In recent times graphics companies are constantly striving to achieve more and more detail in computer rendering. New technologies are steadily being developed that advance rendering architectures and techniques. And features such as new compression algorithms, memory architectures and forms of occlusion culling are constantly being worked on and improved. Unfortunately, we still have a considerable distance to travel in achieving what is widely considered as the ultimate goal: photo-realistic quality graphics on the PC. Because of this, today's graphics must rely on some basic technologies and in this article we will discuss texturing and, more specifically, texture filtering.

With a perfect computer, we would be able to display an infinitely high amount of geometry and pixels. In doing this we would have incredible detail, assigning millions, or even billions, of pixels each a unique color value that would be used to render richly detailed scenes. Everything would appear just as it is in real life. Unfortunately, this is not yet possible. There are many limitations in calculations, memory, and bandwidth, which mean we must come up with innovative ways to simulate detail--in other words, we must cheat.

One of our primary cheats in achieving greater detail is texturing. While most would not consider texturing a cheat, in reality it is little more than just that. Consider what a texture is, a 2D image of a 3D object(s) or surface. It is really like a photograph of something. Take a brick wall, for example. You want to show other people what it looks like, but you do not want to take the bricks with you. A much easier solution would be to bring a picture of the brick or a model of the wall. And this is exactly how a texture works. You are showing a 2D image of a 3D object or surface. The brick wall and its geometry and detail are too complex to efficiently manage and display so instead, by using a texture, we simulate this minutiae. The texture cheats by making an object look as though it has details in color and geometry that really do not exist.

When considering texturing, one must understand both what a texture is and what it is comprised of at its lowest level. The simplest answer is that a texture is comprised of numerous color fragments. These color fragments, or texels (texture elements), are what we sample to determine the individual color value of each rendered pixel, with the total number of texels depending entirely on the resolution of the texture. And sampling a texture is just like sampling anything else--by doing so you are taking a small portion of the texture to represent a larger area. This basically allows for reduced computations and increased overall performance.