A Study on Automatic Image Arrangement in Urban Area - toward the New Concept of Urban Visualization -

Recently, visualization of urban scenes attracts attention from the perspective of tourism and cultural city activation. Considering the popularity of the digital still cameras, a more impressive form of urban scenes, which allows the user to view the scenes from connected various points, is required. The still images can be connected in three dimensional space relatively. By compensating some lacks based on human spatial cognitive ability, a lively urban visualization is expected. In order to implement the urban visualization technique by using many still images, multiple resolution matching of images and spatial arrangement of the images are required. This study proposes a technique of urban visualization consisted of multiple resolution image matching, spatial image arrangement, and effective image projection for impressive visualization, and then explores the possibility of the new concept of urban representation based on the technique. The matching strategy for multiple resolution matching is composed of multiple resolution image creation by wavelet analysis, resolution ratio seeking, coarse to fine strategy matching, estimation of templates deformation, voted block matching. According to the matching results, relative orientation is applied, and then images are arranged in a three dimensional space. To keep the stability of the solution, calculation method for relative orientation is modified. The proposed technique was applied to images taken around a building, and formed a closed network. The matching result was enough to visualize the urban scene compared with manual matching. Through the application, the significance of the technique and limit of the technique was confirmed.