HMM-based Offline Recognition of Handwritten Words Crossed Out with Different Kinds of Strokes
Laurence Likforman-Sulem, Alessandro Vinciarelli · ENLIGHTEN (Jurnal Bimbingan dan Konseling Islam) · 2008
In this work, we investigate the recognition of words that have been crossed-out by the writers and are thus degraded. The degradation consists of one or more ink strokes that span the whole word length and simulate the signs that writers use to cross out the words. The simulated strokes are superimposed to the original clean word images. We considered two types of strokes: wave-trajectory strokes created with splines curves and line-trajectory strokes generated with the delta-lognormal model of rapid line movements. The experiments have been performed using a recognition system based on hidden Markov models and the results show that the performance decrease is moderate for single writer data and light strokes, but severe for multiple writer data.