Visual Structure
From pixels to color regions, gradient lines and drawable visual objects.
A family of experiments that joined region geometry and line trajectories into a scene-like description of visual content.
The question
Can an image be represented by a set of drawable objects rather than a grid of pixels?
The work explored two complementary forms of structure: coherent color regions and gradient-following lines. A region carries area, color and a boundary; a line carries trajectory, orientation and local geometry. Together they form a scene that can be rendered again and examined as a collection of explicit elements.
From color fields to objects
Color spots & receptive fields
Historical experiments group image areas by coherent color information and local receptive-field behavior, recovering boundaries and broader masses rather than relying only on isolated edge pixels.
Lines and local geometry
Separate experiments extract continuous trajectories from gradient information. Direction, curvature and extent become attributes of graphical primitives.
Combined representation
Color supplies the masses; lines restore the structure.
The combined idea is analogous to an artist building a picture: larger regions first, then lines that make form legible. Its purpose was not merely stylistic image transformation; the aim was to create a structured, renderable representation potentially useful to later recognition stages.
What the record establishes
The archive contains implementations and visual outputs for the constituent methods, together with a documented approach to combining them. It does not establish a production-ready recognition system, trained classifier, or numerical improvement over a reference baseline.
The distinctive engineering move is to separate visual content into interpretable drawing entities that retain geometric properties.
Related work
These connections describe related engineering ideas, not interchangeable products.