Vector-Based Occlusion Detection for Automatic Facade Texture Mapping

The demand of a photo-realistic city model is high in recent years due to the development of location-based service has gain much intention. A photo-realistic city model contains not only the geometry, e.g. buildings, roads, trees, terrain, etc., but also the texture information, e.g. ortho-image, facade and roof textures. The purpose of this paper is focus on the facade texture generation using airborne oblique imagery provided that the 3D polyhedral building models are existed. In which, the most critical issue is how to cope with the occlusion problem and perform texture compensation for the occluded areas without radiometric distortion. In this research, the occlusion problem is solved by means of a vector-based occlusion detection algorithm. The self-occlusion and non-self-occlusion situations were all considered. The compensated textures are come from other images taken at different locations or time, thus the color and gray value may have unbalance and discontinuous problems. Thus, the histogram matching and seamless mosaicking techniques are applied to create a harmonize texture image. Experimental results demonstrate that the proposed scheme can effectively and automatically generate the facade texture when occlusion effect occurred. The texture quality can also meet the requirement of cyber-city applications.