A Graph Theoretic Algorithm for Recovering Drawing Order of Multi-stroke Handwritten Images

This paper proposes a novel algorithm to recover a drawing order from static handwriting images with multi-strokes including double-traced lines (D-lines). Such an algorithm is developed by employing the graph theoretic approach. Then, main issue is to find the smoothest path of each stroke from an undirected graph model corresponding to input handwriting image. First, the structure of graph at each vertex is locally analyzed. Such a structure is decomposed to ones corresponding to single-strokes of handwriting. In addition, any graph models including D-lines is transformed to semi-Eulerian graph models by employing the idea of D-line index. Then, the problem of recovering drawing order reduces to maximum weight perfect matching problem of graph. For solving the problem, we here employ a probabilistic tabu search algorithm. The effectiveness and usefulness are examined by some experimental studies.