A Restoration Method for Broken Handwritten Line-figures with Connecting Terminal Points

In some pattern recognition processes bmed on recognizing structural features, a serious problem is poorly-cbawn lines which cause broken parts when a digital bitmap (a binary image) is madeof the line-jgure. This broken pattern changes the pattern structure, and increases the likelihood of rejection or misreading during the recognition process. Therefore, we have developed a restoration method for broken handwritten line-figures. In the restoration, broken parts areselected and connected smoothly. In an experiment using handwritten numerals, about 70 % of the patterns that were restored were recognized correctly. also a not-broken-part in the upper right, as shown in Fig. 2(b). Figure 2(b) shows that an extra hole is formed at the top of the pattern, and as a result, the structure of the pattern differs from the standard pattern in the recognition dictionary. Therefore, in structural analysis recognition, changing the pattern structure can seriously d c e the probability of recognition. To solve this problem, a restoration method that can be focused on selected broken-parts is needed. This paper proposes a restoration method where only the broken part is selected by using the direction and length of the broken part. The effectiveness of this method when applied to a handwritten line-figure is demonstrated.