Recovery of Information on the Drawing Order of Single-Stroke Cursive Handwritten Characters from Their 2D Images

概要

論文の詳細を見る
Handwriting is easier to recognize when it is done on-line than when it is done off-line, because for on-line handwriting one can use the time information obtained during the writing process, whereas for off-line handwriting one must depend solely on the static 2D image, in which the time information has already been lost. The time information is essentially the order in which the stroke segments constituting the handwritten characters are drawn. We propose in this paper a novel approach to recovering the time information from a static 2D image of single-stroke cursive script, based on a good continuity criterion implemented by applying the SLALOM approximation method. Departing from convention, we will evaluate smoothness not in the local area of crossing points but in the global area of the handwriting image, thus overcoming difficulties resulting from ambiguities or possible noise at the crossing points. To cope with the increased computational burden resulting from the complexity of the script, we divide the image into small components and rely on the principle of "divide and conquer."
一般社団法人情報処理学会の論文
1995-09-15