A New Approach for Perceptually-based Fitting Strokes into Straight Segments

Abstract

Fitting the strokes of a sketch into geometrical primitives is still an open problem, even for sketches which depict bare line-drawings without annotations. Such sketches comprise only discrete strokes, sequences of points obtained between a pen down and a pen up. It is commonly accepted that the best perceptual fittings depend on the context. Hence, we will only be able to extract the best line-drawing from a sketch by considering a complex recognition flow, where lines must be iteratively fitted according to different tentative relationships until the most plausible line-drawing is reached. The recognition task considered in this paper is determining whether a stroke represents a straight line. The goal is doing it in a way that allows for iterative recognition flows. The novel contributions are that our approach is more fast and robust than accurate, uses perceptual criteria to classify strokes, and returns likeliness instead of a simple yes/no.