The Research On Logical Structure Analysis Of Document Image

There are various kinds of tables in the ordinary documents. How to realize the automatic positioning, the analysis and the content recognition of table image has become an important branch of DIA domain.Based on massive practice, this paper has carried on the beneficial exploration to the table layout structure analysis and the table logical structure analysis. In the table layout structure analysis part, this paper extracts the lines from the table image at first, then gets the intersection points of lines, and at last gains each unit information. The result of table layout structure analysis was represented through the matrix of intersection point of lines. This method not only reduced the difficulty of question description, moreover has manifested the overall characteristic of table layout structure. All this has provided the enormous convenience for the following processing. In the table logical structure analysis part, this paper proposed a new table structure expression method, and through nesting chain can describe the relationships between the nesting structures. And all this has provided the convenient description way for the complex form structure description. This method fully used the dependent relations between the title field and the data field, as well as line intersection characteristic in each basic layout structure. This method not only can analysis filled-in table image's logical layout, but also may segment table image according to the basic layout structure.The practice proved that, this paper stated the method to obtain the good processing effect, has satisfied the form structure automatic reduction practical application request.