A mesh denoising method and device based on adaptive feature preservation

Abstract

The application provides a mesh denoising method and device based on adaptive feature preservation, which determines feature regions and non-feature regions in a triangular mesh model based on a normal difference method; determines a first sub-neighborhood set corresponding to each feature vertex in the feature regions, and obtains a feature normal vector of each feature vertex after filtering processing; and updates the vertex coordinates in the feature regions based on the feature normal vector. The application can minimize the difference between the filtered result and the original normal vector while smoothing and correcting the normal vector of the mesh model, thereby effectively reducing the loss of feature information and improving the precision and visual effect of the mesh model.

Description

Technical Field

[0001] This invention relates to the field of model processing technology, and specifically to a mesh denoising method and apparatus based on adaptive feature preservation. Background Technology

[0002] Polygonal meshes, comprising three-dimensional points and their connectivity information, are widely used in the geometric representation of three-dimensional shapes. Mesh data is generated by converting the object to be scanned into discrete point cloud data using various 3D scanning devices. However, the network model generated during the point cloud and data acquisition process is susceptible to noise interference from various sources, such as vibrations from the 3D scanner or computational errors caused by reconstruction detection techniques. Furthermore, noisy data often interferes with the components of the mesh model. Therefore, mesh denoising is an extremely important task in order to effectively utilize curved surface meshes in image geometry processing.

[0003] Most existing mesh denoising methods fail to preserve the feature information of the object being scanned. During the denoising process of the mesh model, the feature edges of the object being scanned are also smoothed out, resulting in distortion of the mesh model and high measurement errors of the scanned object.

[0004] Therefore, there is an urgent need for a mesh denoising method and apparatus based on adaptive feature preservation to solve the problems of mesh model distortion and high measurement error of scanned objects in the existing mesh denoising process. Summary of the Invention

[0005] In view of this, it is necessary to provide a mesh denoising method and apparatus based on adaptive feature preservation to solve the problems of mesh model distortion and high measurement error of scanned objects in the existing mesh denoising process.

[0006] To address the aforementioned technical problems, this invention provides a mesh denoising method based on adaptive feature preservation, comprising:

[0007] The characteristic and non-characteristic regions in the triangular mesh model are determined based on the normal difference method.

[0008] An adaptive feature-preserving neighborhood selection method is used to determine the first sub-neighborhood set corresponding to each feature vertex in the feature region, and the first sub-neighborhood in the first sub-neighborhood set is filtered to obtain the feature normal vector of each feature vertex after filtering.

[0009] Update the vertex coordinates within the feature region based on the feature normal vector;

[0010] Wherein, the first sub-neighborhood is the first face connected to the feature vertex, and the first sub-neighborhood of the edge in the feature region is the second face connected to the edge and the third face sharing the edge with the second face.

[0011] In one possible implementation, determining the feature regions and non-feature regions in the triangular mesh model based on the normal difference method includes:

[0012] Construct the topological relationships of the triangular mesh model data to obtain the connection relationship diagram between the facets and vertices;

[0013] Determine the first neighborhood of the face and the vertex, and calculate the first normal vector of the triangular mesh model based on the first neighborhood according to the position information of the adjacent face and vertex;

[0014] Determine the second neighborhood of the face and the vertex, and calculate the second normal vector of the triangular mesh model based on the second neighborhood according to the position of the face and the vertex in the second neighborhood;

[0015] The first normal vector and the second normal vector are filtered to obtain the first target normal vector and the second target normal vector after filtering.

[0016] If the absolute value of the difference between the first target normal vector and the second target normal vector is less than a preset difference threshold, then the face and the vertex are determined to be located in a non-feature region. If the absolute value of the difference between the first target normal vector and the second target normal vector is greater than the preset difference threshold, then the face and the vertex are determined to be located in a feature region.

[0017] The range of the second neighborhood is larger than the range of the first neighborhood.

[0018] In one possible implementation, the filtering method is a bilateral normal filtering method.

[0019] In one possible implementation, the bilateral normal filtering formula is:

[0020]

[0021] Here, nor() represents the normalization operation. c represents the normal vector after filtering in the first sub-neighborhood. j n represents the center point of the neighborhood patch. j W represents the normal vector of the neighborhood patch. s W represents the spatial domain weight. r This represents the range weight.

[0022] In one possible implementation, the W s and W r The calculation formulas are as follows:

[0023]

[0024]

[0025] Wherein, the standard deviation σ s σ represents the spatial similarity of neighboring patches. c This represents the normal similarity of the area patches.

[0026] In one possible implementation, the filtering process on the sub-neighborhoods in the sub-neighborhood set includes:

[0027] The filtered result with the smallest difference from the unfiltered normal vector is selected as the normal vector of the sub-neighborhood, and the selection formula is as follows:

[0028]

[0029] Where, n i This represents the normal vector before sub-neighborhood filtering.

[0030] In one possible implementation, the formula for calculating the vertex coordinates in updating the vertex coordinates within the feature region based on the normal vector is:

[0031]

[0032] in, These are the updated vertex coordinates.

[0033] One possible implementation also includes:

[0034] Obtain the second sub-neighborhood set corresponding to each non-feature patch in the non-feature region, and perform filtering processing on the second sub-neighborhood in the second sub-neighborhood set to obtain the non-feature normal vector of each feature vertex after filtering processing;

[0035] The second sub-neighborhood is the fourth facet connected to the non-feature facet.

[0036] In one possible implementation, the method further includes updating the coordinates of non-feature vertices within the non-feature region based on the non-feature normal vector, wherein the formula for calculating the coordinates of the non-feature vertices is:

[0037]

[0038] in, For the updated non-feature vertex coordinates, v i The coordinates of the non-feature vertices before the update are given, α is the update step size, and W is the value of the vertex. ij W is the weight function for non-feature normal vector similarity and neighborhood points. ij The calculation formula is:

[0039]

[0040] Where δ is the radius of the triangular mesh model.

[0041] To address the aforementioned problems, the present invention also provides a mesh denoising device based on adaptive feature preservation, comprising:

[0042] The feature calculation module is used to determine the feature regions and non-feature regions in the triangular mesh model based on the normal difference method;

[0043] The normal filtering module is used to determine the first sub-neighborhood set corresponding to each feature vertex in the feature region using an adaptive feature-preserving neighborhood selection method, and to filter the first sub-neighborhood in the first sub-neighborhood set to obtain the feature normal vector of each feature vertex after filtering.

[0044] A vertex update module is used to update the vertex coordinates within the feature region based on the feature normal vector;

[0045] Wherein, the first sub-neighborhood is the first face connected to the feature vertex, and the first sub-neighborhood of the edge in the feature region is the second face connected to the edge and the third face sharing the edge with the second face.

[0046] The beneficial effects of the above embodiments are as follows: Based on the normal difference method, the feature regions and non-feature regions in the triangular mesh model are determined, which can quickly classify the mesh features without relying on complex feature direction estimation. Then, the adaptive feature-preserving neighborhood selection method is used to determine the first sub-neighborhood in the feature region and perform filtering processing to obtain the feature normal vectors of each feature vertex after filtering. Based on the feature normal vectors, the vertex coordinates in the feature region are updated. This can achieve smooth correction of the mesh model method vector while minimizing the difference between the filtered result and the original normal vector, effectively reducing the loss of feature information and improving the accuracy and visual effect of the mesh model. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 A flowchart illustrating an embodiment of the adaptive feature-preserving mesh denoising method provided by the present invention;

[0049] Figure 2 For the present invention Figure 1A flowchart illustrating one embodiment of S101;

[0050] Figure 3 This is a schematic diagram of the scene selected for the first sub-domain in this invention;

[0051] Figure 4 This is a schematic diagram illustrating the effect of an embodiment of the adaptive feature-preserving mesh denoising method in this invention;

[0052] Figure 5 A schematic diagram of an embodiment of the mesh denoising device based on adaptive feature preservation provided by the present invention;

[0053] Figure 6 This is a schematic diagram of another embodiment of the mesh denoising device based on adaptive feature preservation provided by the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0055] Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor systems and / or microcontroller systems.

[0056] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0057] This invention provides a mesh denoising method and apparatus based on adaptive feature preservation, which will be described below.

[0058] Figure 1 This is a flowchart illustrating an embodiment of the adaptive feature-preserving mesh denoising method provided by the present invention.

[0059] Reference Figure 1 This invention provides a mesh denoising method based on adaptive feature preservation, comprising:

[0060] S101. Determine the characteristic and non-characteristic regions in the triangular mesh model based on the normal difference method;

[0061] S102. An adaptive feature-preserving neighborhood selection method is used to determine the first sub-neighborhood set corresponding to each feature vertex in the feature region, and the first sub-neighborhood in the first sub-neighborhood set is filtered to obtain the feature normal vector of each feature vertex after filtering.

[0062] S103. Update the vertex coordinates within the feature region based on the feature normal vector;

[0063] Wherein, the first sub-neighborhood is a set of first faces connected to the feature vertex, and the first sub-neighborhood of the edge in the feature region includes a second face connected to the edge and a third face sharing an edge with the second face.

[0064] It is understandable that the face connected to the feature vertex is the first face of that feature vertex, the face connected to the edge in the feature region (i.e., two feature vertices connected form an edge) is the second face of that edge, and the face that shares an edge with the second face is the third face.

[0065] The above scheme determines the feature and non-feature regions in the triangular mesh model based on the normal difference method, which can quickly classify mesh features without relying on complex feature direction estimation. Then, an adaptive feature-preserving neighborhood selection method is used to determine the first sub-neighborhood in the feature region and perform filtering to obtain the feature normal vectors of each feature vertex after filtering. Based on the feature normal vectors, the vertex coordinates in the feature region are updated. This can achieve smooth correction of the mesh model's method vector while minimizing the difference between the filtered result and the original normal vector, effectively reducing the loss of feature information and improving the accuracy and visual effect of the mesh model.

[0066] Figure 2 For the present invention Figure 1 A flowchart of one embodiment of S101.

[0067] Reference Figure 2 The determination of feature regions and non-feature regions in the triangular mesh model based on the normal difference method includes:

[0068] S201. Construct the topological relationships of the triangular mesh model data to obtain the connection relationship diagram between the face and the vertex;

[0069] S202. Determine the first neighborhood of the face and the vertex, and calculate the first normal vector of the triangular mesh model based on the first neighborhood according to the position information of the adjacent face and vertex;

[0070] S203. Determine the second neighborhood of the face and the vertex, and calculate the second normal vector of the triangular mesh model based on the second neighborhood according to the position of the face and the vertex in the second neighborhood;

[0071] S204. Filter the first normal vector and the second normal vector to obtain the filtered first target normal vector and the filtered second target normal vector.

[0072] S205. When the absolute value of the difference between the first target normal vector and the second target normal vector is less than a preset difference threshold, it is determined that the face and the vertex are located in a non-feature region. When the absolute value of the difference between the first target normal vector and the second target normal vector is greater than the preset difference threshold, it is determined that the face and the vertex are located in a feature region.

[0073] The range of the second neighborhood is larger than the range of the first neighborhood.

[0074] Understandably, in the non-feature regions of a triangular mesh model, such as planes, faces, and vertices, the normal vectors will exhibit similar data. However, in the feature regions of the triangular mesh model, such as edges, chamfers, and corners, the normal vectors of faces and vertices will show significantly different data. Based on this characteristic, we first obtain a small neighborhood (i.e., the first neighborhood) for each face and vertex in the triangular mesh model. Then, based on the positional information of adjacent faces and vertices, we calculate the normal vector n of the triangular mesh model based on the first neighborhood. s Then, the large-scale neighborhood (i.e., the second neighborhood) of each face and vertex in the triangular mesh model is obtained. Similarly, the normal vector n of the triangular mesh model based on the position information of the faces and vertices within the second neighborhood is calculated. l Since the original triangular mesh model contains significant noise, which can affect subsequent feature extraction, the normal vector n calculated above can be modified here. s and normal vector n l Filtering is performed to eliminate the influence of noisy data, thereby obtaining the first normal vector of the first neighborhood after filtering. The second normal vector of the second neighborhood

[0075] Furthermore, a preset difference threshold σ for the normal vectors is used to traverse the triangular mesh model. If the absolute value of the difference between the first target normal vector and the second target normal vector is less than the preset difference threshold, then the face and the vertex are determined to be located in a non-feature region. If the absolute value of the difference between the first target normal vector and the second target normal vector is greater than the preset difference threshold, then the face and the vertex are determined to be located in a feature region. The expression is as follows:

[0076]

[0077] The above scheme divides the triangular mesh model into feature regions and non-feature regions, and analyzes the first and second neighborhoods respectively based on the normal difference method, thereby realizing feature extraction of the triangular mesh model.

[0078] Furthermore, the filtering method is a two-sided normal filtering method, which is a nonlinear filtering method that can effectively smooth the triangular mesh model while preserving edge features.

[0079] In one embodiment, the bilateral normal filtering formula is:

[0080]

[0081] Here, nor() represents the normalization operation. c represents the normal vector after filtering in the first sub-neighborhood. j n represents the center point of the neighborhood patch. j W represents the normal vector of the neighborhood patch. s W represents the spatial domain weight. r This represents the range weight.

[0082] In one possible embodiment, the W s and W r The calculation formulas are as follows:

[0083]

[0084]

[0085] Wherein, the standard deviation σ s σ represents the spatial similarity of neighboring patches. c This represents the normal similarity of the area patches.

[0086] Figure 3 This is a schematic diagram of the scenario selected for the first sub-domain in this invention. Figure 4 This is a schematic diagram illustrating the effect of an embodiment of an adaptive feature-preserving mesh denoising method.

[0087] Reference Figure 3 and Figure 4As shown, this technical solution uses an adaptive feature-preserving neighborhood selection method to select the neighborhood of faces and vertices. That is, the first sub-neighborhood is the set of first faces connected to the feature vertex. The first sub-neighborhood of the edge in the feature region includes the second face connected to the edge and the third face sharing the edge with the second face. The traditional neighborhood selection method of the triangular mesh model takes the current face as the center point and selects the face connected to the vertex and edge of the current face as the neighborhood face. However, when the traditional neighborhood method is applied to the feature region, it considers both feature faces and non-feature faces at the same time. After multiple iterations, the feature blurring phenomenon will occur. The adaptive feature-preserving neighborhood selection method can smooth the triangular mesh face while preserving the edge features, thereby improving the mesh quality and visual effect of the triangular mesh model.

[0088] Furthermore, the filtering process for the sub-neighborhoods in the sub-neighborhood set includes:

[0089] The filtered result with the smallest difference from the unfiltered normal vector is selected as the normal vector of the sub-neighborhood, and the selection formula is as follows:

[0090]

[0091] Where, n i This represents the normal vector before sub-neighborhood filtering.

[0092] Understandably, calculating the normal vector in non-feature regions using the existing neighborhood selection method, and applying the adaptive feature-preserving neighborhood selection method in feature regions, eliminates the need for additional cumbersome feature direction calculations, thus enabling fast and effective mesh smoothing.

[0093] In one possible implementation, the formula for calculating the vertex coordinates in updating the vertex coordinates within the feature region based on the normal vector is:

[0094]

[0095] in, These are the updated vertex coordinates.

[0096] For feature regions, considering the similarity between normal vectors and projecting neighboring points onto the plane formed by the normal vectors can reduce the error of the triangular mesh model.

[0097] One possible implementation also includes:

[0098] Obtain the second sub-neighborhood set corresponding to each non-feature patch in the non-feature region, and perform filtering processing on the second sub-neighborhood in the second sub-neighborhood set to obtain the non-feature normal vector of each feature vertex after filtering processing;

[0099] The second sub-neighborhood is the fourth facet connected to the non-feature facet.

[0100] In one possible implementation, the method further includes updating the coordinates of non-feature vertices within the non-feature region based on the non-feature normal vector, wherein the formula for calculating the coordinates of the non-feature vertices is:

[0101]

[0102] in, For the updated non-feature vertex coordinates, v i The coordinates of the non-feature vertices before the update are given, α is the update step size, and W is the value of the vertex. ij W is the weight function for non-feature normal vector similarity and neighborhood points. ij The calculation formula is:

[0103]

[0104] Where δ is the radius of the triangular mesh model.

[0105] To remove the influence of noise data when calculating the coordinates of vertices in non-feature regions, the vertices in non-feature regions are restricted to moving only along the direction of the normal vector, which can maintain the smoothness of the plane within the non-feature region.

[0106] Figure 5 This is a schematic diagram of an embodiment of the mesh denoising device based on adaptive feature preservation provided by the present invention.

[0107] Reference Figure 5 According to another aspect of this application, a mesh denoising apparatus based on adaptive feature preservation is also provided, comprising:

[0108] The feature calculation module 501 is used to determine the feature regions and non-feature regions in the triangular mesh model based on the normal difference method.

[0109] The normal filtering module 502 is used to determine the first sub-neighborhood set corresponding to each feature vertex in the feature region using an adaptive feature-preserving neighborhood selection method, and to filter the first sub-neighborhood in the first sub-neighborhood set to obtain the feature normal vector of each feature vertex after filtering.

[0110] Vertex update module 503 is used to update the vertex coordinates within the feature region based on the feature normal vector;

[0111] Wherein, the first sub-neighborhood is the first face connected to the feature vertex, and the first sub-neighborhood of the edge in the feature region is the second face connected to the edge and the third face sharing the edge with the second face.

[0112] The beneficial effects of the above embodiments are as follows: the feature calculation module 501 determines the feature region and non-feature region in the triangular mesh model based on the normal difference method, which can quickly classify mesh features without relying on complex feature direction estimation; the normal filtering module 502 uses the adaptive feature-preserving neighborhood selection method to determine the first sub-neighborhood in the feature region and performs filtering processing to obtain the feature normal vector of each feature vertex after filtering processing; the vertex update module 503 updates the vertex coordinates in the feature region based on the feature normal vector, which can achieve smooth correction of the mesh model method vector while minimizing the difference between the filtered result and the original normal vector, effectively reducing the loss of feature information and improving the accuracy and visual effect of the mesh model.

[0113] like Figure 6 As shown, the present invention also provides a mesh denoising device 6600 based on adaptive feature preservation. The mesh denoising device 6600 based on adaptive feature preservation includes a processor 601601, a memory 602602, and a display 603603. Figure 5 Only a portion of the components of the adaptive feature-preserving mesh denoising device 600600 are shown; however, it should be understood that implementation of all shown components is not required, and more or fewer components may be implemented instead.

[0114] In some embodiments, memory 602 can be an internal storage unit of the adaptive feature-preserving mesh denoising device 600, such as a hard disk or memory of the adaptive feature-preserving mesh denoising device 600. In other embodiments, memory 602 can also be an external storage device of the adaptive feature-preserving mesh denoising device 600, such as a pluggable hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the adaptive feature-preserving mesh denoising device 600.

[0115] Furthermore, the memory 602 may include both internal storage units of the adaptive feature-preserving mesh denoising device 600 and external storage devices. The memory 602 is used to store application software and various types of data for which the adaptive feature-preserving mesh denoising device 600 is installed.

[0116] In some embodiments, processor 601 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 602 or process data, such as the adaptive feature-preserving mesh denoising method of the present invention.

[0117] In some embodiments, display 603 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 603 is used to display information from the adaptive feature-preserving mesh denoising apparatus 600 and to display a user interface for visualization. Components 601-603 of the adaptive feature-preserving mesh denoising apparatus 600 communicate with each other via a system bus.

[0118] In some embodiments of the present invention, when the processor 601 executes the adaptive feature-preserving mesh denoising method in the memory 602, the following steps can be implemented:

[0119] The characteristic and non-characteristic regions in the triangular mesh model are determined based on the normal difference method.

[0120] An adaptive feature-preserving neighborhood selection method is used to determine the first sub-neighborhood set corresponding to each feature vertex in the feature region, and the first sub-neighborhood in the first sub-neighborhood set is filtered to obtain the feature normal vector of each feature vertex after filtering.

[0121] Update the vertex coordinates within the feature region based on the feature normal vector;

[0122] Wherein, the first sub-neighborhood is the first face connected to the feature vertex, and the first sub-neighborhood of the edge in the feature region is the second face connected to the edge and the third face sharing the edge with the second face.

[0123] It should be understood that when the processor 601 executes the adaptive feature-preserving mesh denoising program in the memory 602, in addition to the functions mentioned above, it can also perform other functions, as detailed in the description of the corresponding method embodiments above.

[0124] Accordingly, this application also provides a computer-readable storage medium for storing a computer-readable program or instruction. When the program or instruction is executed by a processor, it can implement the steps or functions of the mesh denoising method based on adaptive feature preservation provided in the above-described method embodiments.

[0125] The above embodiments provide a mesh denoising method based on adaptive feature preservation, which can realize the technical solutions described in the above embodiments of the mesh denoising device based on adaptive feature preservation. The specific implementation principles of each module or unit can be found in the corresponding content in the embodiments of the mesh denoising device based on adaptive feature preservation, and will not be repeated here.

[0126] The above provides a detailed description of the adaptive feature-preserving mesh denoising device provided by the present invention. Specific examples have been used to illustrate the principle and implementation of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation and application scope based on the idea of ​​the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A mesh denoising method based on adaptive feature preservation, characterized in that, include: The characteristic and non-characteristic regions in the triangular mesh model are determined based on the normal difference method. An adaptive feature-preserving neighborhood selection method is used to determine the first sub-neighborhood set corresponding to each feature vertex in the feature region, and the first sub-neighborhood in the first sub-neighborhood set is filtered to obtain the feature normal vector of each feature vertex after filtering. Obtain the second sub-neighborhood set corresponding to each non-feature patch in the non-feature region, and perform filtering on the second sub-neighborhood in the second sub-neighborhood set to obtain the non-feature normal vector of each non-feature vertex after filtering. The coordinates of vertices within the feature region are updated based on the feature normal vector, and the coordinates of non-feature vertices within the non-feature region are updated based on the non-feature normal vector; Wherein, the first sub-neighborhood is a first face connected to the feature vertex, the first sub-neighborhood of the edge in the feature region includes a second face connected to the edge and a third face sharing an edge with the second face, the second sub-neighborhood is a fourth face connected to the non-feature face, and the edge is used to connect two feature vertices; The determination of feature and non-feature regions in the triangular mesh model based on the normal difference method includes: Construct the topological relationships of the triangular mesh model data to obtain the connection relationship diagram between the facets and vertices; Determine the first neighborhood of the face and the vertex, and calculate the first normal vector of the triangular mesh model based on the first neighborhood according to the position information of the adjacent face and vertex; Determine the second neighborhood of the face and the vertex, and calculate the second normal vector of the triangular mesh model based on the second neighborhood according to the position of the face and the vertex in the second neighborhood; The first normal vector and the second normal vector are filtered to obtain the first target normal vector and the second target normal vector after filtering. If the absolute value of the difference between the first target normal vector and the second target normal vector is less than a preset difference threshold, then the face and the vertex are determined to be located in a non-feature region. If the absolute value of the difference between the first target normal vector and the second target normal vector is greater than the preset difference threshold, then the face and the vertex are determined to be located in a feature region. The range of the second neighborhood is larger than the range of the first neighborhood.

2. The mesh denoising method based on adaptive feature preservation according to claim 1, characterized in that, The filtering method described is a bilateral normal filtering method.

3. The mesh denoising method based on adaptive feature preservation according to claim 2, characterized in that, The bilateral normal filtering formula is as follows: Here, nor() represents the normalization operation. Indicates the first sub-neighborhood. i The normal vector after patch filtering Indicates the first j The center point of each neighboring surface Indicates the first i The center point of each piece of surface Indicates the first i Normal vector of each patch before filtering Indicates the first j Normal vector of each neighborhood patch before filtering Represents the spatial domain weights. Represents the range weights and standard deviations. Represents the spatial similarity of neighboring patches. This represents the normal similarity of neighboring faces.

4. The mesh denoising method based on adaptive feature preservation according to claim 3, characterized in that, The filtering process for the first sub-neighborhood in the first sub-neighborhood set includes: The filtered result with the smallest difference from the original normal vector is selected as the normal vector of the first sub-neighborhood. The selection formula is as follows: .

5. The mesh denoising method based on adaptive feature preservation according to claim 1, characterized in that, In updating the vertex coordinates within the feature region based on the normal vector, the formula for calculating the vertex coordinates is as follows: in, These are the updated vertex coordinates.

6. The mesh denoising method based on adaptive feature preservation according to claim 1, characterized in that, The formula for calculating the coordinates of the non-feature vertices is as follows: in, These are the updated coordinates of the non-feature vertices. These are the coordinates of the non-feature vertices before the update. To update the step size, The weight function is the similarity of non-feature normal vectors and the weight function of neighborhood points. The calculation formula is: in, The radius of the triangular mesh model.

7. A mesh denoising device based on adaptive feature preservation, characterized in that, include: The feature calculation module is used to determine the feature regions and non-feature regions in the triangular mesh model based on the normal difference method; The normal filtering module is used to determine the first sub-neighborhood set corresponding to each feature vertex in the feature region using an adaptive feature-preserving neighborhood selection method, and to filter the first sub-neighborhood in the first sub-neighborhood set to obtain the feature normal vector of each feature vertex after filtering. Obtain the second sub-neighborhood set corresponding to each non-feature patch in the non-feature region, and perform filtering on the second sub-neighborhood in the second sub-neighborhood set to obtain the non-feature normal vector of each non-feature vertex after filtering. The vertex update module is used to update the vertex coordinates within the feature region based on the feature normal vector, and to update the non-feature vertex coordinates within the non-feature region based on the non-feature normal vector; Wherein, the first sub-neighborhood is the first face connected to the feature vertex, the first sub-neighborhood of the edge in the feature region is the second face connected to the edge and the third face sharing the edge with the second face, the second sub-neighborhood is the fourth face connected to the non-feature face, and the edge is used to connect two feature vertices; The determination of feature and non-feature regions in the triangular mesh model based on the normal difference method includes: Construct the topological relationships of the triangular mesh model data to obtain the connection relationship diagram between the facets and vertices; Determine the first neighborhood of the face and the vertex, and calculate the first normal vector of the triangular mesh model based on the first neighborhood according to the position information of the adjacent face and vertex; Determine the second neighborhood of the face and the vertex, and calculate the second normal vector of the triangular mesh model based on the second neighborhood according to the position of the face and the vertex in the second neighborhood; The first normal vector and the second normal vector are filtered to obtain the first target normal vector and the second target normal vector after filtering. If the absolute value of the difference between the first target normal vector and the second target normal vector is less than a preset difference threshold, then the face and the vertex are determined to be located in a non-feature region. If the absolute value of the difference between the first target normal vector and the second target normal vector is greater than the preset difference threshold, then the face and the vertex are determined to be located in a feature region. The range of the second neighborhood is larger than the range of the first neighborhood.

Images