6 is used to superimpose the wrinkle detection color image on the standardized facial image to obtain the wrinkle detection result image of the subject.

[0083] This invention provides a facial wrinkle detection method and system based on an adaptive hybrid Hessian filter. The scale range and scaling step size of the hybrid Hessian filter are adaptively adjusted according to the subject's age. The adjusted hybrid Hessian filter is then used to filter high-resolution facial images. This method accurately and rapidly detects the distribution, depth, and width of wrinkles across the entire face. This method effectively avoids the problems of traditional hybrid Hessian filters in wrinkle detection in facial images of subjects of different ages, which can lead to meaningless increases in algorithm time and the tendency for shadowed areas of facial features to be identified as coarse wrinkles. Finally, a mathematical model for facial wrinkle scoring applicable to the entire face is proposed using two wrinkle parameters: wrinkle area and wrinkle depth.

[0084] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0085] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A facial wrinkle detection method based on an adaptive hybrid Hessian filter, characterized in that: include: Performing image standardization processing on the facial image of the subject to be detected to obtain a facial standardized image; Obtaining facial feature points of the subject on a standardized facial image; Adaptively adjust the scale range of the hybrid Hessian filter based on the subject's age, specifically including: determining the maximum value σ of the scale range of the hybrid Hessian filter based on the subject's age K and minimum value σ1 and scale step; according to the formula σ k =σ1+(k-1)×Δ adaptively calculate the scale range of the hybrid Hessian filter; where Δ=s, s represents the scale step; 1≤k≤K, K is the maximum value of the scale range σ K and the minimum value σ1 and the scale step size, σ k Represents the scale value of the mixed Hessian filter corresponding to the kth element; Based on the hybrid Hessian filter, the facial standardized image is filtered using the adjusted scale range to obtain a filtering result, and the filtering result is image processed to obtain a processed filtered image; According to the facial feature points of the subject, the grayscale values ​​of the eyebrows, eyes, nose, lips and other areas outside the face in the processed filtered image are set to zero to obtain a wrinkle detection grayscale image of the skin area, and the wrinkle detection grayscale image of the skin area is pseudo-colored to obtain a wrinkle detection color image; The wrinkle detection color map is superimposed on the standardized facial image to obtain the subject's wrinkle detection result map.

2. The facial wrinkle detection method based on adaptive hybrid Hessian filter according to claim 1, characterized in that: Obtain the subject's facial feature points on the facial standardization image, including: Based on Dlib81 facial feature points, 81 facial feature points of the subject are obtained on the facial standardized image.

3. The facial wrinkle detection method based on adaptive hybrid Hessian filter according to claim 1, characterized in that: Determine the maximum value σ of the mixed Hessian filter scale range based on the subject's age K And the minimum value σ1 and the scale step size, specifically including: When the subject is younger than 30 years old, the maximum value σ of the mixed Hessian filter scale range is set. K is determined to be 2, the minimum value σ1 is determined to be 1, and the scale step is determined to be 0.1; When the subject is older than or equal to 30 years old and younger than 40 years old, the maximum value σ of the mixed Hessian filter scale range is set. K is determined to be 5, the minimum value σ1 is determined to be 1, and the scale step is determined to be 0.3; When the subject is older than or equal to 40 years old and younger than 50 years old, the maximum value σ of the mixed Hessian filter scale range is set. K is determined to be 4, the minimum value σ1 is determined to be 1, and the scale step is determined to be 0.5; When the subject is older than or equal to 50 years old and younger than 60 years old, the maximum value σ of the mixed Hessian filter scale range is set. K The value is determined to be 8, the minimum value σ1 is determined to be 1, and the scale step is determined to be 0.5; When the subject is older than or equal to 60 years old, the maximum value σ of the mixed Hessian filter scale range is set. K The value of is determined to be 10, the minimum value σ1 is determined to be 1, and the scale step is determined to be 0.

5.

4. The facial wrinkle detection method based on adaptive hybrid Hessian filter according to claim 1, characterized in that: Perform image processing on the filtering result to obtain a processed filtered image, specifically including: The filtering result is converted into a binary filtered image, and the connected domain area threshold is used to remove small area anomaly detection results in the binary filtered image to obtain the processed filtered image.

5. The facial wrinkle detection method based on adaptive hybrid Hessian filter according to claim 1, characterized in that: Also includes: The wrinkle detection result image of the subject is converted into a wrinkle detection result grayscale image, and the wrinkle detection result grayscale image is input into the scoring model to determine the degree of aging of the subject's facial skin.

6. The facial wrinkle detection method based on adaptive hybrid Hessian filter according to claim 5, characterized in that: The scoring model is: Among them, score is the score, P is the sum of all pixel points in the wrinkle area, num(i) is the number of pixels with gray value i; M is the sum of the gray value multiplied by the number of all pixels under this gray value.

7. A facial wrinkle detection system based on an adaptive hybrid Hessian filter, characterized in that: include: A standardization processing module is used to perform image standardization processing on the facial image of the subject to be detected to obtain a facial standardized image; A facial feature point acquisition module, used to acquire facial feature points of a subject on a standardized facial image; The scale range adjustment module is used to adaptively adjust the scale range of the hybrid Hessian filter based on the subject's age, specifically including: determining the maximum value σ of the scale range of the hybrid Hessian filter based on the subject's age K and minimum value σ1 and scale step; according to the formula σ k =σ1+(k-1)×Δ adaptively calculate the scale range of the hybrid Hessian filter; where Δ=s, s represents the scale step; 1≤k≤K, K is the maximum value of the scale range σ K and the minimum value σ1 and the scale step size, σ k Represents the scale value of the mixed Hessian filter corresponding to the kth element; A filtering processing module is used to filter the normalized facial image based on a hybrid Hessian filter using the adjusted scale range to obtain a filtering result, and perform image processing on the filtering result to obtain a processed filtered image; a wrinkle detection color map determination module, configured to set the grayscale values ​​of the eyebrows, eyes, nose, lips, and other areas outside the face in the processed filtered image to zero based on the subject's facial feature points to obtain a wrinkle detection grayscale map of the skin area, and to perform pseudo-color processing on the wrinkle detection grayscale map of the skin area to obtain a wrinkle detection color map; The wrinkle detection result image determination module is used to superimpose the wrinkle detection color image on the facial standardized image to obtain the wrinkle detection result image of the subject.

Images