Image analysis using line segments extraction by chain code differentiation

Abstract

This paper proposes a new fast method for line segment extraction from edge maps. It has a parallel nature and can be used on parallel machines easily. The method uses the chain codes in the edge map, namely macrochains, for line segment detection. In the first phase, it breaks the macrochains into several microchains by employing the extreme points of the first derivative of shifted-smoothed chain code function. Straight-line segments approximate the resulting microchains. In the second phase, the line segments are grouped together based on their proximity (collinearity and nearness) to make longer segments. The final set could be tailored for any minimum segment length and minimum error desired.