Image glossary
Raster vs vector
Also known as bitmap vs vector, raster graphics, vector graphics
A raster image stores a fixed grid of pixels and degrades when enlarged; a vector image stores shapes as mathematical descriptions and renders sharply at any size.
Two different things to store
A raster image is a grid of coloured samples. Its resolution is fixed at creation, and enlarging it means inventing samples that were never captured. JPEG, PNG, WebP, AVIF and every photograph are raster.
A vector image is a set of instructions: this path, these control points, this fill, this stroke width. There is no resolution, because the renderer computes the pixels fresh at whatever size is asked for. SVG, PDF artwork and most logo source files are vector.
Which suits what
Photographs must be raster. There is no compact set of shapes that describes a face or a landscape; the information really is per-pixel, and a vector approximation of a photograph is both enormous and worse.
Logos, icons, diagrams, charts and typography should be vector wherever a source exists. A few hundred bytes of path data stays perfectly sharp on a business card and on a billboard, and recolouring it is editing a fill value rather than re-exporting an image.
The practical consequence people hit most often: asked to enlarge a logo supplied as a small PNG, no tool can fully recover the crisp geometry, and an AI upscaler will confidently produce something logo-shaped with the wrong letterforms. Find the vector. It is almost always the correct answer and almost always faster than the alternatives.
Crossing between them
Rasterising a vector is exact and routine: you choose a size and the renderer produces pixels. Anything can be turned into a raster.
Going the other way — tracing a raster into vectors — is an approximation. It works acceptably on flat, hard-edged artwork with few colours, and produces nonsense on a photograph. It is a recovery technique for a lost source file, not a workflow.
Tools that work with this
Related terms
- Aspect ratioAspect ratio is the proportional relationship between an image's width and height, expressed as two numbers such as 16:9, and it is what determines whether a platform crops your image.
- Super-resolutionSuper-resolution is the use of a trained model to increase an image's resolution by generating plausible detail consistent with its training data, rather than by interpolating between existing pixels.
- Bicubic interpolationBicubic interpolation is a resampling method that estimates each new pixel from a weighted cubic function of the sixteen surrounding source pixels, producing smoother results than simpler methods without adding any real detail.
- PNGPNG is a lossless raster image format with full alpha transparency, excellent on screenshots and line art and very inefficient on photographs.
Back to the full glossary.