CN106251393B - A kind of gradual Photon Mapping optimization method eliminated based on sample - Google Patents

Abstract

The invention discloses a progressive photon mapping optimization method based on sample elimination, which includes a ray tracing stage; a photon tracing stage; a photon elimination stage; a rendering and imaging stage: traversing a coloring point, querying the photon search radius of the coloring point and within the photon The photon in the figure uses the contribution parameters of the shading point to the screen space to calculate the contribution color of the shading point, and returns the contribution color to the screen space to form a rendered image; photon rendering iteration: narrow the photon search radius of the shading point, and repeat photon tracking From the stage to the rendering and imaging stage, until the set number of iterative renderings is reached, all the rendered images formed are accumulated and averaged to obtain the final rendered image. The invention has the characteristics of easy implementation, simple method and good blue noise property.

Description

A Progressive Photon Mapping Optimization Method Based on Sample Elimination

technical field

The invention relates to the field of graphics realistic rendering, in particular to a progressive photon mapping optimization method based on sample elimination.

Background technique

Photon mapping algorithm is a kind of global illumination algorithm, which is mainly used to achieve realistic rendering effect and achieve photorealistic rendering image quality (Photorealistic Rendering). Rendering in pursuit of real photo-level image quality is the core technology in the field of film and television braking, which is called realistic rendering. Realistic rendering can make 3D film and television animation works more realistic, showing the same authentic effect as in real life. Realistic rendering mainly includes key technologies such as global illumination algorithm and special effect simulation. The global illumination algorithm not only renders the effect of the light source on the object, but also renders the propagation effect of light energy between objects, including color overflow, caustics, and ambient occlusion.

Progressive photon mapping algorithm [Progressive photon mapping] is an optimized photon mapping algorithm, which achieves progressive rendering effects by dividing photon rendering into multiple iterations, and solves the problem of storage and deviation in photon rendering. The first step is the ray tracing stage, starting from the camera, sending rays to the scene through the screen space, tracing the intersection points of the rays and the scene, recording these intersection points and the contribution parameters to the screen space, these intersection points are also called shading points. The iterative phase of photon rendering then proceeds. Each pass of photon rendering is divided into photon emission preprocessing stage and photon collection stage.

Existing photon mapping algorithms require large-scale photon emission to ensure rendering accuracy. Both the photon tracking stage and the rendering stage require a large system execution time and storage space overhead, resulting in low computational efficiency. When choosing to emit a certain number of photons for rendering, the number of photons collected will affect the rendering effect. When the number of collected photons is small, noise occurs due to the random distribution of photons, and when the number of collected photons is large, problems such as deviations will occur at the edge of the feature.

Contents of the invention

Aiming at the noise problem in the existing photon mapping algorithm due to the random emission and rebound of photons, the invention provides a progressive photon mapping optimization method based on sample elimination. The invention can accelerate the progressive photon mapping and optimize the rendering effect.

To achieve the above object, the present invention adopts the following technical solutions:

A progressive photon mapping optimization method based on sample elimination, comprising:

Step 1, ray tracing stage: trace the intersection points of rays and the scene, record these intersection points and the contribution parameters to the screen space; where these intersection points are coloring points;

Step 2, photon tracking stage: start from the light source in the scene, emit photons and perform forward photon tracking, and store the photons on the surface of the diffuse reflection object in the scene in the photon graph, and construct a KD-tree from the photon graph;

Step 3, photon elimination stage: Calculate the elimination weight of each photon in the KD-tree, build the maximum heap according to the elimination weight, filter out the photon with the largest elimination weight, eliminate the current photon with the largest elimination weight and update the elimination weight of its neighbors, Update the maximum heap, filter and eliminate the photon with the largest weight again and perform elimination and update until the set number of eliminated photons is reached;

Step 4: Rendering and imaging stage: Traversing the coloring point, querying the photon within the photon search radius of the coloring point in the photon graph, using the contribution parameter of the coloring point to the screen space to calculate the contribution color of the coloring point, and returning the contribution color to screen space to form the rendered image;

Step 5. Photon rendering iteration: narrow down the photon search radius of the shading point, repeat steps 2 to 4 until the set number of iterative renderings is reached, accumulate all the formed rendered images and calculate the average to obtain the final rendered image.

In the process of step one of the ray tracing stage, starting from the camera, rays are emitted into the scene through the screen space.

During the step one ray tracing stage, when the ray intersects the diffuse reflection surface, the shading point and the contribution parameters to the screen space are saved.

During the first ray tracing phase, when the ray intersects the non-diffuse reflection surface, save the shading point and the contribution parameters to the screen space, and continue to trace the ray until it intersects the diffuse reflection surface.

In the second step, each light source in the scene emits a certain number of photons, uses a random algorithm to determine the outgoing direction of the photons, and traces the light rays of the photons in the scene.

In the second step, when a photon collides with the surface of a non-diffuse reflective object, it is reflected, refracted, scattered or absorbed according to the properties of the object, and at the same time saves the energy, direction, position and surface normal information of the photon to the photon graph.

In the second step, when the photon collides with the diffuse reflection surface, the photon undergoes diffuse reflection or absorption, and at the same time saves the energy, direction, position and surface normal information of the photon to the photon graph.

In said step 3, the elimination weight of a photon in the KD-tree is the sum of the weights of all neighbors of the current photon, wherein the calculation process of the weight of any neighbor of the current photon is:

Step (3.1): Calculate the ratio of the distance between any neighbor photons of the current photon to the influence parameter of 2 times the photon density;

Step (3.2): Use 1 to subtract the ratio obtained in step (3.1), and finally calculate the weight of any neighbor of the current photon.

The weight calculation formula of all neighbors of the current photon is:

Among them, w _ij is the weight between the current photon i and j neighbor photons, j is an integer greater than or equal to 1; d _ij is the distance between the jth neighbor photon of the current photon i; r _max,i is the photon density The influence parameter of ; A _i is the search radius of the nearest neighbor; N _i is the number of the nearest neighbor, which is an integer greater than or equal to 1.

In the step 4, the contribution parameter of the shading point to the screen space includes the outgoing irradiance of the shading point, and the calculation formula of the outgoing irradiance of the shading point is:

r _t =α*r _t-1

in, Indicates the outgoing irradiance calculated at the position x of the shading point; w is the incident ray direction of the shading point x; k is the number of photons within the search radius, and k is an integer greater than or equal to 1; w _i is the i-th photon among the k photons φ _i is the energy of the ith photon; f _r (x,w,w _i ) is the bidirectional reflectance distribution function, that is, the proportion between the incident irradiance and the outgoing irradiance; R is the photon elimination Ratio, that is, the remaining number after photon elimination is divided by the number of emitted photons; r _t is the radius range of photons near the search shading point; t is the current iterative rendering times, and t is an integer greater than or equal to 1; α is the control parameter, which is constant.

The beneficial effects of the present invention are:

(1) The present invention proposes a progressive photon mapping optimization algorithm based on sample elimination, and applies the sample processing method to the photon mapping method, so that the photon distribution can have a certain blue noise attribute while preserving randomness, so that the resulting rendering The noise in the picture is reduced; due to the randomness of the photon emission process, many uneven phenomena appear in the photon distribution. Between the two steps of photon emission and photon rendering, the present invention adds a step of photon elimination, by calculating the relative The weight of distance, eliminate the photon closest to other photons, and choose an optimal elimination ratio, so that the photon map used in each photon calculation in the progressive photon mapping method has blue noise characteristics, so that the photon used in each step The distribution is more optimized, which speeds up the progressive photon mapping and optimizes the rendering effect.

(2) The optimized photon sample processing method of the present invention controls the elimination weight of photons according to the density of photon distribution to ensure the brightness of the global illumination. At the same time, the global photon elimination enables the photon mapping algorithm based on the sample method to be applied to all renderings of the global illumination In the calculation of the effect; the sample method based on elimination of sampling in the present invention can be closely combined with the progressive photon mapping in the process, so that the photon mapping optimization method based on the sample method can be extended to the progressive photon mapping method; by optimizing each step The iterative photon distribution speeds up the convergence and can achieve the effect of reducing the number of iterations by half.

Description of drawings

FIG. 1 is a flow chart of the progressive photon mapping optimization method based on sample elimination in the present invention.

Fig. 2(a) is the rendering rendering of the Cornell Box scene rendered 64 times using the progressive photon mapping algorithm.

Fig. 2(b) is a rendering effect diagram of rendering a Cornell Box scene 32 times using the present invention.

Fig. 2(c) is a rendering effect diagram of 64 passes of rendering a Cornell Box scene using the present invention.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention.

The method of sample elimination enables the production of Poisson sample geometries with stronger blue noise properties. The method of sample elimination calculates a weight between every two sample points according to the distance between the sample points. The closer the sample points are, the greater the weight is. The elimination weight of a sample point is the distance between him and the surrounding neighbors. sum of weights. The method adopts a greedy strategy to eliminate the sample point with the largest weight from some randomly generated samples each time. And update the elimination weights of the neighbors around this sample point. Finally, a fixed number of sample point sets with blue noise properties can be obtained. As a new sample processing method, the sample elimination method has the characteristics of easy implementation, simple method, and good blue noise properties. The photon set in the photon mapping method is regarded as a sample point set for sample processing, and the processed photon set can remove the random noise in the distribution to a certain extent.

For a rendering instance process, it is the process of calculating a modeled 3D scene into a picture. Before the rendering task starts, the 3D scene data is converted into an expression that the rendering engine can recognize. A scene data file package containing complete information contains information such as cameras, geometry, light sources, materials, textures, etc. that can be recognized by the rendering engine. Rendering starts after the scene is ready.

The Cornell Box scene (referred to as the Box scene) is a simple example of rendering global illumination effects. In this scene, a point light source is placed on the top of the box, and two small boxes are placed in the middle of the box. Set the rendering variable to 64 passes, set the elimination ratio to 30%, and the pixel sampling rate to 4.

Fig. 1 is a flowchart of the progressive photon mapping optimization method based on sample elimination of the present invention. As shown in the figure, after the rendering is started, the specific steps are as follows:

Step 1. The ray tracing stage; start from the camera, emit light rays into the scene through the screen space, trace the intersection points of the light rays and the scene, and record these intersection points and the contribution parameters to the screen space.

Step 1 is as follows:

From the camera directly in front of the box, emit light into the scene, and determine the direction of the light through the connection between the camera and the screen space. Since one pixel samples 4 sampling points, 4 light rays are emitted through one pixel, and these rays are traced. Objects in the scene are all It is a diffuse reflection surface, which saves this intersection point at the first intersection point of the light and the scene, and saves the contribution parameters of the surface to the screen space. These intersection points are also called shading points.

Step 2. Photon tracking stage: track photons by emitting a certain number of photons from the light source and performing forward photon tracking, and store photons in the photon graph on the surface of the diffuse reflection object to construct a KD-tree of photons.

details as follows:

Send 20,000 photons into the scene from the light source above the box, then generate a random direction for each photon, and track the photon's trajectory in the scene. Save the energy, position, direction, and plane discovery information of the photons that collided with the plane into the photon graph. The photons are absorbed or diffusely reflected when they collide with the surface of the object. The diffusely reflected photons continue to be tracked in the scene. The maximum number of bounces is used to end the photon tracing process and complete the generation of the photon map.

Step 3. Photon elimination stage; calculate the elimination weight of each photon, construct the maximum heap according to the elimination weight, select the photon with the largest elimination weight, eliminate this photon and update the elimination weight of the photon’s neighbors, update the maximum heap, and select the largest elimination weight again The photons are eliminated and updated, and this process is repeated until the set photon elimination number is reached.

details as follows:

Step (3.1): Traverse each photon, search for the nearest 10 neighbor photons around the photon, and calculate the weight of each neighbor photon and the current photon according to the formula (1), and the elimination weight of the current photon is the sum of these neighbor weights ;

The weight calculation formula of all neighbors of the current photon is:

Among them, w _ij is the weight between the current photon i and j neighbor photons, j is an integer greater than or equal to 1; d _ij is the distance between the jth neighbor photon of the current photon i; r _max,i is the photon density The influence parameter of ; A _i is the search radius of the nearest neighbor; N _i is the number of the nearest neighbor, which is an integer greater than or equal to 1.

In the present invention, a low-sampling tree is used to save r _max,i , and the r _max,i closest to him is selected according to the position of the current photon each time; the initial value of r _max,i can be according to the first formula (1) Calculated by the two formulas, A _i is the search radius of the nearest neighbor, N _i is the number of the nearest neighbor, the calculated r′ _{max is directly used in the first iteration, and i} is r _{max, i} ; in the remaining iterations, according to The r _{max, i} value in the queried low-sampling tree and the currently calculated r′ _{max, the average of the value of i} are used as r _{max, i} , and the data value in the low-sampling tree is updated after calculation;

Step (3.2): According to the elimination weights of all photons, a maximum heap is established;

Step (3.3): Select the photon with the largest elimination weight;

Step (3.4): Updating the elimination weight of the neighbor photon of the elimination photon selected in step (3.4), the method is to subtract the weight of the neighbor photon of the elimination photon from the elimination weight:

Step (3.5): update the maximum heap, repeat step (3.3), step (3.4) and step (3.5) until the set elimination ratio of 30% is reached, that is, 6000 photons are eliminated.

Step 4. Rendering imaging stage; traverse all the shading points recorded in step 1, search the photon graph for each shading point and find the photons near the shading point with a radius of r _t to estimate the illumination effect of the light source, including the direct All photons of illumination and indirect illumination, t is the current iteration number, and by controlling the parameters, each iteration reduces the search radius to eliminate the deviation. Finally, the color is calculated according to the contribution parameters of the recorded shading points and returned to the screen space for accumulation.

details as follows:

Step (4.1): traverse each shading point, and at a shading point, search for photons within the radius r _t near the shading point; t is the current iterative rendering times, and t is an integer greater than or equal to 1; α is The control parameter is a constant, and the value is 0.9 during the running of the example;

r _t =α*r _t-1 formula (2)

Step (4.2): Calculate the outgoing irradiance of the current coloring point according to the formula (3) of the outgoing radiance;

in, Indicates the outgoing irradiance calculated at the position x of the shading point; w is the incident ray direction of the shading point x; k is the number of photons within the search radius, and k is an integer greater than or equal to 1; w _i is the i-th photon among the k photons φ _i is the energy of the ith photon; f _r (x,w,w _i ) is the bidirectional reflectance distribution function, that is, the proportion between the incident irradiance and the outgoing irradiance; R is the photon elimination Ratio, which is the number remaining after photon elimination divided by the number of photons emitted.

Step (4.3): Use the contribution information of the shading point to the screen space to calculate the contribution color of the shading point, return the color to the screen space, and accumulate and average the image of the previous iteration to form an image, which is displayed in the rendering result window.

Step 5. Photon rendering iterations, repeat steps 2 to 5 until the set number of iterative renderings reaches 64 times, and the rendering is terminated. Figure 2(c) is the result obtained by using this method to render 64 times.

Fig. 2(a) is a rendering rendering of the Cornell Box scene rendered 64 times using the progressive photon mapping algorithm; Fig. 2(b) is a rendering rendering rendering of the Cornell Box scene 32 times using the method of this patent. By comparing Fig. 2(a) to Fig. 2(c), it can be known that the image obtained by the present invention is clearer and the rendering effect is better.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented through computer programs to instruct related hardware, and the programs can be stored in a computer-readable storage medium. During execution, it may include the processes of the embodiments of the above-mentioned methods. Wherein, the storage medium may be a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM).

Although the specific implementation of the present invention has been described above in conjunction with the accompanying drawings, it does not limit the protection scope of the present invention. Those skilled in the art should understand that on the basis of the technical solution of the present invention, those skilled in the art do not need to pay creative work Various modifications or variations that can be made are still within the protection scope of the present invention.

Claims (8)

1. A progressive photon mapping optimization method based on sample elimination, characterized in that, comprising: Step 1, ray tracing stage: trace the intersection points of rays and the scene, record these intersection points and the contribution parameters to the screen space; where these intersection points are coloring points; Step 2, photon tracking stage: start from the light source in the scene, emit photons and perform forward photon tracking, and store the photons on the surface of the diffuse reflection object in the scene in the photon graph, and construct a KD-tree from the photon graph; Step 3, photon elimination stage: Calculate the elimination weight of each photon in the KD-tree, build the maximum heap according to the elimination weight, filter out the photon with the largest elimination weight, eliminate the current photon with the largest elimination weight and update the elimination weight of its neighbors, Update the maximum heap, filter and eliminate the photon with the largest weight again and perform elimination and update until the set number of eliminated photons is reached; Step 4: Rendering and imaging stage: Traversing the coloring point, querying the photon within the photon search radius of the coloring point in the photon graph, using the contribution parameter of the coloring point to the screen space to calculate the contribution color of the coloring point, and returning the contribution color to screen space to form the rendered image; Step 5. Photon rendering iteration: narrow down the photon search radius of the shading point, repeat steps 2 to 4 until the set number of iterative renderings is reached, accumulate all the formed rendered images and average them to obtain the final rendered image; In the fourth step, the contribution parameter of the shading point to the screen space includes the outgoing irradiance of the shading point, and the calculation formula of the outgoing irradiance of the shading point is: r _t =α*r _t-1 in, Indicates the outgoing irradiance calculated at the position x of the shading point; w is the incident ray direction of the shading point x; k is the number of photons within the search radius, and k is an integer greater than or equal to 1; w _i is the i-th photon among the k photons φ _i is the energy of the ith photon; f _r (x, w, w _i ) is the bidirectional reflectance distribution function, that is, the proportion between the incident irradiance and the outgoing irradiance; R is the photon elimination Ratio, that is, the remaining number after photon elimination is divided by the number of emitted photons; r _t is the radius range of photons near the search shading point; t is the current iterative rendering times, and t is an integer greater than or equal to 1; α is the control parameter, which is constant. 2. A progressive photon mapping optimization method based on sample elimination as claimed in claim 1, characterized in that, in said step one, during the ray tracing phase, starting from the camera, the light is emitted into the scene through the screen space. 3. a kind of progressive photon mapping optimization method based on sample elimination as claimed in claim 1, is characterized in that, in described step one ray tracing stage process, when light intersects with diffuse reflective surface, save coloring point and pair Contribution parameter for screen space. 4. a kind of progressive photon mapping optimization method based on sample elimination as claimed in claim 1, is characterized in that, in described step one ray tracing stage process, when light intersects with non-diffuse reflective surface, save coloring point and Contribution parameter to screen space, and continue tracing rays until they intersect the diffuse surface. 5. a kind of progressive photon mapping optimization method based on sample elimination as claimed in claim 1, is characterized in that, in described step 2, each light source in the scene emits a certain number of photons, utilizes random algorithm to determine photon Outgoing direction, and trace photon rays in the scene. 6. A progressive photon mapping optimization method based on sample elimination as claimed in claim 1, wherein in said step 2, when the photon hits the surface of a non-diffuse reflective object, it is reflected according to the properties of the object , refraction, scattering or absorption, while saving photon energy, direction, position and surface normal information into the photon map. 7. A kind of progressive photon mapping optimization method based on sample elimination as claimed in claim 1, characterized in that, in said step 2, when the photon hits the diffuse reflection surface, the photon is diffusely reflected or absorbed, and the photon is saved at the same time The energy, direction, position, and surface normal information of the photon map. 8. A kind of progressive photon mapping optimization method based on sample elimination as claimed in claim 1, wherein in said step 3, the elimination weight of a photon in the KD-tree is the sum of the weights of all neighbors of the current photon ; Among them, the calculation process of the weight of any neighbor of the current photon is: Step (3.1): Calculate the ratio of the distance between any neighbor photons of the current photon to the influence parameter of 2 times the photon density; Step (3.2): Use 1 to subtract the ratio obtained in step (3.1), and finally calculate the weight of any neighbor of the current photon.