Computer images are represented in one of two fundamentally different ways, and the distinction runs through the whole history of computer graphics. A raster (or bitmap) image is a rectangular grid of picture elements, or pixels, each storing a color value. A vector image instead describes a picture as a set of geometric primitives — points, straight lines, curves, and filled shapes — defined by mathematical coordinates rather than a fixed array of samples.
The vector approach is captured directly in the W3C’s Scalable Vector Graphics specification, which states that “SVG allows for three types of graphic objects: vector graphic shapes (e.g., paths consisting of straight lines and curves), images and text.” Because the shapes are stored as equations and control points, a vector image can be scaled, rotated, or printed at any resolution without losing sharpness; the renderer recomputes the pixels for whatever device is in use. This makes vector formats the natural fit for fonts, logos, technical drawings, maps, and illustration — anywhere crisp edges must survive arbitrary zoom.
The raster approach is the one that screens and cameras speak natively. A display is itself a grid of physical pixels driven from a framebuffer, and image-capture and photographic formats are inherently sampled. The PNG specification (RFC 2083) describes its subject as “a lossless, portable, well-compressed storage format for raster images,” and the same pixel-grid model underlies JPEG, GIF, and the bitmap stored in any framebuffer. Raster images excel at photographic and continuous-tone content that has no clean geometric description, but they are tied to their resolution: enlarge a bitmap and its individual pixels become visible, a defect vector graphics do not share.
In practice the two representations are complementary and constantly converted into each other. Almost all final display is raster, because the screen is a raster device, so vector content must be rasterized — turned into pixels — before it can be shown, the job of the graphics pipeline’s rasterization stage. Conversely, tracing or vectorizing turns a bitmap back into editable shapes. The early arcade and home-computer eras leaned heavily on sprites and bitmaps because hardware drew pixels directly, while CAD systems and plotters were vector-native from the start; modern graphics stacks routinely mix both, storing artwork as vectors and emitting pixels on demand.