Surface simplification with semantic features using texture and curvature maps

We propose a polygonal surface simplification algorithm that can preserve semantic features without user control. The semantic features of a model are important for human perception, which are insensitive to small geometric errors. Using an edge detects: Its three kinds of maps are employed to extract these features. First, an image map is generated boundary lines represent changes of chroma in the texture image by using edge detector. Second, the discrete curvatures at 3D vertices are mapped to the curvature map, and their data is also analyzed by an edge detector. Finally, a feature map is generated by combining the image and curvature maps. By finding areas of the 2D map that correspond to areas of the 3D model, semantic features can be preserved after simplification. We demonstrate this experimentally.