Automatic Generation of Ortho-Photo Texture from Digital Elevation Model

We propose the automatic generation of the ortho-photo data which support realistic scenes for DEM by texture mapping. This ortho-photo data is automatically generated by pattern recognition techniques using Bayesian classifier which uses the features extracted from a DEM and its geo-referenced ortho-photo data as training sets. We defined the various features of each texel such as its height, slope angle, slope direction, surface curvature, hue, saturation and brightness from the training datasets. The proposed method makes possible for mapping texture of a realistic ortho-photo data to virtual terrain data which are unable to take satellite photo or aerial photo. These case are often in of computer game and digital movie area. Also, generating ortho-photo with the enlarged DEM, it does not cause the aliasing from the difference of resolution. It makes very similar images with real photography by shading and efficiently handles ortho-photo data and elevation data occupied enormous storage in cloud computing environment.