KR20070119853A - Real-time three-dimensional painting image rendering using octree texture - Google Patents

Abstract

A real time three-dimensional painting image rendering using octree-texture is provided to render a three-dimensional painting image by using three-dimensional parameter information. An octree-texture of a mesh inputted from a CPU(Central Processing Unit) is generated and two-dimensional texture coordinates are indexed to generate a neighbor texture and a density texture. The lowermost node of the octree-texture is searched from a GPU. Water and dye density information of the corresponding node and a neighbor node are obtained by using index information of the lowermost node.

Description

Real-Time Three-Dimensional Painting Image Rendering Using Octree Texture}

The present invention relates to a real-time three-dimensional painting image rendering method using an octree texture, specifically, by storing the three-dimensional surface information of the mesh as an octree structure in the volume texture, by performing a particle simulation on the three-dimensional surface using a GPU It is about the technique of rendering painting effects in real time.

The method of simulating the interaction of water, paint, paper, etc. to create natural painting images that seem to be drawn by humans is one of the major research areas of non-realistic rendering. The typical painting simulation method simulates the effect of water and paint concentrations, different brush touch methods, and different color paints, rendering a painting image that looks like a human. However, simulating complex and physical interactions requires a great deal of computation, which places many constraints on real-time applications. In addition, the existing methods deal only with two-dimensional simulations occurring on paper, so new techniques are needed in the three-dimensional implementation.

The technical problem to be solved by the present invention is to solve the above-mentioned problems, it is to create a painting image in real time by searching the octree texture using a GPU, and performing a simulation using a pixel shader.

In addition, by using the three-dimensional parameter information is to render a three-dimensional painting image, not a two-dimensional painting image.

Real-time painting image rendering method according to a feature of the present invention for achieving the above object

A preprocessing step of generating an octree texture of a mesh input as an input from a CPU and indexing two-dimensional texture coordinates to generate neighboring textures and density textures; Searching at the GPU for the lowest node of the octree texture comprising a three-dimensional surface of a mesh; Obtaining water and paint concentration information of a corresponding node and a neighboring node using index information of the found node; And applying a simulation equation using the above information.

Hereinafter, embodiments of the present invention will be described in detail.

An embodiment of the present invention proposes a method of rendering an input mesh as a 3D painting image by using an octree texture and particle simulation. This method has the advantage of being able to render in real time because most of the process is performed on the GPU, and use three-dimensional elements as simulation parameters.

In the first step, we first create a cube bounding volume of the input 3D mesh and divide the lower node using the volume as the top node. In order to determine whether to split the node, we check whether the polygon exists in the volume of the node using the existing efficient hexahedron and polygon cross test method.

Create a tree up to the depth you set and store it in a 3D texture. This octree texture has an 8-bit RGBA channel and the data stored is different for each node type.

If it is the lowest node and has a polygon in the volume, two-dimensional texture coordinates are stored in two channels of RGB. Two-dimensional texture coordinates are assigned in order from (0,0). And store 1 in the alpha channel. An intermediate node with eight child nodes stores the three-dimensional coordinates of eight child nodes in the octree texture in the RGB channel. And store 0.5 in the alpha channel. Finally, empty nodes that have nothing store zeros on all RGBA channels.

The neighbor texture generates a total of 26 sheets, and stores the RGBA values of adjacent nodes in three directions of X, Y, and Z around one node in the texture. The concentration texture stores the initial concentrations of water and paint of the node.

The first steps so far are all preprocessing done by the CPU.

The second step is to search the octree texture to get the surface information of the mesh to render. This is done through the pixel shader and searches the tree structure stored in the texture memory to access density information of the surface or neighbor node information.

Let's say that the three-dimensional texture coordinate we are looking for is M and the depth of the node we are visiting is D. Index 2 * 2 * 2 area at depth D I _D Speak I ₀ Is (0,0,0).

Therefore, the formula for calculating the coordinates of M in the current visit I _D is

In Equation 1, the type of the current node may be determined based on an alpha value stored in P.

If the alpha value is 0, it will not proceed any further because it is an empty node. If the alpha value is 1, it is the lowest node that contains polygons, so 2D texture coordinates are obtained through the R, G, and B channels. If the alpha value is 0.5, the R, G, and B values are I _{D + 1} because they are intermediate nodes with child nodes. Therefore, proceed to the next depth.

All polygons belong to the lowest node that contains the polygon, so in our example we never refer to an empty node. Therefore, the loop is accessed as deep as the tree to access the lowest node.

In the next step, water and paint concentration information of the node and neighbor nodes is obtained using the index information of the node obtained in the previous step. Two of the RGB channels of the lowest node, including polygons, contain U and V texture coordinates. By accessing the concentration texture using the coordinates, we can get the water and paint concentration information of the node.

In order to simulate water and paint advection, we need information from neighboring nodes. The neighboring texture is accessed using the U and V coordinates obtained in the above step to obtain index information of the neighbor node. When the density texture is accessed again using the U and V coordinates obtained from the neighbor texture, the concentration information of water and paint of the neighbor node can be obtained.

The final step is to perform the particle simulation based on the information obtained from the octree texture, neighbor texture, and density texture. Particle simulation is done by Lattice Boltzmann Equation. The equation is

Parameters can be applied according to the user's choice. The movement properties of the particles can be set using values such as brightness, distance from the camera, and normals according to the light direction.

As described above, it is possible to render a 3D painting image in real time by the method of the present invention.

Claims (2)

Real-time non-realistic rendering technique that renders 3D painting images in real time How to apply 3D spatial information parameter to Lattice Boltzmann Equation