A material generation method and system based on physical rendering

Abstract

The application discloses a material generation method and system based on physical rendering, and solves the problem that a fixed refractive index cannot simulate dynamic effects such as water flow and smoke by using a space-time variable refractive index field. The application solves the problem that the internal multiple scattering is ignored by a thin surface BTDF, leading to the energy underestimation of materials such as frosted glass and the visual effect of 'rigid' by using an energy compensation term. Meanwhile, the problem that the physical method such as path tracing has large calculation overhead and is difficult to render dynamic fluid in real time is solved by using a screen space operation, a neural network acceleration and the like.

Description

Technical Field

[0001] This invention relates to the field of image rendering, and more specifically to a material generation method and system based on physically based rendering. Background Technology

[0002] In the field of computer vision and graphics, realistic image rendering technology has extremely important application value. Whether in film and television special effects, game engines, architectural design visualization, or emerging digital content industries such as virtual reality (VR), augmented reality (AR), and the metaverse, realistic image synthesis capabilities are the core technological support for improving user experience and enhancing immersion. Currently, image generation technologies are mainly divided into two categories: one is traditional computer graphics methods based on physical modeling, which can achieve highly realistic effects such as lighting, material reflection, and shadows, and is widely used in the film industry and high-end simulation systems; the other is methods based on artificial intelligence learning that have emerged in recent years, such as diffusion models and generative adversarial networks. These methods, through large-scale data training, can generate high-quality and diverse images in a relatively short time. However, both types of methods have obvious limitations. How to ensure visual quality while maintaining physical consistency and computational efficiency has become the core challenge in the field of image synthesis.

[0003] Existing image representation and rendering methods face numerous limitations in practical applications. Traditional physically based rendering methods rely on precise modeling of scene geometry, material properties, and lighting conditions. While they can generate visually extremely realistic images, they are computationally complex, typically requiring complete 3D scene information, and perform poorly when handling complex surface materials (such as highly reflective, transparent, or translucent objects). For example, the reflection and refraction behavior of transparent materials such as glass, water surfaces, and windows involves complex ray tracing processes, leading to a significant increase in rendering time. On the other hand, while deep learning-based image generation methods have advantages in speed and versatility, they often lack accurate control over lighting, material, and geometric relationships because they do not rely on explicit physical modeling. This can easily result in image outcomes that do not conform to physical laws, such as incorrect shadow directions and abnormal reflection paths. Furthermore, these methods make it difficult to perform structured editing of the generated images; users cannot flexibly adjust lighting intensity, material type, or viewpoint changes, thus limiting their application in professional design and engineering simulation. Summary of the Invention

[0004] To address the aforementioned issues, this patent proposes an extended intrinsic image representation method that overcomes the limitations of traditional intrinsic image decomposition, which only includes reflection and illumination maps. By introducing a transmission map as a key component, it enables image synthesis of transparent materials such as glass and windows. This method not only retains the controllability and realism of physically based modeling but also combines the efficiency and versatility of deep learning in image generation. This novel image representation method achieves accurate modeling and dynamic control of transparent materials while maintaining high image quality, significantly enhancing the realism and operability of image synthesis.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A physically based rendering-based material generation method includes the following steps:

[0007] Based on the rendering equation, the reflection and transmission characteristics of each surface point in the image rendering scene are calculated.

[0008] Based on the reflection and transmission characteristics, a material model is constructed, which includes diffuse reflection, specular reflection and specular transmission components, and introduces metallicity and transparency parameters.

[0009] Using the constructed material model, combined with the spatiotemporal variable refractive index field, a transparent material is modeled.

[0010] Based on the material model and the modeled transparent material, screen space image synthesis is performed to generate the final image.

[0011] Preferably, the rendering equation includes:

[0012]

[0013] Where, ω i ω is the incident direction, ω0 is the exit direction, L(ρ, ω o L(p, ω) represents the emitted radiance along the ω0 direction at position p. i ) indicates that at position p, along ω i Incident radiance in the direction, f(p, ω) o ω i ) represents the two-way scattering distribution function. It is the surface normal vector at point p. This represents the angle ω on the unit sphere for all incident directions. i Integrate the points.

[0014] Preferably, the constructed material model includes:

[0015] fd(p,ω o ωi )=(1-t1)(1-m)f d (p,ω o ω i )+f s (p, ω) o ω i )+t1(Γ t ·f t (p, ω) o ω i ))

[0016] Where, f(p, ω) o ω i ) is the two-way scattering distribution function, t1 is the transparency parameter, m is the metallicity parameter, and f is the... d (p, ω) o ω i ) is the diffuse reflection component, f s (p, ω) o ω i f is the specular reflection component. t (p,ω o ω i ) is the specular transmission component, Γ t It is an energy compensation term.

[0017] Preferred, Γ t Used to compensate for changes in energy distribution caused by dynamic fields:

[0018]

[0019] Where k is the dielectric attenuation coefficient, It is the gradient of the spacetime variable refractive index field. It is the gradient magnitude, ds is the path length of the ray, ∫ ray • is the path integral along the path of the refracted ray.

[0020] Preferably, the specular transmission component includes:

[0021]

[0022] in, It is the extended normal distribution function, F(ht, ω). o ) is the Fresnel transmission coefficient, G(ht, ω) o ω i ) is the geometric decay function;

[0023] In the geometric decay function G(ht, ω) o ω i In the diagram, ht represents the incident direction ω during specular transmission. i and the direction of emission ω oThe half-vector is expressed as follows:

[0024]

[0025] Wherein, η(p,t) is the spatiotemporal variable refractive index field, that is, the surface refractive index that changes with time and space.

[0026] Preferably, a spatiotemporally variable refractive index field η(p,t) is introduced into the screen space rendering to achieve physically accurate dynamic fluid refraction:

[0027]

[0028] Where η(p,t) is the spatiotemporal variable refractive index, η0 is the fundamental refractive index, dimensionless, with a value range of η0 > 1, and Δη is the refractive index fluctuation amplitude, dimensionless, with a value range of [0, 0.5η0]. It's a wave arrow. ω is the three-dimensional spatial coordinate, t is the time, φ is the phase shift, and λ1 and λ2 are the coefficients of the fluid medium term and the complex fluid term.

[0029] Preferably, the steps for synthesizing the final image include:

[0030] I=(1-T)(1-M)I diff +I spec +T(Γ t ·(I tran +ΔE·I sss ))

[0031] Where I represents the final synthesized image, T is the transparency map, and M is the metallic map, representing the metallic properties of each pixel. diff It is an image synthesized from diffuse reflection components, I spec It is an image synthesized from specular reflection components, I tran It is an image synthesized from the specular transmission components, Γ t It is the energy compensation term, ΔE·I sss It is a quasi-Monte Carlo subsurface scattering model.

[0032] Preferably, the quasi-Monte Carlo subsurface scattering model includes:

[0033]

[0034] in, ω represents the surface normal vector. i It is the incident direction, f t It is the specular transmission function, and ΔE is the energy loss rate due to neglecting multiple scattering. It is the spherical harmonic reconstruction operator, f sss (p,ω o ωi E is the scattering function. in It represents the incident irradiance, * represents the convolution operation, and I represents the incident irradiance. sss A composite image representing the subsurface scattering components.

[0035] The present invention also provides a material generation system based on physically based rendering, the system being used to implement the above method, comprising: a calculation module, a first construction module, a second construction module, and a compositing module;

[0036] The calculation module is used to calculate the reflection and transmission characteristics of each surface point in the image rendering scene based on the rendering equation;

[0037] The first construction module is used to construct a material model based on the reflection characteristics and the transmission characteristics. The material model includes diffuse reflection, specular reflection and specular transmission components, and introduces metallicity and transparency parameters.

[0038] The second construction module is used to model transparent materials by utilizing the constructed material model and combining it with a spatiotemporal variable refractive index field.

[0039] The compositing module is used to perform screen space image compositing based on the material model and the modeled transparent material to generate the final image.

[0040] Preferably, the rendering equation includes:

[0041]

[0042] Where, ω i ω is the incident direction, ω0 is the exit direction, L(p, ω o L(p, ω) represents the emitted radiance along the ω0 direction at position p. i ) indicates that at position p, along ω i Incident radiance in the direction, f(p, ω) o ω i ) represents the two-way scattering distribution function. It is the surface normal vector at point p. This represents the angle ω on the unit sphere for all incident directions. i Integrate the points.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0044] This invention addresses the problem of traditional methods using fixed refractive indices, which cannot simulate dynamic effects such as water flow and smoke, by employing a spatiotemporally variable refractive index field. Furthermore, it solves the problem of underestimating the energy of materials like frosted glass and resulting in a "rigid" visual effect by using an energy compensation term to overcome the issue of thin-surface BTDF neglecting internal multiple scattering. Simultaneously, it addresses the computational overhead of physical methods such as path tracing, which makes real-time rendering of dynamic fluids difficult, through screen-space manipulation and neural network acceleration. Attached Figure Description

[0045] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. 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.

[0046] Figure 1 This is a schematic diagram of the material model structure in an embodiment of the present invention. Detailed Implementation

[0047] 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 some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0049] Example 1

[0050] This embodiment provides a material generation method based on physically based rendering, the steps of which include:

[0051] S1. Based on the rendering equation, calculate the reflection and transmission characteristics of each surface point in the image rendering scene.

[0052] The rendering equation is a core equation in computer graphics used to describe the interaction of light in a virtual scene. It is an integral equation describing the light radiation at a point p on a non-emitting surface (i.e., the light radiation emanating from the point towards the camera in the direction ω0). Rendering equation:

[0053]

[0054] Where, ω iωp is the incident direction, i.e., the direction vector from surface point p towards the light source. ω0 is the outgoing direction, i.e., the direction vector from surface point p towards the camera or observer. L(p, ω0) o L(p, ω) represents the emitted radiance along the ω0 direction at position p. i ) indicates that at position p, along ω i The incident radiance in the direction of ω. o ω i ) represents the two-way scattering distribution function (BSDF), which describes the path of light from point p on a surface from the incident direction ω. i Scattering characteristics in the outgoing direction ω0. It is the surface normal vector at point p. This represents the situation on a unit sphere with respect to all possible incident directions ω. i Integrate because light can strike surface point p from any direction.

[0055] The core idea of ​​the rendering equation is that the radiance of light leaving a surface from a certain direction is the sum of the radiance of all light incident on that surface after reflection or transmission. In other words, the outgoing light at each surface point is the sum of the contributions from all possible incident light rays scattered at that point. This scattering is described by the Disney Principled BSDF model, which depends on the surface's material properties, such as diffuse reflection, specular reflection, and refraction.

[0056] The Disney Principled BSDF model is a widely used appearance model in various rendering engines. Based on physical principles and empirical observations, while not strictly adhering to physical precision, it provides a practical and intuitive parametric method for representing reasonable materials, prioritizing artistic control and ease of use. This model supports a variety of materials and lighting effects, including diffuse reflection, subsurface scattering, retroreflection, specular reflection, refraction of transparent surfaces, and gloss compensation for retroreflection.

[0057] S2. Based on reflection and transmission characteristics, construct a material model, which includes diffuse reflection, specular reflection and specular transmission components, and introduce metallicity and transparency parameters.

[0058] This step builds an extended physically based rendering (ePBR) material model, such as... Figure 1 As shown. The goal of the ePBR material model is to create a unified model that can describe various material properties, including diffuse reflection, specular reflection, and specular transmission. This model combines diffuse reflection, specular reflection, and specular transmission, and introduces metallicity and transparency parameters to flexibly represent various materials. The formula for the ePBR material model is:

[0059] f(p,ω o ω i )=(1-t1)(1-m)f d (p,ω o ω i )+f s (p,ω o ω i )+t1(Γ t ·f t (p,ω o ω i ))

[0060] Where, f(p,ω) o ,ω i ) is the final two-way scattering distribution function (BSDF), describing the surface's reflection and transmission characteristics. t1 is the transparency parameter, representing the surface's transparency, ranging from [0,1]. When t=0, the surface is completely opaque; when t=1, the surface is completely transparent. m is the metallicity parameter, representing the surface's metallic properties, ranging from [0,1]. When m=1, the surface is a perfect conductor (metal); when m=0, the surface is a dielectric material (non-metal). f d (p,ω o ω i f is the diffuse reflection component, described by the Lambertian model. s (p,ω o ω i ) is the specular reflection component, described by the microplane model. t (p,ω o ω i Γ is the specular transmission component, described by an extended microplane model. When t = 0 and m = 0, it is an opaque medium; when t = 0 and m = 1, it is a conductor (metal); when t = 1 and m = 0, it is a transparent glass; when t = 1 and m = 1, it is a transparent metallic material, which does not exist in nature. t It is an energy compensation term used to compensate for changes in energy distribution caused by dynamic fields.

[0061]

[0062] Where k is the dielectric attenuation coefficient, It is the gradient of the spacetime variable refractive index field. ∫ is the gradient magnitude, and ds is the path length of the ray. ray • is the path integral along the path of the refracted ray.

[0063] (1) Lambertian model of diffuse reflection

[0064] The Lambertian model is a simple diffuse reflection model that assumes the surface reflects light uniformly in all directions (ignoring complex effects such as subsurface scattering and grazing incidence retroreflection). This means that the intensity of the reflected light is the same in all directions, regardless of the direction of the incident light.

[0065]

[0066] Where α represents the diffuse albedo of the surface, i.e., the proportion of light reflected from the surface, ranging from [0, 1]. π E It is a normalization factor used to ensure the energy conservation of the diffuse reflection component.

[0067] (2) Specular reflection micro-surface model

[0068] The microplane model assumes that the surface is composed of a large number of tiny planes with different orientations.

[0069] The intensity of specularly reflected light depends on the distribution and orientation of these microfacets.

[0070]

[0071] Where ω0 is the emission direction, ω i It is the incident direction, and hr is the ω in specular reflection. i The half-vector of ω0, i.e. It is the surface normal vector at point p.

[0072] D(hr) is the Normal Distribution Function (NDF), which describes the normal distribution of microfacets on a surface. Using the GGX distribution:

[0073]

[0074] Here, r is the surface roughness parameter, ranging from [0,1]. The greater the roughness, the less smooth the surface, and the more dispersed the specular reflection. π E It is the normalization factor.

[0075] F(hr,ω o The Fresnel reflection coefficient describes the change in light intensity as it is reflected from a surface, based on the surface's refractive index. It is approximated using the Schlick approximation.

[0076] F(hr,ω o )=F0+(1-F0)·(1-|ω o ·hr|) 5

[0077] Where F0 is the base value of reflectivity. η is the refractive index of the surface.

[0078] G(hr,ω o ,ω i The geometric attenuation function describes the reduction in effective reflective area due to microfacet occlusion. Using the Smith method combined with the Schlick approximation:

[0079]

[0080] k is a parameter related to surface roughness. r is the surface roughness parameter.

[0081] (3) Specular transmission

[0082] In the model of transparent thin surfaces, transmission properties are described by an extended microfacet model. The thin surface assumption simplifies light transmission calculations while maintaining a reasonable model of the transparent material. This model effectively handles the transmission properties of transparent surfaces such as glass and windows, providing a foundation for subsequent image synthesis.

[0083]

[0084] in, This is the extended normal distribution function (eNDF), used to describe the micro-facet distribution of transmitted light, and can be estimated using a joint spherical deformation strategy. F(ht,ω) o G(ht,ω) is the Fresnel transmission coefficient, describing the intensity change of light as it is transmitted through a surface. It is based on the surface's refractive index and uses the Schlick approximation. o ,ω i ) is the geometric attenuation function, which describes the reduction in effective transmission area due to microfacet occlusion, using the same Smith method and Schlick approximation as the reflection part.

[0085] In the geometric decay function G(ht,ω) o ,ω i In the diagram, ht represents the incident direction ω during specular transmission. i and the direction of emission ω o The half-vector is expressed as follows:

[0086]

[0087] Wherein, η(p,t) is the spatiotemporal variable refractive index field, that is, the surface refractive index that changes with time and space.

[0088] Traditional thin-surface BTDF models use a solid refractive index (e.g., glass η = 1.5), which cannot simulate the dynamic refraction effects of fluid media (water flow, smoke, heat waves). To solve this problem, a spatiotemporally variable refractive index field η(p,t) is introduced in screen-space rendering to achieve physically accurate dynamic fluid refraction while maintaining real-time performance.

[0089]

[0090] Where η(p, t) is the spatiotemporal variable refractive index, describing the refractive index of any point p in space at time t. η0 is the fundamental refractive index, dimensionless, with a value range of η0 > 1, such as η0 = 1.33 for water. Δη is the refractive index fluctuation amplitude, dimensionless, with a value range of [0, 0.5η0], such as Δη = 0.1 for turbulent water flow. It's a wave arrow. φ represents the three-dimensional spatial coordinates, ω is the angular frequency, t is time, and φ is the phase shift. λ1 and λ2 are the coefficients of the fluid medium term and the complex fluid term, respectively.

[0091] It is a neural network that integrates fluid physics equations with dynamic visual observations through implicit neural representations, thereby enabling the expression of non-periodic dynamic refraction of complex fluids (such as turbulence and eddies). θ It is a multilayer perceptron, which takes spatiotemporal coordinates and physical guidance characteristics as inputs and outputs a dynamic correction of the refractive index. It is a physical guiding characteristic, a guiding feature vector calculated from the fluid dynamics equations.

[0092]

[0093] in, ρ is the fluid velocity field, and ρ is the fluid density field. It is the velocity divergence. This indicates fluid expansion (such as thermal upwelling). Indicates compression (e.g., the center of a vortex). It represents the density gradient modulus, reflecting the sharpness of the fluid mixing interface (the larger the value, the sharper the interface). It represents the acceleration field, thereby capturing transient changes in fluids (such as turbulent bursts). It is a local average velocity, used to eliminate noise and reflect macroscopic flow trends. It is a temperature field.

[0094] The loss function is as follows:

[0095]

[0096] Among them, I renderRender the composite image, I gt For a real image, ||·||2 is the L2 norm. It is the residual of the refractive index transport equation, which describes the physical evolution of the refractive index in fluid motion.

[0097]

[0098] in, It is the time derivative of the refractive index. It is the fluid velocity field. It is the spatial gradient of refractive index. It is the gradient with respect to the refractive index, that is:

[0099]

[0100] Here, ||·||1 is the L1 norm.

[0101] η MLP It is the refractive index distribution predicted by the model, η prior It is a physical prior of refractive index, expressed as:

[0102]

[0103] Where ρ is density, T is temperature, and κ is... ρ It is the density-refractive index coefficient, κ T It is the temperature-refractive index coefficient. λ is the square of the L2 norm. λ3 and λ4 are weighting coefficients.

[0104] S3. Using the constructed material model and combining it with the spatiotemporal variable refractive index field, model the transparent material.

[0105] How to model a transparent thin surface, such as a window or a glass tabletop? This surface model is based on the two-way transmission distribution function (BTDF), which describes the transmission characteristics of light at points on the surface. To simplify calculations, the thin surface assumption is used, neglecting internal reflections and light deflection. Under the thin surface assumption, light transmission is mainly determined by the micro-facet distribution of the surface, similar to the treatment of specular reflection, but transmission occurs on the other side of the surface. Assumptions:

[0106] Ignore internal reflections: Assume that there is no internal reflection of light between two parallel surfaces of a transparent thin surface.

[0107] Ignore the deflection of light: Assume that the refraction effects of light when entering and leaving the transparent surface cancel each other out, so the directional deflection of the incident and outgoing light can be ignored.

[0108] Applicable scenarios: This assumption applies to real-world transparent surfaces, such as windows or glass tabletops, where the thickness of these surfaces is negligible.

[0109] S4. Based on the material model and the transparent material after modeling, perform screen space image synthesis to generate the final image.

[0110] The final image is synthesized using geometric, material, and lighting information in screen space. This method avoids complex 3D scene reconstruction, operating directly in 2D screen space, thus improving efficiency and reducing computational costs. The steps for synthesizing the final image are shown in the following equation:

[0111] I=(1-T)(1-M)Id iff +I spec +T(Γ t ·(I tran +ΔE·I sss ))

[0112] Where I represents the final composite image. T is the transparency map, representing the transparency level of each pixel. M is the metallic map, representing the metallic properties of each pixel. diff It is an image synthesized from diffuse reflection components. spec It is an image synthesized from specular reflection components. tran This is an image synthesized from the specular transmission components. Γ t It is an energy compensation term used to compensate for changes in energy distribution caused by dynamic fields.

[0113]

[0114] Where κ is the dielectric attenuation coefficient, It is the gradient of the spacetime variable refractive index field. ∫ is the gradient magnitude, and ds is the path length of the ray. ray • is the path integral along the path of the refracted ray.

[0115] Traditional thin-surface BTDF neglects internal multiple scattering, resulting in an energy underestimation of over 18% for high-roughness transparent materials (frosted glass) and a lack of internal optical paths, leading to a "rigid" visual effect. To address this issue, this embodiment proposes a Quasi-Monte Carlo Subsurface Scattering (QMC-SSS) model, namely ΔE·I sss .

[0116]

[0117] in, ω represents the surface normal vector. i It is the incident direction, f t It is the specular transmission function. ΔE is the energy loss rate due to neglecting multiple scattering.

[0118]

[0119] in, It is the spherical harmonic reconstruction operator, f sss (p,ω o ,ω i E is the scattering function. in It represents the incident irradiance, * represents the convolution operation, and I represents the incident irradiance. sss A composite image representing the subsurface scattering component is used to simulate the optical effect of light passing through a translucent material (such as skin, wax, or frosted glass), undergoing multiple scatterings within the material, and then re-emitting from the surface.

[0120]

[0121] Where, σ s Here, k is the scattering coefficient, N is the scattering order, and Φ is the maximum scattering order. k It is a k-th order phase function. ω o It is the direction of emission, ω i It is the incident direction, and p is a point on the surface.

[0122] Using the Henyey-Greenstein anisotropic phase function:

[0123]

[0124] Where θ is the scattering angle, cosθ=ω o ·ω i g k It is the anisotropy factor, with a value range of [-1, 1].

[0125] (1) Diffuse reflection I diff

[0126] I diff =A·E

[0127] Where A is the albedo map in screen space, representing the inherent color of the surface. E is the diffuse irradiance map, representing the intensity of diffuse light reaching the surface.

[0128]

[0129] It can be directly estimated using the screen-space G-buffer, and is typically used in deferred rendering.

[0130] (2) Specular reflection I spec

[0131]

[0132] Where A and B are pre-calculated coefficients related to surface roughness, and F0 is the base reflectivity. It is a normalized filter kernel based on surface roughness. A mr It is a specular reflection image, generated through screen-space ray tracing. The filter kernel K is applied to the specular reflection image A. mr The convolution operation is used to simulate the blurring effect of specular reflection. (Specular reflection image A) mr The generation steps are as follows:

[0133] ① Use screen-space ray tracing to calculate the reflected color of each pixel.

[0134] ② A depth map and a normal map are needed to calculate the reflection direction and to trace the reflected rays in screen space.

[0135] ③ If the reflected light extends beyond the screen space or hits an object behind it, image restoration technology may be needed to fill the hole.

[0136] (3) Specular Transmission I tran

[0137]

[0138] Where A and B are pre-calculated coefficients related to surface roughness, and F0 is the base reflectivity. It is a normalized filter kernel based on surface roughness. A bg This is a background radiation map, representing the background color of transmitted light. If a transparent surface is added to the scene, the background radiation map can be obtained directly; if an opaque surface is made transparent, image inpainting techniques may be needed to estimate the background content. For background radiation map A bg Two convolution operations are applied to simulate the blurring effect of transmitted light.

[0139] Example 2

[0140] This embodiment also provides a material generation system based on physically based rendering, including: a calculation module, a first construction module, a second construction module, and a compositing module; the calculation module is used to calculate the reflection and transmission characteristics of each surface point in the image rendering scene based on the rendering equation; the first construction module is used to construct a material model based on the reflection and transmission characteristics, the material model including diffuse reflection, specular reflection, and specular transmission, and introducing metallicity and transparency parameters; the second construction module is used to model transparent materials using the constructed material model and a spatiotemporally variable refractive index field; the compositing module is used to perform screen space image compositing based on the material model and the modeled transparent materials to generate the final image.

[0141] The rendering equations include:

[0142]

[0143] Where, ω i ω is the incident direction, ω0 is the exit direction, L(p, ω o L(p, ω) represents the emitted radiance along the ω0 direction at position p. i ) indicates that at position p, along ω i Incident radiance in the direction, f(p, ω) o ω i ) represents the two-way scattering distribution function. It is the surface normal vector at point p. This represents the situation on a unit sphere with respect to all possible incident directions ω. i Integrate the points.

[0144] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A material generation method based on physically based rendering, characterized in that, Includes the following steps: Based on the rendering equation, the reflection and transmission characteristics of each surface point in the image rendering scene are calculated. Based on the reflection and transmission characteristics, a material model is constructed, comprising diffuse reflection, specular reflection, and specular transmission components, and incorporating metallicity and transparency parameters; the constructed material model includes: in, It is a two-way scattering distribution function. It is a transparency parameter. It is a metallicity parameter. It is the diffuse reflection component. It is the specular reflection component. It is the specular transmission component. It is an energy compensation term; The specular transmission component includes: in, It is an extended normal distribution function. It is the Fresnel transmission coefficient. It is a geometric decay function; Represents any point in space; yes Surface normal vector at point; In geometric decay function middle, The incident direction in specular transmission and launch direction The half-vector is expressed as follows: in, It is a spacetime variable refractive index field, that is, the surface refractive index that changes with time and space; Introducing a spatiotemporally variable refractive index field in screen space rendering To achieve physically accurate dynamic fluid refraction: in, It is a spacetime variable refractive index. It is the fundamental refractive index, dimensionless, with a range of values. , It represents the amplitude of refractive index fluctuation, dimensionless, and its range of values. , It's a wave arrow. It is a three-dimensional coordinate system. It is angular frequency. It is time. It's a phase shift. and These are the coefficients of the fluid medium term and the complex fluid term; It's a neural network. It is a multilayer perceptron, which takes spatiotemporal coordinates and physical guidance characteristics as inputs and outputs a dynamic correction of refractive index. It is a physical guiding characteristic, a guiding feature vector calculated from the fluid dynamics equations; Using the constructed material model, combined with the spatiotemporal variable refractive index field, a transparent material is modeled. Based on the material model and the modeled transparent material, screen space image synthesis is performed to generate the final image.

2. The material generation method based on physically based rendering according to claim 1, characterized in that, The rendering equations include: in, It is the direction of incidence. It is the direction of launch. Indicates the location place, along The emitted radiation brightness in the direction, Indicates the location place, along The incident radiance in the direction of the radiation. Represents the two-way scattering distribution function. yes The surface normal vector at the point, This represents all incident directions on a unit sphere. Integrate the points.

3. The material generation method based on physically based rendering according to claim 1, characterized in that, Used to compensate for changes in energy distribution caused by dynamic fields: in, It is the dielectric attenuation coefficient. It is the gradient of the spacetime variable refractive index field. It is the gradient magnitude. Light rays are divided into paths. It is the path integral along the path of the refracted ray.

4. The material generation method based on physically based rendering according to claim 1, characterized in that, The steps involved in synthesizing the final image include: in, This represents the final synthesized image. It's a transparency chart. It is a metallicity map, representing the metallic properties of each pixel. It is an image synthesized from diffuse reflection components, I spec It is an image synthesized from specular reflection components. It is an image synthesized from the specular transmission components. It is an energy compensation term. It is a quasi-Monte Carlo subsurface scattering model.

5. The material generation method based on physically based rendering according to claim 4, characterized in that, The quasi-Monte Carlo subsurface scattering model includes: in, Represents the surface normal vector. It is the direction of incidence. It is the specular transmission function. This is due to the energy loss rate caused by neglecting multiple scattering. It is a spherical harmonic reconstruction operator. It is the scattering function. It is the incident irradiance. It's a convolution operation. A composite image representing the subsurface scattering components.

6. A material generation system based on physically based rendering, the system being used to implement the method according to any one of claims 1-5, characterized in that, include: The module consists of a computation module, a first building module, a second building module, and a synthesis module. The calculation module is used to calculate the reflection and transmission characteristics of each surface point in the image rendering scene based on the rendering equation; The first construction module is used to construct a material model based on the reflection characteristics and the transmission characteristics. The material model includes diffuse reflection, specular reflection and specular transmission components, and introduces metallicity and transparency parameters. The second construction module is used to model transparent materials by utilizing the constructed material model and combining it with a spatiotemporal variable refractive index field. The compositing module is used to perform screen space image compositing based on the material model and the modeled transparent material to generate the final image.

7. The material generation system based on physically based rendering according to claim 6, characterized in that, The rendering equations include: in, It is the direction of incidence. It is the direction of launch. Indicates the location place, along The emitted radiation brightness in the direction, Indicates the location place, along The incident radiance in the direction of the radiation. Represents the two-way scattering distribution function. yes The surface normal vector at the point, This represents all incident directions on a unit sphere. ωi Integrate the points.

Images