A method, device, equipment and readable medium for 3D face reconstruction

Abstract

This specification provides a method, apparatus, device, and readable medium for three-dimensional reconstruction of a face. The method comprises: obtaining multiple facial images of a target subject with the same facial expression from different perspectives; for each facial image, determining the target subject's eye region in the facial image based on a face detection algorithm; detecting an eyelash region within the eye region and marking target pixels in the eyelash region; performing three-dimensional reconstruction based on the marked multiple facial images to obtain a three-dimensional facial point cloud; determining the eyelash locations of the eyelash regions within the three-dimensional facial point cloud; deleting at least the eyelash locations to obtain a target three-dimensional facial point cloud; and constructing a target three-dimensional facial model based on the target three-dimensional facial point cloud. By deleting the marked eyelash data from the three-dimensional point cloud and creating a three-dimensional facial model based on the deleted three-dimensional point cloud, the effect of the presence of eyelashes on the quality of the three-dimensional facial model can be reduced to a certain extent.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of computer technology, and in particular to a method, apparatus, device, and readable medium for three-dimensional reconstruction of a face. Background Art

[0002] 3D facial reconstruction technology is widely used in film, television, gaming, finance, live streaming, and other fields, such as 3D animation and game production, creating virtual avatars for live streamers, and creating 3D facial emoticons. While modeling based on real-life data improves modeling efficiency, the presence of eyelashes can easily create noise in high-precision models (HDMs). Occluding eyelids from eyelashes can lead to partial loss of eyelid data, reducing the quality and accuracy of the low-polygon model that fits the HDM, making it difficult to achieve the desired eyelid effect. Summary of the Invention

[0003] To overcome the problems existing in the related art, this specification provides a method, apparatus, device and readable medium for three-dimensional reconstruction of a face.

[0004] According to a first aspect of the embodiments of this specification, a method is provided, comprising:

[0005] Obtain multiple facial images of the target object with the same expression from different perspectives;

[0006] For each face image, determining the eye region of the target object in the face image based on a face detection algorithm;

[0007] Detecting an eyelash area in the eye area and marking target pixel points in the eyelash area;

[0008] Perform three-dimensional reconstruction based on multiple labeled face images to obtain a three-dimensional point cloud of the face;

[0009] Determining the eyelash location of the marked eyelash region in the three-dimensional point cloud of the face;

[0010] Deleting at least the eyelash site to obtain a three-dimensional point cloud of the target face;

[0011] A target 3D face model is constructed based on the target face 3D point cloud. According to a second aspect of the embodiments of this specification, a device for 3D face reconstruction is provided, the device comprising:

[0012] An acquisition module is used to acquire multiple facial images of the target object with the same expression from different perspectives;

[0013] A first determination module is configured to determine, for each facial image, an eye region of the target object in the facial image based on a face detection algorithm;

[0014] a detection module, configured to detect an eyelash area in the eye area;

[0015] A marking module, used to mark target pixels in the eyelash area;

[0016] A first reconstruction module is used to perform three-dimensional reconstruction based on the multiple labeled face images to obtain a three-dimensional face point cloud;

[0017] A second determination module is used to determine the eyelash position of the marked eyelash area in the three-dimensional point cloud of the face;

[0018] a deletion module, configured to delete at least the eyelash site to obtain a three-dimensional point cloud of the target face;

[0019] The second reconstruction module is used to construct a target three-dimensional face model based on the target face three-dimensional point cloud.

[0020] According to a third aspect of the embodiments of this specification, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the first aspect above when executing the program.

[0021] According to a third aspect of the embodiments of this specification, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and the computer program is used to instruct related hardware to implement the method described in the first aspect above.

[0022] The technical solutions provided by the embodiments of this specification may have the following beneficial effects:

[0023] The embodiments of this specification mark the eyelash area in multiple acquired facial images to remove the eyelash sites in the facial 3D point cloud based on the marking, and create a 3D facial model based on the facial 3D point cloud after deleting the eyelash data. This can reduce the impact of the presence of eyelashes on the quality of the facial 3D model to a certain extent and achieve the desired eyelid effect.

[0024] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the specification and, together with the description, serve to explain the principles of the specification.

[0026] Figure 1 This is a schematic diagram of a three-dimensional reconstruction based on data collected by real people, as shown in this specification.

[0027] Figure 2This is a flowchart of a method for 3D reconstruction of a face according to an exemplary embodiment of this specification.

[0028] Figure 3 This is a schematic diagram of a method for detecting and marking eyelash areas according to an exemplary embodiment of this specification.

[0029] Figure 4 This is a flowchart of another method for 3D reconstruction of a face according to an exemplary embodiment of this specification.

[0030] Figure 5 This is a hardware structure diagram of the computer device where the three-dimensional face reconstruction device in the embodiment of this specification is located.

[0031] Figure 6 This is a block diagram of a 3D face reconstruction device according to an exemplary embodiment of the present specification. DETAILED DESCRIPTION

[0032] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with this specification. Rather, they are merely examples of apparatus and methods consistent with certain aspects of this specification, as detailed in the appended claims.

[0033] The terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit this specification. As used in this specification and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0034] It should be understood that although the terms first, second, third, etc. may be used in this specification to describe various information, such information should not be limited to these terms. These terms are merely used to distinguish information of the same type from one another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information without departing from the scope of this specification. Depending on the context, the term "if" as used herein may be interpreted as "when," "when," or "in response to determining."

[0035] 3D facial reconstruction technology is widely used in film, television, gaming, finance, live streaming, and other fields, such as 3D animation production, 3D game production, avatar creation for anchors, and 3D facial emoticon packs. In traditional modeling processes, the quality of high-precision 3D facial models relies on the expertise of professional 3D modelers and is time-consuming and costly. 3D modeling methods based on real-person data can improve modeling efficiency and are broadly divided into the following steps:

[0036] First, multiple cameras are used to capture photos of the object to be modeled from different perspectives at the same time, generating multi-view photos. A 3D reconstruction algorithm is then used to convert the multi-view photos into a 3D point cloud. This point cloud is then matched with a standard facial model to produce a high-precision original 3D model (also known as a high-poly model). High-poly models typically contain millions or even tens of millions of triangular facets generated based on color and texture information, making them difficult to directly use in rendering engines. Finally, the facial model is retopologically retopologically, resulting in a model (called a low-poly model) that matches the high-poly model but has fewer triangular facets. This low-poly model can be used to generate expressions directly in the rendering engine.

[0037] In the above process, the quality of the high-poly model determines the realism and details of the low-poly model. Figure 1 Figure 1 shows a schematic diagram of a 3D reconstruction based on real-life data. The presence of eyelashes in a real-life photo 110 can easily create noise in the high-polygon model 120. The occlusion of the eyelids by the eyelashes can lead to partial loss of eyelid data, resulting in bulges in the generated low-polygon model 130. This reduces the quality and accuracy of the low-polygon model, making it difficult to achieve the desired eyelid effect.

[0038] In response to the problems existing in the related art, this specification proposes a method for three-dimensional reconstruction of a face, which marks eyelash data in a two-dimensional image, deletes the corresponding eyelash locations in the three-dimensional point cloud based on the marks, and creates a three-dimensional face model based on the deleted three-dimensional point cloud, thereby reducing the impact of the presence of eyelashes on the quality of the three-dimensional face model to obtain the desired eyelid effect.

[0039] Next, the embodiments of this specification are described in detail.

[0040] like Figure 2 As shown, Figure 2 This is a flowchart of a method for 3D reconstruction of a face according to an exemplary embodiment of the present specification, comprising the following steps:

[0041] Step S201, obtaining multiple facial images of a target subject with the same expression from different perspectives;

[0042] To convert two-dimensional images into three-dimensional data, it is necessary to capture multiple images of the target object to be modeled (target object) from different perspectives. These images must be taken simultaneously and correspond to the same expression of the target object. The image capture device can be a standard camera or a depth camera. Considering the variations in facial expressions in practical applications of three-dimensional facial models, multiple sets of facial images of the target object with different expressions can be captured. Each set of facial images includes multiple facial images from different perspectives with the same expression, and a three-dimensional facial model is created for each set of facial images. In other words, each expression corresponds to a corresponding three-dimensional facial model.

[0043] Step S202: for each face image, determining the eye region of the target object in the face image based on a face detection algorithm;

[0044] Every face image needs to be processed for eyelash area detection and labeling. First, the eye area of the target object in the face image is determined based on the face detection algorithm. Usually, face detection is based on facial key points. Facial key point detection is also called facial key point detection, positioning or face alignment. It means locating the key areas of the face, including eyebrows, eyes, nose, mouth, facial contours, etc., given a face image. Facial key points reflect the facial features of various parts of the face. With the development of technology and the increase in accuracy requirements, the number of facial key points has grown from the initial 5 to 200 today. The common annotation scheme is the 68-point annotation method. The version used by the Dlib algorithm divides facial key points into internal key points and contour key points. The internal key points include eyebrows, eyes, nose, and mouth, a total of 51 key points, and the contour key points include 17 key points.

[0045] The Active Shape Model (ASM) and Active Appearance Model (AAM) algorithms are two classic algorithms for facial landmark detection. The ASM algorithm uses manual calibration to first calibrate the training set, then obtains a shape model through training. It then matches specific objects through keypoint matching. The AAM improves on the ASM algorithm by creating not only a shape model but also a texture model during the model building phase, using this fusion of the two models for matching.

[0046] Step S203, detecting the eyelash area in the eye area and marking the target pixel points in the eyelash area;

[0047] After determining the eye region in the face image, the eyelash region is detected in the eye region. Eyelashes have a certain growth direction and are clearly distinguishable from other facial regions. By performing edge / texture analysis on the eye region, the eyelash region can be highlighted. In one embodiment of the present specification, detecting the eyelash region in the eye region includes: cropping an eye image containing the eye region of the target object; detecting the eyelash region based on the eye image output by a specified image channel, where the specified image channel is an image channel in which the contrast between the eyelash region and other regions is higher than that of other channels. Edge detection emphasizes the contrast of the image, which is the difference in brightness. Based on the contrast, the difference between the edge and non-edge regions can be more prominent. The eye image can be converted into HSV format, and the eye image of the V channel is taken for the detection of the eyelash region; or it can be converted into YUV format, and the eye image of the Y channel is taken for the detection of the eyelash region. In one embodiment of the present specification, the specified image channel is the V channel in the HSV image format.

[0048] There are many methods for edge / texture analysis of images, such as first-order edge detection operators, Sobel edge detection operators, and Canny edge detection operators. Gabor filters, which have frequency and directional representations similar to those of the human visual system and are insensitive to light changes, are a common method for edge / texture analysis. In another embodiment of the present application, the eyelash region is detected based on an eye image output from a specified image channel using multiple preset Gabor filters.

[0049] Because a Gabor filter can only detect texture in a fixed direction, and eyelashes grow in a chaotic and chaotic manner, multiple Gabor filters are needed to detect the eyelash region. The detection direction of the Gabor filters is achieved by varying the parameter θ. The process of detecting the target eye image using multiple preset Gabor filters is generally as follows: the target eye image is simultaneously input into each Gabor filter to perform convolution processing on the target eye image. The output value corresponding to each pixel in the output image of each Gabor filter is used to represent the probability that the pixel belongs to the edge feature in the direction detected by the Gabor filter. The larger the corresponding output value, the higher the probability. Therefore, each pixel has multiple output values. The largest output value is taken as the target output value and compared with a preset threshold. Pixels above the threshold are determined to be eyelash pixels. The convolution kernel size, various Gabor filter parameters, the number of Gabor filters, and the threshold of the Gabor filter output values are set based on the characteristics of the image to be processed, such as the resolution of the image to be processed. The region composed of the determined eyelash pixels is the eyelash region.

[0050] The target pixels can be pixels in a specific eyelash region or pixels in a specified area surrounding the eyelash region. This is because the eyelash pixels obtained through edge / texture analysis typically belong to eyelashes with clear edges and distinct roots, and because eyelashes have a certain width, the detected eyelash pixels are generally located at the edge of the eyelashes rather than in the center, which can lead to missed detection of eyelash pixels. In one embodiment of the present specification, the method further includes: determining a target eyelash region, where the target eyelash region includes the eyelash region and a specified area surrounding the eyelash region, and the target pixels are pixels in the target eyelash region. By also marking the pixels in the specified area surrounding the eyelash region, the coverage of eyelashes in the facial image can be improved, thereby improving the accuracy of eyelash detection in the facial image. The specified area surrounding the eyelash region can be obtained by dilating the eyelash region using a preset convolution kernel, the size of the convolution kernel being set by the user.

[0051] Marking each pixel in the target eyelash area to distinguish it from other areas of the face can be achieved by marking it with color, for example, marking each pixel in the target eyelash area with green, which is more distinguishable from human skin.

[0052] like Figure 3 The schematic diagram of a method for detecting and marking eyelash regions according to an exemplary embodiment of this specification is shown. First, after facial key point detection, a facial image 310 is cropped into an eye image 320 that only includes the eye region 321. Eye image 320 is then converted to HSV format, and a V-channel image 330 is taken as the target eye image. Next, the V-channel eye image is input into 36 preset Gabor filters. Each pixel corresponds to 36 output values, and the output values range from 0 to 255. The maximum value of the corresponding output values is taken for each pixel, generating a corresponding response image 340. The pixels corresponding to the eyelashes can be clearly seen in response image 340. Response image 340 is then expanded to cover more eyelash pixels, resulting in an expanded eyelash region image 350. Compared to response image 340, the eyelash region covers more pixels. Finally, the eyelash pixels determined in the expanded eyelash region image 350 are marked green at corresponding positions in the facial image 310 to obtain a marked facial image 360 .

[0053] Step S204: performing three-dimensional reconstruction based on the labeled multiple facial images to obtain a three-dimensional point cloud of the face;

[0054] Step S205, determining the eyelash location of the marked eyelash area in the three-dimensional point cloud of the face;

[0055] Step S206, deleting at least the target eyelash site to obtain a target face three-dimensional point cloud;

[0056] Step S207: construct a target three-dimensional face model based on the target face three-dimensional point cloud.

[0057] The three main data formats used in 3D reconstruction are voxels, point clouds, and grids. 2D images must first be converted into a 3D data format before 3D reconstruction can be performed. In one embodiment of this specification, multiple facial images marked with the same expression are converted into a 3D facial point cloud, with each point having 3D coordinates, color information, and texture information. The color information for each point in the 3D facial point cloud is obtained by fusing the color information from multiple facial images. For example, if an eyelash pixel marked as green in facial image A is not marked as skin color in facial image B, then the color of the point in the 3D facial point cloud corresponding to that pixel will be between green and skin color. On this basis, to ensure the accuracy of eyelash location determination, points with colors similar to the marked colors are also determined as eyelash locations.

[0058] The similarity between colors can be obtained by calculating the Euclidean distance between two colors in RGB space. The formula is:

[0059]

[0060] Among them, C1 and C2 represent color 1 and color 2 respectively, C 1R 、C 1G 、C 1B Represents the R channel, G channel and B channel of color 1, C 2R 、C 1G 、C 1B However, since the RGB space is linear and the human visual system is nonlinear, the RGB space cannot reflect the human eye's perception of color. The color distance calculated based on the above formula cannot reflect the human eye's perception of the difference between two colors.

[0061] In order to make the calculation result close to the human eye's perception of color, we can calculate the weighted Euclidean distance in the RGB space. The weighted Euclidean distance calculation formula is as follows:

[0062]

[0063] ΔR=C 1，R -C 2，R

[0064] ΔG=C 1，G -C 2，G

[0065] ΔB=C 1，B-C 2，B

[0066]

[0067] In one embodiment of this specification, a parameter ΔC, representing the similarity between the color of each point in the 3D facial point cloud and the marker color, is calculated based on the weighted Euclidean distance formula (the smaller the ΔC, the more similar the two colors are), and compared with a preset upper limit. Points with ΔC less than the preset upper limit are identified as eyelash sites. Removing eyelash sites from the 3D facial point cloud can reduce the impact of eyelashes on the quality of the 3D facial model. Other methods that can represent the similarity between two colors and that approximate the human eye's color perception are also applicable and are not limited by this specification.

[0068] The presence of eyelashes can generate noise around them, which can also affect the quality and accuracy of modeling. Eliminating noise interference can further improve the quality of the 3D facial model. In one embodiment of this specification, it is also necessary to determine the noise points within a preset range of the eyelash loci, and then delete the eyelash loci and the noise points to obtain the target face 3D point cloud. The noise points within the preset range of the eyelash loci are all points contained in a sphere with the eyelash loci as the center and the preset range as the radius. To improve the efficiency of searching and determining noise points, a KDTree can be constructed to search for noise points. In one embodiment of this specification, noise points are determined based on the KDTree. Because areas closer to the eyelash tips generate more noise points, different noise search ranges can be set for eyelash loci located in different parts of the eyelashes to minimize the impact of noise. In one embodiment of this specification, the preset range is set based on the part of the eyelashes where the eyelash loci are located. In another embodiment of this specification, the preset range increases linearly from the eyelash root to the eyelash tip.

[0069] A target 3D face model is constructed based on a 3D face point cloud with eyelashes removed or with noise around the eyelashes removed. Furthermore, the target 3D face point cloud can be converted into voxels or a 3D mesh, and a 3D face model can be created based on the voxels or 3D mesh.

[0070] like Figure 4 FIG2 is a flowchart of a method for 3D reconstruction of a face according to another exemplary embodiment of the present specification, comprising the following steps:

[0071] Step S401, obtaining multiple facial images of a target subject with the same expression from different perspectives;

[0072] Step S402: for each face image, determining the eye region of the target object in the face image based on a face detection algorithm;

[0073] Step S403 , cropping the eye image containing the eye area of the target object and converting it into a V channel image in the HSV format;

[0074] Step S404: Detecting the eyelash region in the V channel image using a plurality of preset Gabor filters and performing dilation processing on the eyelash region;

[0075] Step S405 , marking the pixels of the eyelash area after the expansion process as green and performing three-dimensional reconstruction based on the marked multiple facial images to obtain a three-dimensional point cloud of the face;

[0076] Step S406, determining the eyelash location of the marked eyelash area in the three-dimensional point cloud of the face and the noise points within a preset range of the eyelash location;

[0077] Step S407, deleting eyelash sites and noise points to obtain a three-dimensional point cloud of the target face;

[0078] Step S408: construct a target 3D face model based on the target face 3D point cloud.

[0079] Corresponding to the embodiments of the aforementioned method, this specification also provides an embodiment of a three-dimensional face reconstruction device and a terminal used therein.

[0080] The embodiments of the face 3D reconstruction device of this specification can be applied to electronic devices. The device embodiments can be implemented through software, hardware, or a combination of software and hardware. Taking software implementation as an example, as a device in a logical sense, it is formed by the processor reading the corresponding computer program instructions in the non-volatile memory into the memory and running it. From the hardware level, if Figure 5 The figure shows a hardware structure diagram of an electronic device where the 3D face reconstruction device of the embodiment of this specification is located. Figure 5 In addition to the processor 510, memory 530, network interface 520, and non-volatile memory 540 shown, the electronic device where the device 531 is located in the embodiment may also include other hardware according to the actual function of the electronic device, which will not be described in detail.

[0081] like Figure 6 As shown, Figure 6 This is a block diagram of a 3D face reconstruction device according to an exemplary embodiment of the present specification, the device comprising:

[0082] An acquisition module 610 is used to acquire multiple facial images of a target subject with the same expression from different perspectives;

[0083] A first determination module 620 is configured to determine, for each facial image, the eye region of the target object in the facial image based on a face detection algorithm;

[0084] A detection module 630 is used to detect an eyelash area in the eye area;

[0085] A marking module 640 is used to mark target pixels in the eyelash area;

[0086] A first reconstruction module 650 is configured to perform 3D reconstruction based on the labeled facial images to obtain a 3D facial point cloud;

[0087] The second determining module 660 is used to determine the eyelash location of the marked eyelash area in the three-dimensional point cloud of the face;

[0088] A deletion module 670 is configured to delete at least the eyelash site to obtain a three-dimensional point cloud of the target face;

[0089] The second reconstruction module 680 is used to construct a target 3D face model based on the target 3D point cloud.

[0090] The implementation process of the functions and effects of each module in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0091] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this specification. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0092] Accordingly, this specification also provides a computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the program, the method described in any of the aforementioned method embodiments is implemented. Furthermore, this specification also provides a computer-readable storage medium storing a computer program configured to instruct related hardware to implement the method described in any of the aforementioned method embodiments.

[0093] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0094] Those skilled in the art will readily recognize other embodiments of the present invention upon consideration of the present invention and practice of the invention claimed herein. This specification is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of this specification and include common knowledge or customary techniques in the art not claimed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present invention being indicated by the claims.

[0095] It should be understood that the present description is not limited to the exact structure that has been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present description is limited only by the appended claims.

[0096] The above description is only a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this specification should be included in the scope of protection of this specification.

Claims

1. A method for 3D reconstruction of a human face, characterized in that: The method comprises: Obtain multiple facial images of the target object with the same expression from different perspectives; For each face image, determining the eye region of the target object in the face image based on a face detection algorithm; Detecting an eyelash area in the eye area and marking pixel points in the eyelash area, where the pixel points in the eyelash area belong to target pixel points; Perform three-dimensional reconstruction based on multiple labeled face images to obtain a three-dimensional point cloud of the face; Determining the eyelash location of the marked eyelash region in the three-dimensional point cloud of the face; Deleting at least the eyelash site to obtain a three-dimensional point cloud of the target face; A target three-dimensional face model is constructed based on the target face three-dimensional point cloud.

2. The method according to claim 1, characterized in that The method further comprises: Determining a noise point located within a preset range of the eyelash position; Deleting at least the eyelash site to obtain a three-dimensional point cloud of the target face includes: The eyelash sites and the noise points are deleted to obtain a three-dimensional point cloud of the target face.

3. The method according to claim 1, characterized in that The method further comprises: A target eyelash area is determined, where the target eyelash area includes the eyelash area and a designated area around the eyelash area, and the target pixel points include pixel points of the eyelash area included in the target eyelash area and pixel points of the designated area around the eyelash area.

4. The method according to claim 1, wherein Detecting the eyelash area in the eye area includes: cropping an eye image containing an eye region of the target object; The eyelash region is detected based on an eye image output by a designated image channel, where the designated image channel is an image channel in which the contrast between the eyelash region and other regions is higher than that of other channels.

5. The method according to claim 4, characterized in that The designated image channel is the V channel in the HSV image format.

6. The method according to claim 4, characterized in that The eyelash area is detected based on the eye image output by the designated image channel based on a plurality of preset Gabor filters.

7. The method according to claim 2, characterized in that The noise points are determined based on KDTree.

8. The method according to claim 7, characterized in that The preset range is set according to the eyelash part where the eyelash site is located.

9. The method according to claim 8, characterized in that The preset range increases linearly from the eyelash root to the eyelash tip.

10. A device for three-dimensional reconstruction of a human face, characterized in that: The device comprises: An acquisition module is used to acquire multiple facial images of the target object with the same expression from different perspectives; A first determination module is configured to determine, for each facial image, an eye region of the target object in the facial image based on a face detection algorithm; a detection module, configured to detect an eyelash area in the eye area; a marking module, configured to mark pixel points in the eyelash region, where the pixel points in the eyelash region belong to target pixel points; A first reconstruction module is used to perform three-dimensional reconstruction based on the multiple labeled face images to obtain a three-dimensional face point cloud; A second determination module is used to determine the eyelash position of the marked eyelash area in the three-dimensional point cloud of the face; a deletion module, configured to delete at least the eyelash site to obtain a three-dimensional point cloud of the target face; The second reconstruction module is used to construct a target three-dimensional face model based on the target face three-dimensional point cloud.

11. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 9 is implemented.

12. A computer-readable storage medium storing a computer program, wherein the computer program is used to instruct related hardware to perform the method according to any one of claims 1 to 9.

Images