Vector Graphics¶
Images are specified as coordinate-defined geometric objects that can be edited and rendered for different output conditions, rather than as one fixed pixel grid.
Core Idea¶
Vector graphics specify an image through coordinate-defined geometric objects and their presentation relationships rather than only through a fixed grid of final pixel samples. A line is defined by endpoints or a path command, a curve by control points or other parameters, and a region by its boundary; fill, stroke, ordering and transformations determine how those objects contribute to the image. The source description can then be drawn for a particular size and device. What makes the model repeatable is not a particular SVG file, font, map, editor or vector monitor, but the separation of geometric image specification from device-specific realization.[1][2][3]
That separation has qualified benefits. The same source outline can be rescaled, transformed and edited at the object level without merely enlarging an existing bitmap. The rendered image, however, is still finite-resolution: a rasterizer selects pixels, and font hinting may alter outlines at particular sizes to protect legibility. Nor is every described contour an exact account of a real-world shape. Geometry can approximate a coast, letterform or photographed object. The vector model stores the chosen geometry, not a guarantee of unlimited visual fidelity, universally small files or zero representational loss.[3][4]
W3C's SVG language exemplifies the model with paths that can be moved, lined, curved, closed, filled and stroked. But SVG can also contain embedded images and text: the file format is not identical to the vector-graphics identity. OpenType outline glyphs and symbolized GIS features show the same underlying organization in two unlike settings, with different coordinate conventions and rendering goals.[1][2][5][4][6]
Structural Signature¶
Sig role-phrases: coordinate frame → geometric objects → object organization → presentation mapping → output realization → model boundary.
- Coordinate frame. Positions and transformations are interpreted in a declared space: SVG user coordinates, font design units, or a geographic reference system. The numbers are not self-interpreting; applying the wrong frame changes where and how the object appears.[2][5][4]
- Geometric objects. Points, line segments, curves, contours and regions carry the image's shape before final device sampling. Replace them with only a frozen pixel array and the vector description is lost.[2][5][4]
- Object organization. Paths can contain subpaths, glyphs can be composite, and map features can be collected or layered. A one-object graphic is a valid degenerate case, but multi-object images depend on arrangement and composition, not just an unordered list of coordinates.[2][5][4]
- Presentation mapping. Fill, stroke, glyph treatment or cartographic symbol connects geometric objects to visible marks. Some data sets retain only geometry and defer this mapping; they become a graphic when presentation is supplied.[1][2][6]
- Output realization. A renderer or drawing device turns the source specification into marks under chosen size, scale and device conditions. The operation is central to seeing the graphic, even though it need not be stored within its source model.[3][6]
- Model boundary. A vector description preserves selected shapes and object identities but cannot promise that finite samples, simplified lines or absent tonal information preserve every feature of an underlying target. This is a use/recognition condition, not necessarily a field inside the file.[3][4]
What It Is Not¶
It is not simply any use of a mathematical vector. The term refers to graphical objects described by coordinates and geometric relations; an array of numerical features without image geometry is not a vector graphic. It is not a particular file extension: SVG supports vector paths but also raster images, and other formats can describe paths or outlines. It is not vector-display hardware: old displays and plotters could draw lines directly, while contemporary vector sources are often rasterized for pixel screens or printers.[1][3]
It is not intrinsically a map, a data visualization or a rendering algorithm. A vector drawing can be decorative rather than an encoding of independently measured information; a GIS feature file may store geometry before its symbolization produces a map; and rendering names the process that produces visible output from such a source. It is not a guarantee of lossless resizing in the perceptual sense: the stored coordinates may be reused, but low-resolution output, hinting, stroke treatment and feature simplification still affect what a viewer sees.[4][6][3]
Scope of Application¶
In illustration and interface graphics, paths and regions organize shape, color, grouping and transformations. SVG specifies a concrete two-dimensional language in which paths, images and text can be combined; only the geometrically described parts exemplify vector graphics in the strict sense here. Complex masks and filters can make an SVG work visually rich without proving every element remains a simple, infinitely rescalable path.[1][2]
In typography, outline fonts store glyph contours in design coordinates and transform them into bitmaps for the requested point size and device. The source outline is reusable, while rasterization and optional grid fitting adapt the actual pixels. In GIS, coordinate-described point, line and polygon features can be symbolized into a map: the geospatial data structure and the completed visual representation should be kept distinct.[5][3][4][6]
The same model is available for technical diagrams and other drawings whose content is naturally decomposed into discrete geometric objects. It is less natural for uncontrolled continuous-tone capture, such as a photograph, unless the author deliberately replaces sampled variation with chosen contours and regions. That replacement can be valuable stylization or analysis, but it is a new modeling decision rather than a lossless conversion.
Clarity¶
The entry makes three layers visible. Source geometry says what objects and relationships are stored. Presentation mapping says how those objects become strokes, fills, symbols or visible glyphs. Output realization says how a particular device turns that design into pixels or physical marks. Conflating these layers produces errors such as calling a low-resolution screenshot “vector” because it once came from an outline, or claiming a raw GeoJSON boundary file already determines the appearance of a map.[3][4][6]
It also resolves the ambiguity in “scalable.” A vector source can be re-evaluated at another size without enlarging a previously sampled pixel array. That does not ensure equal perceived quality across sizes: in typography, size-dependent grid fitting may deliberately modify an outline to avoid dropout or uneven stems. Scalability is a property of the descriptive model; successful rendering is a further condition.[3]
Manages Complexity¶
The model compresses a potentially large set of output pixels into a smaller set of structured decisions when the image really is composed of relatively few objects. A glyph can be stored as contours and control points, then reused at several sizes. A map layer can retain roads or parcels as separate line or polygon features and change their symbolization without redrawing each map pixel by hand. The reduction is from many device samples to reusable geometric relations and style decisions.[5][3][4][6]
Compression is conditional. A very intricate drawing may require many control points or layers; trying to approximate continuous photographic variation with geometry may proliferate objects and lose the simplicity that motivated the model. File size is governed by actual complexity, encoding, metadata and optional embedded content, not by the word “vector” alone. The model manages complexity best when meaningful image parts line up with reusable geometric objects.
Abstract Reasoning¶
To recognize vector graphics, ask whether the candidate stores or generates coordinate-defined image objects that can be independently transformed or restyled before final output. Identify the coordinate frame, geometric primitives, object relationships, styling rule and rendering step. Then ask which visual features are preserved by that description and where approximation or sampling enters. A file containing only an embedded bitmap fails the test despite an SVG wrapper; a glyph contour passes despite its eventual display as pixels.[1][5][3]
This decomposition supports practical inferences. If an illustration is stored as paths, changing its output size can trigger fresh rendering from source geometry. If only a raster export survives, its object identities and control points cannot be recovered automatically just by enlarging it. For a map, changing symbology may alter appearance without changing the underlying geographic coordinates; changing the geometry is a different intervention. These are consequences of the representation/rendering distinction, not guarantees of perfect fidelity.[6][4]
Knowledge Transfer¶
The literal structure transfers from outline typography to a styled GIS map: both retain coordinate-defined objects, separate those objects from final visible marks, and render for a particular output context. The object meanings differ. A glyph contour is a designed letterform in font units; a GeoJSON road or parcel is a geospatial feature under a coordinate convention, whose symbol may be selected by a map renderer. A property of one realization—font grid fitting, for example—does not automatically carry to the other.[5][3][4][6]
The nearest portable skeleton is geometric representation or delayed rendering. Live Representation may illuminate many map and diagram cases, but its full target/medium/faithfulness signature is not a necessary genus for generative abstract vector artwork whose intended graphic is made in the medium rather than mapped from an independently identified target. A more precise cross-domain “late-bound geometric description” abstraction is a future-prime question requiring independent non-graphics cases. The named vector-graphics entry remains bound to image geometry and visible realization.
Examples¶
Outline typography at multiple output sizes¶
OpenType's TrueType glyf table specifies simple glyph outlines through contour and Bézier control-point data; composite glyphs may reference other glyphs. A scaler converts design coordinates for a requested size, optional instructions adjust a grid-fitted outline, and scan conversion selects a device bitmap. The same source glyph can therefore support multiple output sizes without storing a separate fixed bitmap for each size, while the pixel result can still depend on size and hinting.[5][3]
The coordinate frame is the font's design-unit system. Geometric objects are outline contours and control points. Object organization is contour order or composite-glyph references. Presentation mapping includes the glyph's visible fill and size-specific grid-fitting choices. Output realization is scaling, optional hinting and scan conversion to the device's pixels. The model boundary is that the finite outline and output grid do not guarantee identical appearance at every point size.[5][3]
Mapped back: This is a vector graphic because reusable coordinates and contours, rather than one final bitmap, carry the letterform into differently scaled realizations. Font hinting demonstrates why source scalability and rendered perfection must not be conflated.
A symbolized GIS map layer¶
RFC 7946 defines GeoJSON geometries such as Point, LineString and Polygon and a Feature structure that keeps geometry separate from properties. QGIS documentation describes renderers and symbols that draw vector features as visible map marks. Consider a map view that assigns line symbols to road LineStrings and fill symbols to parcel Polygons. The road and parcel coordinates are not their screen pixels, and their cartographic appearance is a further choice.[4][6]
The coordinate frame is GeoJSON's specified geographic coordinate convention. Geometric objects are the coordinate-defined line and polygon features. Object organization comes from Feature collections and map layers. Presentation mapping selects line and fill symbols, potentially according to feature attributes. Output realization draws those symbols at the selected map view or export scale. The model boundary is that simplified boundaries and symbol choices preserve some spatial structure but need not preserve photographic texture, survey precision or every detail at every scale.[4][6]
Mapped back: This is the same geometric-description-to-visible-output structure as the glyph, but the objects are spatial features with geographic semantics. A bare GeoJSON file is vector data; it becomes the claimed vector-graphics map case when a presentation mapping renders the features.
Structural Tensions¶
Object-level editability versus continuous-tone coverage. Discrete paths keep parts independently editable and transformable. But to reproduce dense, irregular tone with those parts can require proliferating shapes, while a raster samples the tone directly and sacrifices the original object decomposition. Diagnostic: Is the intended image best understood as stable geometric objects or spatially varying samples?
Reusable source geometry versus finite output fidelity. The source outline can be evaluated at new sizes, which avoids enlarging an old pixel grid. A low-resolution rasterizer may still produce dropouts, uneven strokes or a need for hinting; treating source reuse as visual perfection hides this cost. Diagnostic: Which feature is invariant in the source, and which changes when it is realized on the target device?[3]
Geometric simplification versus representational accuracy. A map with fewer path vertices is easier to manipulate and render, but omitting a bend, hole or small parcel boundary may change the task's answer. More detail can increase complexity and clutter at small display scales. Diagnostic: At what use scale does the omitted geometry alter a relevant spatial inference or visual decision?
Structural–Framed Character¶
- Evaluative weight: “Vector” names a descriptive image model, not an inherent aesthetic or technical superiority over raster media; the right model depends on the content and output task.
- Human-practice dependence: Artists, typographers and cartographers choose which objects and styles to encode, but the coordinate/object/rasterization relations can be checked independent of their institutional identities.
- Institutional origin: SVG, OpenType and GeoJSON are standardized realizations; no single standard constitutes the whole vector-graphics pattern.[1][5][4]
- Vocabulary travel: The word “vector” travels to linear algebra, GIS and design, but its graphics meaning is not the generic mathematical vector and must keep an image-geometry referent.
- Import versus recognition: A case is recognized by recoverable coordinate-defined graphical objects and a rendering pathway. Calling a bitmap a vector because it sits in a vector-capable container imports the label without the structure.
Its character: structural as a reusable representation schema within computer graphics, yet domain-specific. Geometric image objects and visible realization are constitutive, and their literal tests do not automatically transfer to all uses of “vector” or “representation.”
Structural Core vs. Domain Accent¶
The skeletal relation is to keep a manipulable geometric specification separate from a later output realization. The domain-bound mechanism is image geometry: coordinate frames, paths/contours or feature shapes, graphical organization and styling, and a renderer or drawing device. Without that image-related chain, the phrase “vector graphics” has only metaphorical reach.[2][5][6]
The application overlay varies—font legibility, illustration editability, or geographic map communication—but is not the identity. The entry does not clear the prime bar because its recognition tests depend on geometric image objects and graphic realization. General representation is an actual live prime but is not asserted as strict parent for generative art; a broader late-bound geometric-description pattern is an explicit future-prime question, not an already established bridge.
Instantiates / Related Primes¶
Representation is a close and often instantiated neighbor when vector geometry maps an independently identified target—such as roads or parcels—onto a graphic medium with known omissions. It is not a necessary superclass of every decorative or generative vector image under that prime's full independent-target and faithfulness conditions. Visualization (graphics) is likewise nearby when the graphic deliberately encodes information, but an abstract vector ornament need not perform that job. Rendering (computer graphics) names the production process from a description; the source vector graphic is not itself that process.
Neighborhood in Abstraction Space¶
Vector Graphics sits in a moderately populated region (58th percentile for distinctiveness): it has near-neighbors but no dense thicket of look-alikes.
Family — Spatial Perception & Navigation (21 abstractions)
Nearest neighbors
- Zooming User Interface — 0.86
- Responsive-Layout Breakage — 0.85
- Multivariate Glyph — 0.85
- Wireless triangulation — 0.85
- Motion Compensation — 0.85
Computed from structural-signature embeddings · 2026-10-08
Not to Be Confused With¶
- Raster graphics: stores or presents samples on a pixel grid. A rendered vector image will often become raster on a modern screen, but the source model is not thereby retroactively a pixel-only source.[3]
- SVG: a standardized graphics language capable of paths, images and text; a document using only an embedded bitmap does not make its visual payload vector geometry.[1]
- Vector display: hardware that directly draws geometric strokes; vector data can instead be rasterized for ordinary screens and printers.[3]
- GIS vector data: coordinate-defined features may be studied as spatial data before they are given cartographic symbols. A styled map is the image case, not an automatic property of every geometry file.[4][6]
- Mathematical vector: an element of a vector space can participate in graphical coordinates, but its bare mathematical identity does not specify a graphic.[2]
References¶
[1] W3C SVG Working Group, “Scalable Vector Graphics (SVG) 2: Introduction”, §1.1. Original standards-body description of vector shapes, embedded images, text and graphical-object styling. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h
[2] W3C SVG Working Group, “Scalable Vector Graphics (SVG) 2: Paths”, §§9.1–9.3. Original path geometry, commands, fill and stroke specification. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i
[3] Microsoft and Adobe, “OpenType 1.9.1: TrueType Fundamentals”, From Font File to Paper, Scaling, Grid-fitting and Scan Converter sections. Original outline-to-bitmap account. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q
[4] H. Butler et al., “RFC 7946: The GeoJSON Format”, §§1.4–1.5, 3.1–3.2 and 4, 2016. Original standards-track geometry and Feature specification. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p
[5] Microsoft and Adobe, “OpenType 1.9.1: glyf—Glyph Data Table”, glyph-header and simple/composite-glyph sections. Original font-format specification. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l
[6] QGIS Project, “PyQGIS Developer Cookbook: Using Vector Layers”, §6.8 Appearance (Symbology) of Vector Layers. Original project documentation for rendering point, line and polygon features. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m