Recovering drawing order from static handwritten images using probabilistic tabu search
Takayuki Nagoya, Hiroyuki Fujioka · 2011
This paper considers the problem for recovering a drawing order from static handwriting images with single stroke. Such a stroke may include the so-called double-traced lines (D-lines). The problem is analyzed and solved by employing the so-called graph theoretic approach. Then the central issue is to obtain the smoothest path of stroke from a graph model of input handwriting image. First, the graph model is constructed from the input handwriting image by employing thinning algorithm. Then, we locally analyze the structure of graph at each vertex. In particular, the method to identify D-lines is developed by introducing the idea of `D-line index'. The method enables us to transform any graph models including D-lines to semi-Eulerian graph models. Then, the restoration problem reduces to maximum weight matching problem of graph, thus a probabilistic tabu search algorithm is developed to solve the problem. The effectiveness and usefulness are examined by some experimental studies.