Strokes classification of handwritten text based on the Frechet distance
N.M. Pronina, Leonid Moiseevich Mestetskiy · 2023
The paper aims to develop an interpretable similarity metric of the calligraphic handwritten elements, strokes. The need for such a metric often arises in text navigation tasks. To extract strokes, the text is converted into a binary image and a skeletal representation is constructed for all connected components. It provides accurate information about how the pen moved while writing the text. Next, the skeletal representation is sliced into a geometric graph. The strokes are formed as subgraphs. As a result, the new method is proposed which is based on the use of Frechet distance between plane curves. The presented metric has been applied to the tasks of stroke classification and letter recognition based on stroke bigrams and trigrams.