Multi-View Image Precise Texture Mapping Method Based on Frame Buffer

Detailed texture mapping methods for building 3D models normally select the best matching image for each single face. However, face associated single images often have partial occlusion problems that impact the quality of mapped textures. An accurate multi-view image texture mapping method based on frame buffer rendering is presented here. First, the coordinates of the single independent building model and the triangle mesh model produced using oblique photogrammetric data were registered. Following this, rendering primitives based on spatial characteristics of the single independent building model were extracted. Subsequently, rendering cameras for the triangle mesh model were established depending on the rendering primitives. Using the Render to Texture (RTT) method, textures corresponding to rendering primitives from a frame buffer object (FBO) are obtained while rendering the triangle mesh model using the rendering cameras. Finally, the textures of the surface of a detailed building model were remapped using rendering primitives, and automatic texture mapping was achieved from the triangle mesh model to form a detailed single independent building model. The usage of oblique photogrammetry data demonstrates that the occupation of the defective textures is reduced by over 14% using our proposed method compared to the traditional method, and the size of the textures were compressed to 14.21% with more accurate texture mapping quality, thereby satisfying the texture reconstruction requirements for a detailed building model.