Computer graphics three kinds of rendering (drawing) technology, rendering technology

In computer graphics, there are many ways to represent a real three-dimensional object, such as point cloud, mesh, voxel, curve, etc. this also affects the selection of the drawing primitives when drawing three-dimensional objects, the following is my summary of some of the classification of the drawing technology, I hope to be helpful to everyone.

research on real-sense graphics algorithm render three elements of a real image:

1. Precise Geometrical model

2. Fine Texture Mapping

3. Reasonable and realistic illumination and shading

The early illumination model of local illumination model is based on experience, which can only reflect the direct illumination of the light source, and the illumination model is a local illumination model.

factors to be considered in the local illumination model: ambient light, diffuse reflection, high-light and specular reflection, light attenuation,

Produce color (RGB), multi-source, hard shadow, soft shadow, Volume Shadow

Shadows are divided into the dynamic shadow algorithm and the Static Shadow algorithm (the difference is whether the light is moving)

Shadow Generation Technology , currently commonly used in three kinds: Planar Shadow, Shadow Mapping and Shadow Volume, the former similar to the projection, the simplest calculation, the disadvantage can only be plotted on the plane of the shadow; Shadow Mapping use the station in the light source to see the generated depth map along the source normal to detect whether the scene of the body pixels in the shadow, the disadvantage is that the light source and object position is relatively fixed, and in extreme cases, the accuracy is poor, not suitable for accurate to the pixel dynamic time occasions; Shadow Volume is currently the most suitable for the precise performance of dynamic time scene technology, the most applicable, its typical application example is Doom 3, the lack of shadow volume introduced additional vertices and polygons, increased storage and processing strength, while rendering the shadow is more hard, if you want to achieve soft shadows, still need other technical cooperation

The global illumination Model is accurate, which simulates the interaction of illumination between objects in order to achieve satisfying illumination effect.

the composition of the incident light of the global illumination Model: direct illumination of the light source, reflected light of other objects, transmitted light.

Texture Map texture mapping refers to the process of assigning a texture or data to a 3D object, the principle of which is to change the parameter values in the light equation to achieve the purpose of changing the appearance of the object.

generic texture pipeline for a single texture:
Common projection functions are: spherical projection, cylindrical projection, plane projection, natural projection

method of texture mapping:

L Alpha Map: To achieve transparent, translucent effects of the image

L Light Map

L Gloss Map: Used to simulate uneven surface gloss of objects

L Environment Map:

L Bump Mapping: The basic idea is to change the color component in the light equation by changing the surface normals rather than using textures.

Process texture is mainly used to simulate the common marble,stone,wood,cloud and other textures in nature, and most process textures are based on some kind of noise function.

3D texture is no longer a two-dimensional planar image, it is a volumetric texture, can be seen as a variety of different cube materials, as dozens of, hundreds of layers of 2D texture overlap, each layer is a simple 2D texture. Each point in a 3D texture has a three-dimensional relative coordinate within the texture's interior space.

Multi-pass texture mapping: In computer graphics theory, all light equation factors need to be calculated quickly, and a color sample is generated. In fact, the illumination factor can be calculated in different pathways, each of which modifies the previous result. image-based rendering technology (IMAGE-BASERENDERING,IBR)

For real-time applications, the greatest benefit of using an image to represent an object is to draw two proportional to the number of pixels to be drawn, regardless of the model itself. advantages and disadvantages of this technology:

1. Based on the reality image, the virtual environment is high fidelity.

2. The processing time of the scene is independent of the complexity of the scene, and it is easy to realize real-time interaction;

3. Lack of a unified space coordinate system;

4. It is difficult to realize space roaming and interaction in any way.

Specific technology:

Bulletin Board: Many special effects can be drawn on the surface of the polygon toward the point of view, this direction to determine the direction of the polygon technology is called Bulletin Board Technology, wherein the polygon is a bulletin board.

According to the different plane normal direction, the bulletin board can be divided into several categories:

Ø Notice board facing the view plane:

Ø notice board facing the viewpoint;

Ø axisymmetric Bulletin board.

Genie : Is the simplest IBR unit, is an image that can be moved on the screen; If you add a depth value to the sprite texture, it is called the depth sprite.

Alternative Technology : In fact, it is also a bulletin board technology.

anti-aliasing technology : distortion of the discontinuous state of the graphics boundary becomes aliasing (jagged), and anti-aliasing is to avoid this distortion.

GPU Bottleneck Detection:

Ø Grating operation Detection

The terminal of a pipeline is raster operation, responsible for reading and writing of depth and stencil caches, comparison and reading and writing of color caches, and alpha blending and testing. Many of the loads in rasterization increase the burden of frame buffer bandwidth. During testing, it is possible to detect whether the frame rate is affected by changing the bit depth of the color buffer, or the depth of the bit deep buffer, or both. Ø Slice coloring Stage The element shading cost can be understood as the overhead of pixel, which is a resolution function that can detect whether the chip coloring caused a performance bottleneck by changing the resolution

Volume Rendering technology (volume rendering)

Visual objects generally refers to spatially dispersed 3D data 3D space of discrete data between the connection relationship can be divided into: structured data: logically organized into three-dimensional array of spatial discrete data, namely, these spatial discrete data elements have a logical relationship between the three-dimensional array, Each element can have its own layer number, line number, and column number.

[1] Homogeneous structured data: In X, Y, z three directions, the distance between the grid points are equal.

[2] Regular grid structured data: Distributed on a 3D grid of cuboid, that is, in X, Y, z three directions, the distance between the grid points is not equal, but in the same direction equal.

[3] Matrix mesh structured data: distributed in a rectangular 3D grid point, the size of the box can be different, no rules to follow.

[4] Irregular grid structured data: Logically still organized into an array of arrays, but in the spatial position of irregular can follow. unstructured data: Spatial data is composed of a series of elements (tetrahedron, six polygons, curved surfaces, curves, etc.), but cannot be organized into an array of dimensions. mixed data. visualization techniques can be divided into: polygon drawing (surface rendering) and body drawing (Volume rendering), point drawing (points rendering)

A) the algorithm of polygon drawing is cubic block method (Cuberille), moving cube method (marching Cubes) and Decomposition Cube method (Diving Cubes).

Basic idea: The body data is divided into several parallel two-dimensional slicing data and processed separately.

Polygon drawing after extracting the parts of interest in an equivalent plane, it is possible to generate high-quality images using photorealistic techniques, allowing researchers to easily observe and analyze them. The method can only display the contour of one value at a time, and it needs to be represented by geometric entities, which cannot reflect the interrelation of various factors in the whole data field. Polygon drawing does not preserve some of the physical properties of the original body data, discarding a lot of useful information inside the object.

b) Body mapping does not require the extraction of geometric elements to restore the surface of the body data, but rather directly from the 3D body data generated on-screen 2D images. The algorithm can produce the whole image of 3D body data, including every detail, and has the advantages of high quality and convenient parallel processing.

Body drawing converts the body data field of a 3D space directly into a 2D image without generating intermediate geometry elements. The discrete sampling point in the Body data field is originally not a color attribute, its color value is artificially endowed on the basis of the material classification, therefore it is pseudo-color.

The function of the body drawing to be realized: on the basis of resampling, the contribution of all sampling points to the screen pixels is computed, that is, the light intensity I (intensity) of each pixel. In the color image, the R, G, b Three components of the I values need to be calculated separately, in order to achieve this function, the need to give the optical model of the body data, used to describe how the 3D data is generated, reflected, blocked and scattered light, so as to calculate the total number of points to the screen pixel contribution

c) The typical algorithm for volume rendering:

1. Direct Volume rendering algorithm: Ray casting volume rendering algorithm, texture mapping volume rendering algorithm, throwing snowball algorithm, error cutting-variable shape rendering algorithm.

2. Transform domain Volume rendering: Fourier volume rendering algorithm, compression domain volume rendering, wavelet domain volume rendering, direct volume rendering algorithm is directly sampled in the spatial domain, the transformation domain first transforms the body data from the airspace to another domain, such as the compressed domain, the frequency domain or the small wave domain, then carries on the sampling mixing directly in these transformed domains.


Point drawing: point as the basic element of the drawing, by appropriate sampling, the scene object is represented as a set of dense surface sampling points, with these points are also stored with a bit of color, depth and surface normal vector information. Based on the information of these points, the surface reconstruction is made to produce a realistic graph composed of continuous surfaces rather than discrete point clouds.

Work Flow:

Model---Point sampling---sample point representation, storage-----invisible point culling----visible point continuation----illumination, texture, shadow----surface reconstruction----Graphics

Main algorithm:

² Point Rendering algorithm for image space: Surfel (surface Element), surface footprint Method (surface splatting)

² Point Rendering algorithm for physical space: Qsplat uses a tree-like hierarchy to enclose the data structure of the ball, each node in the tree contains: the position and radius of the ball, the normal vector at each point, the width of the normal cone, and the color value.