Extraction and restoration of digits touching or overlapping lines

There are many cases that characters written in a form document touch or overlap surrounding lines. In the case of slips, the special table-form documents, digits are written in specified regions and they are often contact with lines of the regions. This problem interferes with digit extraction and therefore drops system performance. This paper proposes a new method that correctly extracts and properly restores digits touching or overlapping lines. The new method extracts pre-digits from input data and divides each of them into a digit-component and a line-component. And then, it decomposes a line-component into type A restoration parts, type B restoration parts, non-restoration parts, and candidate generation parts. Finally it restores each part with different methods.