Perceptual Organization as Graph Rectification in a Constraint-based Scheme for Interpreting Sloppy Stick Figures

James V. Mahoney and Markus P.J. Fromherz

Abstract

Machine systems for understanding hand-drawn sketches and diagrams must reliably interpret common curvilinear configurations, such as arrows, geometric shapes, and conventional signs and symbols. These figures are often sloppily drawn and highly variable from one instance to the next. This paper argues that a perceptual organization process may impact a recognition process for such drawings in three major ways: by allowing great variability of form, by increasing matching efficiency, and by enabling easy extensibility to new configurations.

PDF file

Back to the top.