In the dawn of the Computer Era, bored scientists and creative owners of home computers discovered the true joy and meaning of their computer equipment: to kill, maim and wreak havoc in a virtual world. And they saw that it was good. After a while, typing different commands and reading descriptions on a text display got boring, and they created the world of two-dimensional pictorial slaughter. And they were satisfied once more. For a while. Even killing, maiming and wreaking havoc in the astonishing wonders of their flat twodimensional world eventually became dull, and they sought for something new. The salvation came when a wise man spoke: “Thou shall add a third dimension of guts and gore!” And they saw that the graphics looked… pretty awful, although it had potential. This is where this story begins…
3D Graphics in Computer Games
Connecting the dots
Today, 3D-graphics is often implied when you talk about computer graphics in games, but this has certainly not always been the case. In the mid eighties, the vast majority of the games where either 2D or text based, although this was when the first 3D-games saw the light of day. One of the first games was the space war simulation Starglider (1986) which is showing the mere basics of a 3D world. All space ships and objects in the world were built by dots or vertices as they are called in computer graphics. The dots were connected by lines, which made up a wire frame. These games were very simplistic and required some serious imagination from the user.
The space war simulation Starglider from 1986 was one of the first games using 3D graphics.
Paint by Numbers
The next natural step to take was to patch up all holes in the wire frame models, creating painted faces, polygons, between the lines. This is shown in games like the flight simulator F-19 Stealth Fighter (1988).
2 Steps forward, 1 step back
F-19 Stealth Fighter from 1988.
According to pretty intuitional basic laws of physics, a surface gets darker the further away from a light source it is. The surface also changes in light intensity depending on its inclination angle against the light. These basic laws, along with lots of more complicated ones, make up the world as we see it, and help us understand what we see. This was a very important factor to include in the visualization of the polygon, to make an illusion of something remotely realistic. Therefore, virtual light sources were introduced, and the polygons were set to acquire their colour depending on their distance and orientation in relation these lights. This calculation is known as a local lighting model.
Surfaces with only one solid colour aren’t very realistic however. One of the first games to introduce images, textures, on the surfaces was the now considered classic Wolfenstein (1992). This game was one of the first in a now very populated genre called First Person Shooters, which meant that classic run-around-and-shooteverything-games took a step closer to the flight simulators with the viewpoint “inside the head” of the character in the game instead of a side or top-down view. This game has textures on the walls instead of solid colours, but is not a true 3D-game. The walls are not made up of polygons as one might think. Instead, this game uses a technique called raycasting. The idea of raycasting is to visualize a 2D-map of the walls in a way that it looks like it is a 3D-world. Using this
Wolfenstein 3D from 1992.
method it is for example impossible to look up or down in the virtual world, or move the viewpoint up or down. This is why there are no height differences in Wolfenstein, and why the floor is completely flat. (Later refinements of this method, used in for example Duke Nukem 3D (1996) and Blood (1997), allowed for
these kind of movements too, although not to the extent of “real” 3D.) The objects in the world are not 3D either, instead they are simply ordinary 2D-images. This is very obvious if you move around an object and look at it from different directions - the object looks exactly the same! This “fake” 3D-world can be considered a step backward in the generality of 3D graphics, but it’s very fast to calculate however, which made it very popular in the early nineties.
those days wasn’t enough to handle as many voxels as one would like. Relatively large voxels had to be used which resulted in a rather blocky and coarse image.
Adding stickers
Even though there were experiments with different kinds of techniques, none other than the polygon based 3D really gained foothold of the game developers. Therefore the next step was an improvement of
Playing with building blocks
There were also other attempts at creating a more detailed environment. The helicopter simulator Comanche (1993) used a technique based on voxels to draw a landscape. A voxel is a small cube with some attributes like colour assigned to it. These small building blocks are then
Strike Commander from 1993.
Comanche from 1993.
this method. The flight simulator Strike Commander (1993) was one of the first games using texture mapped polygons. This means a 2D-image is “painted” on the polygon, giving it more detail without having to add more polygons with different colours.
Let there be light
used to build a bigger whole. A computer model built with voxels is pretty much equivalent to a real model built with the world’s most boring LEGO - there’s only one type of piece. These blocks were then visualized using highly optimized algorithms. The technique of using voxels gave a pretty detailed landscape even if the processing power of the computers
Quake (1996) added even more realism to the polygons by adding lightmaps. Lightmaps are ordinary textures, but instead of specifying what colours the polygon has across its surface, it contains how much light falls upon different points of the polygon. These lightmaps was precalculated using a much more advanced technique than could be done in real-time, and stored as part of the game data.
Quake from 1996.
monochromatic lightmaps. The hardware also allowed for advanced effects like alpha blending, which means partially transparent polygons, and texture filtering, which is a method for smoothing the textures as they are applied to the polygons to avoid coarseness due to low texture resolution. Some later games, like Unreal (1998), implemented these kinds of effects in the software renderer too, although none really achieved the quality of the output of the hardware.
The technique is called radiosity, and it is a very realistic way to calculate diffuse reflection of light between objects and light sources. Radiosity can calculate effects like this: Imagine a room with white walls and a big red carpet on the floor. Diffuse reflection from the red carpet will hit the walls and tint them a bit red too. This added, at the time, a stunning realism to the environment. Although, this technique could only be used for static objects, since characters and other moving objects would need to have its lighting recalculated as they were moved. Therefore the old method was used for these kinds of objects.
Quake II from 1997.
Playing with Building Blocks Part II
More expensive toys
Games like Quake 2 (1997) refined the lightmap method by adding coloured lightmaps. These are, as one might presume, lightmaps with colour information. This means coloured light sources could be taken in account. This was possible due to the entry of graphics hardware specialized for drawing 3D-polygons. Quake 2 could still be played without this kind of hardware, but then only with
Now the polygon method was pretty much standard in 3D games, but there were still attempts at finding alternatives. The game Delta Force (1998), by the creators of previously mentioned Comanche, offered a refined version of the voxel landscape which had both positive and negative sides. The coarseness was still there, although not as apparent. On the other hand, the viewing distance was almost infinite, something that was, and still is, very hard to accomplish with polygons. Polygonal approaches often uses fog and other tricks to restrict the
Delta Force from 1997
mixing these textured polygons together. Still this is enough to be able to do all kinds of effects more or less realistic. The latest addition to the repertoire of effects in games is different kinds of non-rigid gaseous bodies. For example, the follow-up to the legendary Wolfenstein called Return to Castle Wolfenstein (2001) implements a good looking flame thrower, and the very realistic World War II flight simulator IL-2 Sturmovik (2001) offers some beautiful clouds. It is still just polygons, but different blending tricks allows for these kinds of effects.
distance the viewer can see, but this was not needed with the voxel landscape. Though, due to the coarseness of the output, this technique never really got popular amongst the masses. A sequel, Delta Force 2 (1999), was the last game in the Delta Force series to use this method and the next game, Delta Force: Land Warrior (2000) used the traditional polygon approach.
Return to Castle Wolfenstein from 2001.
Playing with Fire
Newer 3D games are almost exclusively polygon based. Basically there are still the same points, lines and polygons, but nowadays the graphics hardware allows for much greater amounts of these primitives to be rendered at a time. The current games still uses pretty much the same polygons and textures as Quake did in 1996, where practically the only actual IL-2 Sturmovik from 2001. improvement is additional methods of filtering and
Mirror, mirror on the wall
If we dare take a peek into the future of 3D graphics in games, the next revolution seems to be in the field of lighting. One of the games that promises great improvements in this area is Doom 3 (word of mouth talks about a release in late 2002, but nobody probably knows really), which is said to step away from the pre-calculated lightmaps, to the totally dynamic shadowmaps. The terminology can get a bit confusing sometimes, since lightmaps and shadowmaps sounds pretty much like the same thing. The techniques have pretty much nothing in common though. Trying to explain even the basic workings of the shadow maps would get very hairy very quickly, and is therefore not discussed here.
Doom 3 (not yet released).
The near future of 3D games is a very interesting and dynamic area, although it doesn’t seem to offer any breakthroughs of the same calibre as Wolfenstein and Quake did in its days. Still, it may be wise not to make the same mistake as Charles H Duell did in 1899 by saying: “Everything that can be invented has been invented”
Article by Anders Stenberg, Mattias Stridsman, Sara Söderberg
Recommended reading
- General computer graphics: 3D Computer Graphics by Alan Watt - Implementations and 3D computer graphics in games: http://www.gamedev.net http://www.gamasutra.com