Hiroaki HARADA, Japan | ISBN: | Submitted: Jun 29, 1998

Virtual Himeji-castle: A Photorealistic Modeling Technique based on Pictures

Computer Graphics (CG) is a useful technology for creating views of buildings from any viewpoints and for simulating any environmental conditions. Describing and representing the outward appearance of cultural treasures by CG technology will enable people around the world to enjoy their sight without actually travelling to their actual location.

However, it is rather demanding to represent by graphic data cultural treasures such as huge, old, and complex buildings with low costs and with short preparation time.

We have developed a new modeling tool, SketchVision, which is based on photographs or video images. We have successfully represented some old temples in Kyoto and Himeji castle using this tool. The graphic data produced by our tool can be output in VRML (Virtual Reality Modeling Language) so that it can be mode available and shared over the internet.

This paper gives an outline of SketchVision, describes the modeling process to create the virtual Himeji-castle and virtual Kyoto temples with SketchVision, and includes remarks on the precision of the shapes and the textures. In the end of this paper, we comment on the efficiency of the modeling process and its advantages.

In SketchVision, a building is modeled with the following process:

1) Capturing images: Photographs of the respective buildings are taken by standard cameras or video cameras from any positions and angles. Special instruments are not required. These pictures are subsequently digitized and input into the computer.

2) Selecting and fitting to a model: The user selects a suitable 3-dimensional model from a pre-defined object-list, and roughly fits this model to the images manually. During this process, the user can use various graphical 3-D operations to the 3-D models: "shift", "rotation", "expansion", and "distortion".

3) Automatic fitting: After the manual operation is completed, the computer will search for edges in the images so as to fit the 3-D model to the images precisely.

4) Extracting textures: Textures for covering the 3-D models are extracted from the original images automatically.

5) Combining models: Basic 3-D models that were created by the steps 1)-4) are combined to obtain the total, complex model of the building.

We report that we were able to create the VRML models "Virtual Kyoto" and "Virtual Himeji-castle" by using this tool in only 3 months and 2 months, respectively. In conclusion, we found that this modeling tool is efficient and powerful for virtualizing cultural treasures.

[more information]

[other authors]
Masaki WATANABE(*),Shuichi SHIITANI(*),David COLLEEN(**)

[keywords]
Virtual Reality, Computer Graphics, VRML, Photograph, Image Processing