Text Line Separation in Handwritten Korean Address Images

摘要

We describe a method of separating text lines in handwritten Korean address images, based on vertical partition of input images and grouping and splitting of text line subimages, which mean images into which a text line is segmented vertically and horizontally. The method is robust to address images which normally appear in the application fields. In other words, it can deal with undulated, skewed, vertically contacted or vertically overlapped text lines under a stabilized performance. The experiment done with address images extracted from Korean live mails has demonstrated the superiority of the proposed method. The success rate for the 989 testing images was about 88.8%.