The subjects about the automatic extraction of road information from maps have been recognized as effectual means to construct GIS(Geograpliic Information Systems) and support the services of various kinds of information, the managements of our life-lines, the planning of city projects and so on. Many researches were reported with respect to this subject. However, it is not always easy to accomplish this subject successfully because roads are, in general, overlapped and interrelated with other map components complicatedly. In this paper, we propose an experimental approach to extract road information from urban maps. The characteristic in our approach is not to extract roads directly like many traditional approaches, but to distinguish individual roads cooperatively on the basis of recognition of character regions. We do not only mention the recognition method in our approach, but, also discuss the recognition results through some experiments.