Level of detail (LoD) in computer graphics refers to the process of simplifying the different graphical aspects of a three-dimensional (3D) object that is being rendered at a distance. The purpose of implementing level of detail into a program is so the speed of rendering can be increased when there are multiple objects in the virtual scene and some of them are so distant from the viewer that extraneous details would not be visible anyway. Depending on the exact way level of detail is integrated into a program, it can be one of two types: discreet or continuous. Discreet LoD replaces 3D models or textures with separate, usually premade, models that have less detail. Continuous LoD uses algorithms to dynamically increase or reduce the detail on a model as needed.
Nearly all 3D graphical programs use some form of level of detail, especially video games. If properly implemented, LoD is nearly unnoticeable to the viewer. When the calculated distance between an object in the virtual scene and the position of the viewer is above a certain number, the level of detail can be reduced. This reduction usually results in either less polygons making up the 3D object, reduced quality texture images, or a combination of both. The rendering speed of a scene that uses level of detail is considerably faster than the rendering speed of a scene where objects are consistently at full detail.
Discreet level of detail is a method that uses a set number of distances to determine which model of an object will be rendered. The different quality levels of the 3D model can either be created in a modeling program or calculated when the program starts and then stored for later use. This method works well, because it is simple and fast. It also is not favored by some programmers and artists because, occasionally, the jump from one LoD model to another might appear too obvious and cause an object to suddenly appear larger and more detailed in front of the viewer.
Continuous level of detail actually uses algorithms to take the polygons making up an object and either subdivide them for greater detail or consolidate faces for less detail. This version of LoD allows for a smooth transition from a single pixel to full detail as the distance is lowered. The algorithms, however, can sometimes be very processor intensive and also have the possibility to produce undesirable results, such as missing polygon faces or changes to the 3D models that distort the initial geometry.
Other variations of LoD include reducing the image quality of the texture mapped images, using very low-quality small images instead, or even just using a solid color with no texture mapping to approximate the appearance from a distance. In another method, instead of actually loading a reduced quality 3D model for LoD purposes, the model at a distance is replaced by a geometric primitive such as a sphere or rectangle, which renders far faster than arbitrary polygons. There are other algorithms that can be used for more specific types of LoD calculations, such as those needed to modify terrain meshes and those that can be used to approximate objects that are moving quickly across the scene.