New visual angle rendering method based on quadric surface Gaussian splashing and related device

Abstract

The invention discloses a new visual angle rendering method based on quadric surface Gaussian splashing and a related device. The method comprises the following steps: acquiring an image to be rendered; the to-be-rendered image is input into a trained quadric surface Gaussian splash model to obtain a rendered color image, the quadric surface Gaussian splash model is formed by training a real image and a corresponding color image, the real image is rendered by using a color image rendering formula to obtain the color image, and the color image is rendered by using a color image rendering formula. The rendering effect of the method and the related device is good.

Description

Technical Field

[0001] The invention belongs to the field of image processing and relates to a new viewing angle rendering method based on quadratic surface Gaussian splashing and a related device. Background Art

[0002] Dense reconstruction and new perspective rendering of scenes are important research directions in the fields of computer vision and computer graphics. It aims to recover dense structures from multiple images of the scene from different viewpoints and render realistic images from free viewpoints. Recently, 3D Gaussian splashing has surpassed the rendering quality and rendering speed of several methods based on neural radiance fields by combining traditional splashing techniques in graphics with end-to-end optimization techniques. Subsequently, Gaussian splashing methods have developed rapidly in dynamic reconstruction, editing, and large scene reconstruction.

[0003] However, since the splashing technique loses the z-axis component of the Gaussian distribution and uses approximation during rendering, it can render multi-view consistent textures even at the wrong geometric position, which ultimately makes it difficult to ensure the accuracy of its geometric properties only through photometric consistency error optimization. In subsequent work, instead of querying the Gaussian weight on the image plane, the intersection of the ray and the Gaussian principal element in three-dimensional space and querying the corresponding Gaussian weight became the key element for restoring accurate scene geometry using the Gaussian splashing method. In the method with the Gaussian ellipsoid as the principal element, the calculation of the intersection of the ray and the principal element is usually related to the line of sight direction, so it is impossible to provide a multi-view consistent normal. In the method with the Gaussian disk as the principal element, the intersection of the ray and the principal element is the intersection of the line of sight and the disk. At the same time, this method can provide multi-view consistent geometry, so that multi-view geometric consistency information can be more conveniently introduced. However, the disk plane is only a first-order linear approximation of the scene surface, which makes it insufficient for fitting complex surfaces, resulting in over-smoothed reconstruction results, and ultimately leads to poor rendering effects. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a new perspective rendering method and related devices based on quadratic surface Gaussian splashing, which have better rendering effects.

[0005] To achieve the above object, the present invention discloses a new perspective rendering method based on quadratic surface Gaussian splashing, comprising:

[0006] Get the image to be rendered;

[0007] The image to be rendered is input into the trained quadratic surface Gaussian splash model to obtain a rendered color image, wherein the quadratic surface Gaussian splash model is trained by a real image and its corresponding color image, and the real image is rendered using a color graph rendering formula to obtain the color image.

[0008] The further improvement of the new perspective rendering method based on quadratic surface Gaussian splashing of the present invention is:

[0009] Furthermore, before inputting the image to be rendered into the trained quadratic surface Gaussian splash model, the method further includes:

[0010] Build a dataset;

[0011] Construct a quadratic surface Gaussian splash model;

[0012] The quadratic surface Gaussian splash model is trained using the data set to obtain a trained quadratic surface Gaussian splash model.

[0013] Furthermore, the process of constructing the data set is:

[0014] Obtain several real images;

[0015] Generate an internal reference, a sparse point cloud and an external reference using the real image;

[0016] According to the generated internal parameters, sparse point cloud and external parameters, a color image is drawn using the color rendering formula;

[0017] A dataset is constructed based on each real image and its corresponding color image.

[0018] Furthermore, the color rendering formula is:

[0019]

[0020] Among them, α i is the opacity of each Gaussian principal component, c i is the color of each Gaussian principal component, C(p) represents the rendered RGB color of pixel p, G i (p) represents the Gaussian weight of the intersection point between the homogeneous ray corresponding to pixel p and the i-th quadratic surface.

[0021] Furthermore, the loss function in the process of using the data set to train the quadratic surface Gaussian splash model to obtain the trained quadratic surface Gaussian splash model is:

[0022]

[0023] λ K (K(u,v))=1-sigmoid(ln(|K(u,v)|+∈))

[0024] L Kn (u,v)=λ K (K(u,v))L n(u,v)(14)

[0025] Among them, α i is the opacity of each Gaussian principal component, c i is the color of each Gaussian principal component, C(p) represents the rendered RGB color of pixel p, G i (p) represents the Gaussian weight of the intersection point between the homogeneous ray corresponding to the pixel point p and the i-th quadratic surface, K(u,v) represents the curvature value rendered at the pixel point (u,v), ∈ represents a very small positive value, and λ K represents the curvature factor.

[0026] Furthermore, the curvature value K(p) rendered at the pixel point (p) is:

[0027]

[0028] The present invention discloses a new perspective rendering system based on quadratic surface Gaussian splashing, comprising:

[0029] An acquisition module, used to acquire an image to be rendered;

[0030] A rendering module is used to input the image to be rendered into a trained quadratic surface Gaussian splash model to obtain a rendered color image, wherein the quadratic surface Gaussian splash model is trained by a real image and its corresponding color image, and the real image is rendered using a color graph rendering formula to obtain the color image.

[0031] The further improvement of the new perspective rendering system based on quadratic surface Gaussian splashing of the present invention is:

[0032] Furthermore, it also includes:

[0033] The first building module is used to build a data set;

[0034] The second building module is used to build a quadratic surface Gaussian splash model;

[0035] A training module is used to train the quadratic surface Gaussian splash model using the data set to obtain a trained quadratic surface Gaussian splash model.

[0036] The present invention discloses a computer device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the new perspective rendering method based on quadratic surface Gaussian splashing are implemented.

[0037] The present invention discloses a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the new perspective rendering method based on quadratic surface Gaussian splashing are implemented.

[0038] The present invention has the following beneficial effects:

[0039] During specific operation, the new perspective rendering method and related device based on quadratic surface Gaussian splashing described in the present invention input the image to be rendered into the trained quadratic surface Gaussian splashing model to obtain a rendered color image, wherein quadratic surface Gaussian splashing is used and the quadratic surface is used as a scene primitive, which has stronger geometric fitting ability to solve the problem of over-smoothing reconstruction in previous methods, thereby improving the rendering effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0041] Figure 1 A schematic diagram of the method of the present invention;

[0042] Figure 2 The normal map and curvature map output by the method of the present invention;

[0043] Figure 3 The comparison results of quadratic Gaussian splash (QGS), planar Gaussian splash (2DGS) and Gaussian opacity field (GOF) on public datasets are shown in the figure;

[0044] Figure 4 This is a diagram of the reconstruction results of the present invention on a public data set;

[0045] Figure 5 This is a comparison chart of the reconstruction results of the present invention and 2DGS in an outdoor aerial photography large scene;

[0046] Figure 6 This is a rendering result diagram of the present invention in indoor and outdoor scenes;

[0047] Figure 7 It is a system structure diagram of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are 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 creative work are within the scope of protection of the present invention.

[0049] In the description of the present invention, it should be understood that the terms “include” and “comprises” indicate the presence of described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0050] It should also be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0051] It should be further understood that the term "and / or" used in the present specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present invention generally indicates that the associated objects are in an "or" relationship.

[0052] It should be understood that, although the terms first, second, third, etc. may be used to describe preset ranges, etc. in the embodiments of the present invention, these preset ranges should not be limited to these terms. These terms are only used to distinguish preset ranges from each other. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0053] The word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to determining" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)", depending on the context.

[0054] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention described and shown in the drawings here can usually be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0055] Various structural schematic diagrams of the embodiments disclosed in the present invention are shown in the accompanying drawings. These figures are not drawn to scale, and some details are magnified and some details may be omitted for the purpose of clear expression. The shapes of various regions and layers shown in the figures and the relative sizes and positional relationships therebetween are only exemplary, and may deviate in practice due to manufacturing tolerances or technical limitations, and those skilled in the art may additionally design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0056] Embodiment 1

[0057] The new perspective rendering method based on quadratic surface Gaussian splashing of the present invention comprises the following steps:

[0058] 1) Build a dataset;

[0059] 11) Obtain several real images;

[0060] 12) Generate internal parameters, sparse point cloud and external parameters using the real image, wherein the internal parameters include focal length, principal point and distortion; the external parameters include the 6-DOF position of the camera;

[0061] 13) Rendering to generate color images;

[0062] The specific process of step 13) is as follows:

[0063] Use KNN to calculate the K neighborhood points of each point in the sparse point cloud, and use the average distance of the K neighborhood points to initialize the scale of the quadratic parabola, and then randomly initialize the posture of the quadratic parabola, such as Figure 1 As shown in Formula 1, the expression of the quadratic parabola is:

[0064]

[0065] Among them, s1 and s2 represent the scale factors of the quadratic parabola in the x-axis direction and the y-axis direction, s3 represents the scale factor of the quadratic parabola in the z-direction, ρ and θ represent the polar diameter and polar angle after the Cartesian coordinate system where the quadratic parabola is located is converted into the polar coordinate system. From formula (1), we can get the point on the quadratic parabola Calculate the intersection point between the ray corresponding to each pixel in the image and the quadratic parabola, and calculate the geodesic distance l(a(θ0),ρ0) from the intersection point to the vertex of the quadratic parabola based on the position of the intersection point on the quadratic parabola; Figure 1 As shown:

[0066]

[0067] Among them, a(θ0) is the coefficient of the quadratic term of the intersection line, ρ0 is the horizontal distance from the intersection to the origin of the surface, t is the integral variable in the θ0 direction, and for simplicity, u=2aρ0 is set as the intermediate variable. The Gaussian distribution on the surface is determined according to the geodesic distance. The Gaussian distribution formula on the surface is:

[0068]

[0069] Where σ(θ0) is the standard deviation of the Gaussian distribution in the θ0 direction, which is determined by the scale of the principal element of the quadratic parabola. l(a(θ0)) is the geodesic arc length in the θ0 direction calculated by equation (2). Indicates intersection The Gaussian distribution value at, specifically, σ(θ0) is:

[0070]

[0071] According to the Gaussian weight of the intersection point between the ray and the Gaussian principal element, the color image is drawn using the color volume rendering formula shown in formula (5):

[0072]

[0073] Among them, α i is the opacity of each Gaussian principal component, c i is the color of each Gaussian principal component, C(p) represents the rendered RGB color of pixel p, G i (p) represents the Gaussian weight of the intersection point between the homogeneous ray corresponding to pixel p and the i-th quadratic surface.

[0074] It should be noted that the calculation process of the intersection of each ray and the Gaussian principal element is:

[0075] The ray equation is constructed as:

[0076]

[0077] in, are the optical center of the camera and the ray direction corresponding to the pixel respectively.

[0078]

[0079]

[0080] At 2 +Bt+c=0

[0081]

[0082] In equation (7), the ray-quadratic surface intersection equation is first listed and simplified to a general form: Then, by using the closed-form solution of the quadratic equation, we can calculate the closest intersection point of the ray with the surface, t n and the farther intersection point t f .

[0083] Check the nearest intersection point t n The geodesic distance l n Is it less than or equal to the standard deviation σ in the direction of the intersection? n , if it is less than or equal to, then only the closer intersection point t is selected n For volume rendering, otherwise, check the farther intersection point t f The geodesic distance l f Is it less than or equal to the standard deviation σ in the direction of the intersection? f , if it is less than or equal to, then select the farther intersection point t f Used for volume rendering, otherwise, the ray is considered to have no intersection with the principal element.

[0084] 14) Constructing a dataset of real images and their corresponding rendered color images;

[0085] 2) Constructing a quadratic surface Gaussian splash model;

[0086] 3) using the data set to train the quadratic surface Gaussian splash model to obtain a trained quadratic surface Gaussian splash model;

[0087] In step 3), before training the quadratic surface Gaussian splash model, it is necessary to determine the loss function in the training process. The specific determination process is:

[0088] Due to the second-order nature of quadratic surfaces, the normal at the intersection can be calculated simultaneously With curvature for:

[0089]

[0090] in, represents the horizontal and vertical coordinates of the intersection point in the local coordinate system of the quadratic surface, The depth map, normal map and curvature map on the imaging plane are obtained by drawing the volume rendering function, such as Figure 2 As shown, the rendering formulas for depth, normal and curvature are:

[0091]

[0092] Among them, d i ,n i ,K i and G i (p) are the depth, normal, curvature and Gaussian weight of the intersection point between the homogeneous ray corresponding to pixel p and the i-th quadratic surface. The rendered color image is subtracted from the real image to obtain L c ;

[0093] Applying the deep correction loss function shown in formula (11) encourages each quadratic surface to fit together to form a compact geometric fit;

[0094]

[0095] in, Use the normal consistency loss function to encourage the orientation of each quadratic surface to be locally consistent;

[0096]

[0097] in, represents the normal of each quadratic surface and the line of sight intersection, p represents the three-dimensional point projected from the rendered depth map to the camera coordinate system, The operators represent the difference of the three-dimensional point p along the pixel coordinate axis u, v, respectively, × represents the vector outer product, and N(u,v) is the normal differenced from the rendered depth map at the pixel (u,v). Finally, the loss function L used to supervise the normal of each quadratic surface is obtained. n , but formula (13) is highly dependent on the local plane assumption of the three-dimensional point, which is often not true in areas with large depth changes, such as the edge of an object. Therefore, the present invention proposes a curvature-guided normal consistency loss function as shown in formula (14):

[0098] λ K (K(u,v))=1-sigmoid(ln(|K(u,v)|+∈))

[0099] L Kn (u,v)=λ K (K(u,v))L n (u,v)(14)

[0100] Where K(u,v) represents the curvature value rendered by equation (10) at the pixel (u,v), ∈ represents a very small positive value, such as 1e-6, and λ K represents the curvature factor, which is negatively correlated with the curvature value. K Combined with formula (12), the final curvature-guided normal consistency loss function L is obtained Kn , where when the curvature value is large, the depth changes greatly and the local plane assumption does not hold. Therefore, the normal supervision is weakened. Formula (12) is the normal consistency loss function; Formula (13) is the differential normal formula.

[0101] Using equations (5), (11) and (14) as loss functions, and using the data set, the quadratic surface Gaussian splash model is trained to obtain a quadratic surface Gaussian splash model;

[0102] 4) Obtaining a real image to be rendered, and inputting the real image to be rendered into the trained quadratic surface Gaussian splash model to obtain a rendered color image.

[0103] It should be noted that the present invention uses quadratic surfaces as scene primitives, which have stronger geometric fitting capabilities to solve the problem of over-smoothing reconstruction in previous methods. The introduction of geodesic distance to establish Gaussian distribution enables a single primitive to fit more complex textures to solve the problem of blurry rendering in previous methods. A stricter depth sorting is introduced in the quadratic surface Gaussian splash with higher degrees of freedom to solve the problem of inconsistent multi-view rendering results in previous methods.

[0104] Simulation experiment

[0105] This experiment uses the open source TSDF method to fuse depth maps rendered from multiple viewpoints into a triangular patch model and compares it with open source advanced algorithms.

[0106] Figure 2 The normal map (left) and curvature map (right) output by the algorithm proposed in this invention. The normal map reflects the local orientation information of the scene's geometric surface, and the curvature map represents the degree of curvature of the scene's geometric surface. The darker the blue, the higher the curvature and the greater the degree of curvature. Conversely, the smaller the curvature, the flatter the surface. Figure 3 The comparison results of quadratic Gaussian splash (QGS) and planar Gaussian splash (2DGS) and Gaussian opacity field (GOF) on the public datasets DTU, TNT, and Mip-NeRF 360 are shown; Figure 4 This is the reconstruction result of the present invention on the public data sets TNT and Mip-NeRF 360; Figure 5 This is a comparison chart of the reconstruction results of the present invention and 2DGS in an outdoor aerial photography large scene; Figure 6 This is a rendering result diagram of the present invention in indoor and outdoor scenes.

[0107] Embodiment 2

[0108] refer to Figure 7 The new perspective rendering system based on quadratic surface Gaussian splashing of the present invention comprises:

[0109] The first building module is used to build a data set;

[0110] The second building module is used to build a quadratic surface Gaussian splash model;

[0111] A training module, used for training the quadratic surface Gaussian splash model using the data set to obtain a trained quadratic surface Gaussian splash model;

[0112] An acquisition module, used to acquire an image to be rendered;

[0113] A rendering module is used to input the image to be rendered into a trained quadratic surface Gaussian splash model to obtain a rendered color image, wherein the quadratic surface Gaussian splash model is trained by a real image and its corresponding color image, and the real image is rendered using a color graph rendering formula to obtain the color image.

[0114] The division of modules in the embodiments of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each embodiment of the present application may be integrated into a processor, or may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0115] Embodiment 3

[0116] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the new perspective rendering method based on quadratic surface Gaussian splashing are implemented, including: obtaining an image to be rendered; inputting the image to be rendered into a trained quadratic surface Gaussian splashing model to obtain a rendered color image, wherein the quadratic surface Gaussian splashing model is trained by a real image and its corresponding color image, and the real image is rendered using a color image rendering formula to obtain the color image. The memory may include a memory, such as a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk memory, etc. The processor, the network interface, and the memory are interconnected through an internal bus, and the internal bus may be an industrial standard architecture bus, a peripheral component interconnection standard bus, an extended industrial standard architecture bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory is used to store a program. Specifically, the program may include a program code, and the program code includes a computer operation instruction. The memory may include a memory and a non-volatile memory, and provide instructions and data to the processor.

[0117] Embodiment 4

[0118] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the new perspective rendering method based on quadratic surface Gaussian splashing are implemented. Specifically, it includes: obtaining an image to be rendered; inputting the image to be rendered into a trained quadratic surface Gaussian splashing model to obtain a rendered color image, wherein the quadratic surface Gaussian splashing model is trained by a real image and its corresponding color image, and the real image is rendered using a color image rendering formula to obtain the color image. The computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may include a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may include a read-only memory (ROM), a hard disk, a flash memory, an optical disk, a magnetic disk, etc.

[0119] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0120] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0121] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0122] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0123] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and disclosure of the invention. This application is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not disclosed by the present invention. The specification and examples are to be considered exemplary only, and the true scope and spirit of the present invention are indicated by the following claims.

[0124] It should be understood that the present invention is not limited to the exact construction 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 invention is limited only by the appended claims.

[0125] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any simple modification, change and equivalent structural change made to the above embodiment based on the technical essence of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A new perspective rendering method based on quadratic surface Gaussian splashing, characterized in that: include: Get the image to be rendered; The image to be rendered is input into the trained quadratic surface Gaussian splash model to obtain a rendered color image, wherein the quadratic surface Gaussian splash model is trained by a real image and its corresponding color image, and the real image is rendered using a color graph rendering formula to obtain the color image.

2. The new perspective rendering method based on quadratic surface Gaussian splashing according to claim 1 is characterized in that: Before inputting the image to be rendered into the trained quadratic surface Gaussian splash model, the method further includes: Build a dataset; Construct a quadratic surface Gaussian splash model; The quadratic surface Gaussian splash model is trained using the data set to obtain a trained quadratic surface Gaussian splash model.

3. The new perspective rendering method based on quadratic surface Gaussian splashing according to claim 1 is characterized in that: The process of constructing the data set is as follows: Obtain several real images; Generate an internal reference, a sparse point cloud and an external reference using the real image; According to the generated internal parameters, sparse point cloud and external parameters, a color image is drawn using the color rendering formula; A dataset is constructed based on each real image and its corresponding color image.

4. The new perspective rendering method based on quadratic surface Gaussian splashing according to claim 1 is characterized in that: The color rendering formula is: Among them, α i is the opacity of each Gaussian principal component, c i is the color of each Gaussian principal component, C(p) represents the rendered RGB color of pixel p, G i (p) represents the Gaussian weight of the intersection point between the homogeneous ray corresponding to pixel p and the i-th quadratic surface.

5. The new perspective rendering method based on quadratic surface Gaussian splashing according to claim 2 is characterized in that: The loss function in the process of using the data set to train the quadratic surface Gaussian splash model to obtain the trained quadratic surface Gaussian splash model is: λ K (K(u,v))=1-sigmoid(ln(|K(u,v)|+∈)) L Kn (u,v)=λ K (K(u,v))L n (u,v)(14) Among them, α i is the opacity of each Gaussian principal component, c i is the color of each Gaussian principal component, C(p) represents the rendered RGB color of pixel p, G i (p) represents the Gaussian weight of the intersection point between the homogeneous ray corresponding to the pixel point p and the i-th quadratic surface, K(u,v) represents the curvature value rendered at the pixel point (u,v), ∈ represents a very small positive value, and λ K represents the curvature factor.

6. The new perspective rendering method based on quadratic surface Gaussian splashing according to claim 5 is characterized in that: The curvature value K(p) rendered at the pixel point (p) is:

7. A new perspective rendering system based on quadratic surface Gaussian splashing, characterized in that: include: An acquisition module, used to acquire an image to be rendered; A rendering module is used to input the image to be rendered into a trained quadratic surface Gaussian splash model to obtain a rendered color image, wherein the quadratic surface Gaussian splash model is trained by a real image and its corresponding color image, and the real image is rendered using a color graph rendering formula to obtain the color image.

8. The new perspective rendering system based on quadratic surface Gaussian splashing according to claim 7, characterized in that: Also includes: The first building module is used to build a data set; The second building module is used to build a quadratic surface Gaussian splash model; A training module is used to train the quadratic surface Gaussian splash model using the data set to obtain a trained quadratic surface Gaussian splash model.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the new perspective rendering method based on quadratic surface Gaussian splatting as described in any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the new perspective rendering method based on quadratic surface Gaussian splatting as described in any one of claims 1 to 6 are implemented.

Images