Secrets of 3D computer graphics (стр. 1 из 2)
SECRETS OF 3D COMPUTER GRAPHICS
Report: second-year graduate
Rostov-on-Don
2010
Content
Introduction
What Makes a Picture 3D
What Are 3D Graphics
How to Make It Look Like the Real Thing
Depth of Field
Realistic Examples
Making 3D Graphics Move
Fluid Motion for Us Is Hard Work for the Computer
Transforms and Processors: Work, Work, Work
How Graphics Boards Help
Introduction
You're probably reading this on the screen of a computer monitor -- a display that has two real dimensions, height and width. But when you look at a movie like "Toy Story II" or play a game like TombRaider, you see a window into a three-dimensional world. One of the truly amazing things about this window is that the world you see can be the world we live in, the world we will live in tomorrow, or a world that lives only in the minds of a movie’s or game's creators. And all of these worlds can appear on the same screen you use for writing a report or keeping track of a stock portfolio.
How does your computer trick your eyes into thinking that the flat screen extends deep into a series of rooms? How do game programmers convince you that you're seeing real characters move around in a real landscape? In this edition of How Stuff Works, we will tell you about some of the visual tricks 3D graphic designers use, and how hardware designers make the tricks happen so fast that they seem like a movie that reacts to your every move.
What Makes a Picture 3D
A picture that has or appears to have height, width and depth is three-dimensional (or 3D). A picture that has height and width but no depth is two-dimensional (or 2-D). Some pictures are 2-D on purpose. Think about the international symbols that indicate which door leads to a restroom, for example.
The symbols are designed so that you can recognize them at a glance. That’s why they use only the most basic shapes. Additional information on the symbols might try to tell you what sort of clothes the little man or woman is wearing, the color of their hair, whether they get to the gym on a regular basis, and so on, but all of that extra information would tend to make it take longer for you to get the basic information out of the symbol: which restroom is which. That's one of the basic differences between how 2-D and 3D graphics are used: 2-D graphics are good at communicating something simple, very quickly. 3D graphics tell a more complicated story, but have to carry much more information to do it.
Take a look at the triangles above. Each of the triangles on the left has three lines and three angles -- all that's needed to tell the story of a triangle. We see the image on the right as a pyramid -- a 3D structure with four triangular sides. Note that it takes five lines and six angles to tell the story of a pyramid -- nearly twice the information required to tell the story of a triangle.
For hundreds of years, artists have known some of the tricks that can make a flat, 2-D painting look like a window into the real, 3D world. You can see some of these on a photograph that you might scan and view on your computer monitor: Objects appear smaller when they're farther away; when objects close to the camera are in focus, objects farther away are fuzzy; colors tend to be less vibrant as they move farther away. When we talk about 3D graphics on computers today, though, we're not talking about still photographs -- we're talking about pictures that move.
If making a 2-D picture into a 3D image requires adding a lot of information, then the step from a 3D still picture to images that move realistically requires far more. Part of the problem is that we’ve gotten spoiled. We expect a high degree of realism in everything we see. In the mid-1970s, a game like "Pong" could impress people with its on-screen graphics. Today, we compare game screens to DVD movies, and want the games to be as smooth and detailed as what we see in the movie theater. That poses a challenge for 3D graphics on PCs, Macintoshes, and, increasingly, game consoles like the Dreamcast and the Playstation II.
What Are 3D Graphics
For many of us, games on a computer or advanced game system are the most common ways we see 3D graphics. These games, or movies made with computer-generated images, have to go through three major steps to create and present a realistic 3D scene:
1. Creating a virtual 3D world.
2. Determining what part of the world will be shown on the screen.
3. Determining how every pixel on the screen will look so that the whole image appears as realistic as possible.
Creating a Virtual 3D World
A virtual 3D world isn't the same thing as one picture of that world. This is true of our real world also. Take a very small part of the real world -- your hand and a desktop under it. Your hand has qualities that determine how it can move and how it can look. The finger joints bend toward the palm, not away from it. If you slap your hand on the desktop, the desktop doesn't splash -- it's always solid and it's always hard. Your hand can't go through the desktop. You can't prove that these things are true by looking at any single picture. But no matter how many pictures you take, you will always see that the finger joints bend only toward the palm, and the desktop is always solid, not liquid, and hard, not soft. That's because in the real world, this is the way hands are and the way they will always behave. The objects in a virtual 3D world, though, don’t exist in nature, like your hand. They are totally synthetic. The only properties they have are given to them by software. Programmers must use special tools and define a virtual 3D world with great care so that everything in it always behaves in a certain way.
What Part of the Virtual World Shows on the Screen?
At any given moment, the screen shows only a tiny part of the virtual 3D world created for a computer game. What is shown on the screen is determined by a combination of the way the world is defined, where you choose to go and which way you choose to look. No matter where you go -- forward or backward, up or down, left or right -- the virtual 3D world around you determines what you will see from that position looking in that direction. And what you see has to make sense from one scene to the next. If you're looking at an object from the same distance, regardless of direction, it should look the same height. Every object should look and move in such a way as to convince you that it always has the same mass, that it's just as hard or soft, as rigid or pliable, and so on.
Programmers who write computer games put enormous effort into defining 3D worlds so that you can wander in them without encountering anything that makes you think, “That couldn't happen in this world!" The last thing you want to see is two solid objects that can go right through each other. That’s a harsh reminder that everything you’re seeing is make-believe.
The third step involves at least as much computing as the other two steps and has to happen in real time for games and videos. We'll take a longer look at it next.
How to Make It Look Like the Real Thing
No matter how large or rich the virtual 3D world, a computer can depict (изображать на картине, рисовать) that world only by putting pixels on the 2-D screen. This section will focus on just how what you see on the screen is made to look realistic, and especially on how scenes are made to look as close as possible to what you see in the real world. First we'll look at how a single stationary object is made to look realistic. Then we'll answer the same question for an entire scene. Finally, we'll consider what a computer has to do to show full-motion scenes of realistic images moving at realistic speeds.
A number of image parts go into making an object seem real. Among the most important of these are shapes, surface textures, lighting, perspective, depth of field and anti-aliasing.
Shapes
When we look out our windows, we see scenes made up of all sorts of shapes, with straight lines and curves in many sizes and combinations. Similarly, when we look at a 3D graphical image on our computer monitor, we see images made up of a variety of shapes, although most of them are made up of straight lines. We see squares, rectangles, parallelograms, circles and rhomboids, but most of all we see triangles. However, in order to build images that look as though they have the smooth curves often found in nature, some of the shapes must be very small, and a complex image -- say, a human body -- might require thousands of these shapes to be put together into a structure called a wireframe (каркасный (проволочный) метод изображения объекта).
At this stage the structure might be recognizable as the symbol of whatever it will eventually picture, but the next major step is important: The wireframe has to be given a surface.
This illustration shows the wireframe of a hand made from relatively few polygons -- 862 total.
The outline of the wireframe can be made to look more natural and rounded, but many more polygons -- 3,444 -- are required.
Surface Textures
When we meet a surface in the real world, we can get information about it in two key ways. We can look at it, sometimes from several angles, and we can touch it to see whether it's hard or soft. In a 3D graphic image, however, we can only look at the surface to get all the information possible. All that information breaks down into three areas:
Color: What color is it? Is it the same color all over?
Texture: Does it appear to be smooth, or does it have lines, bumps, craters or some other irregularity on the surface?
Reflectance: How much light does it reflect? Are reflections of other items in the surface sharp or fuzzy?
One way to make an image look "real" is to have a wide variety of these three features across the different parts of the image. Look around you now: Your computer keyboard has a different color/texture/reflectance than your desktop, which has a different color/texture/reflectance than your arm. For realistic color, it’s important for the computer to be able to choose from millions of different colors for the pixels making up an image. Variety in texture comes both from mathematical models for surfaces ranging from frog skin to Jell-o gelatin to stored “texture maps” that are applied to surfaces. We also associate qualities that we can't see -- soft, hard, warm, cold -- with particular combinations of color, texture and reflectance. If one of them is wrong, the illusion of reality is shattered.
Adding a surface to the wireframe begins to change the image from something obviously mathematical to a picture we might recognize as a hand.
We'll take a look at lighting and perspective in the next section.
Lighting and Perspective
When you walk into a room, you turn on a light. You probably don't spend a lot of time thinking about the way the light comes from the bulb or tube and spreads around the room. But the people making 3D graphics have to think about it, because all the surfaces surrounding the wireframes have to be lit from somewhere. One technique, called ray-tracing, plots the path that imaginary light rays take as they leave the bulb, bounce off of mirrors, walls and other reflecting surfaces, and finally land on items at different intensities from varying angles. It's complicated enough when you think about the rays from a single light bulb, but most rooms have multiple light sources -- several lamps, ceiling fixtures, windows, candles and so on.
Lighting plays a key role in two effects that give the appearance of weight and solidity to objects: shading and shadows. The first, shading, takes place when the light shining on an object is stronger on one side than on the other. This shading is what makes a ball look round, high cheekbones seem striking and the folds in a blanket appear deep and soft. These differences in light intensity work with shape to reinforce the illusion that an object has depth as well as height and width. The illusion of weight comes from the second effect -- shadows.
Lighting in an image not only adds depth to the object through shading, it "anchors" objects to the ground with shadows.
Solid bodies cast shadows when a light shines on them. You can see this when you observe the shadow that a sundial or a tree casts onto a sidewalk. And because we’re used to seeing real objects and people cast shadows, seeing the shadows in a 3D image reinforces the illusion that we’re looking through a window into the real world, rather than at a screen of mathematically generated shapes.
Perspective
Perspective is one of those words that sounds technical but that actually describes a simple effect everyone has seen. If you stand on the side of a long, straight road and look into the distance, it appears as if the two sides of the road come together in a point at the horizon. Also, if trees are standing next to the road, the trees farther away will look smaller than the trees close to you. As a matter of fact, the trees will look like they are converging on the point formed by the side of the road. When all of the objects in a scene look like they will eventually converge at a single point in the distance, that's perspective. There are variations, but most 3D graphics use the "single point perspective" just described.
In the illustration, the hands are separate, but most scenes feature some items in front of, and partially blocking the view of, other items. For these scenes the software not only must calculate the relative sizes of the items but also must know which item is in front and how much of the other items it hides. The most common technique for calculating these factors is the Z-Buffer. The Z-buffer gets its name from the common label for the axis, or imaginary line, going from the screen back through the scene to the horizon. (There are two other common axes to consider: the x-axis, which measures the scene from side to side, and the y-axis, which measures the scene from top to bottom.)
The Z-buffer assigns to each polygon a number based on how close an object containing the polygon is to the front of the scene. Generally, lower numbers are assigned to items closer to the screen, and higher numbers are assigned to items closer to the horizon. For example, a 16-bit Z-buffer would assign the number -32,768 to an object rendered as close to the screen as possible and 32,767 to an object that is as far away as possible.
In the real world, our eyes can’t see objects behind others, so we don’t have the problem of figuring out what we should be seeing. But the computer faces this problem constantly and solves it in a straightforward way. As each object is created, its Z-value is compared to that of other objects that occupy the same x- and y-values. The object with the lowest z-value is fully rendered, while objects with higher z-values aren’t rendered where they intersect. The result ensures that we don’t see background items appearing through the middle of characters in the foreground. Since the z-buffer is employed before objects are fully rendered, pieces of the scene that are hidden behind characters or objects don’t have to be rendered at all. This speeds up graphics performance. Next, we'll look at the depth of field element.
Depth of Field
Another optical effect successfully used to create 3D is depth of field. Using our example of the trees beside the road, as that line of trees gets smaller, another interesting thing happens. If you look at the trees close to you, the trees farther away will appear to be out of focus. And this is especially true when you're looking at a photograph or movie of the trees. Film directors and computer animators use this depth of field effect for two purposes. The first is to reinforce the illusion of depth in the scene you're watching. It's certainly possible for the computer to make sure that every item in a scene, no matter how near or far it's supposed to be, is perfectly in focus. Since we're used to seeing the depth of field effect, though, having items in focus regardless of distance would seem foreign and would disturb the illusion of watching a scene in the real world.