Road generalization is a method for thinning out road networks to allow easy viewing according to the size of the map. Most conventional road generalization methods mainly focus on the length of a stroke, which is a chain of links with good continuity based on the principle of perceptual grouping applied to network data such as roads and rivers. However, in the case of facility search in a web map service, for example, a “restaurant guide map,” a road generalization mechanism can be more effective if it depends not only on the stroke length but also on the facility search results. Accordingly, in this study, we implement an on-demand road generalization method that adapts to both the facility search results and the stroke length. Moreover, a sufficiently fast response speed is achieved for practical use in web map services. In particular, this study proposes a fat-stroke model that links facility information to individual strokes and implements a road generalization method that uses this model to improve the response time. In addition, we develop a prototype based on the proposed system. The system evaluation results are based on three indicators, namely, response time of the road generalization system, connectivity between strokes, and connectivity between stroke and facilities. Our experimental results suggest that the proposed method can yield improved response times by a factor of 100 or more while affording higher connectivity.

A fisheye map lets users view both detailed and wide areas. The Focus+Glue+Context map is a fisheye map suited for Web map systems; it consists of a detailed map (i.e., Focus), wide-area map (i.e., Context), and an area to absorb the difference in scales between Focus and Context (i.e., Glue). Because Glue is compressed, the road density is too high to draw all of the roads in this area. Although existing methods can filter roads to draw, they have problems with rendering the road density and connectivity in Glue. This paper proposes an improved method to filter roads in Glue by applying a generalization method based on weighted strokes. In addition, a technique to speed up the proposed method by using a weighted stroke database is described. A prototype Web map system with a high level of response was developed and evaluated in terms of its connectivity, road density, and response.