基于深度图像的建模绘制技术研究

【作者】 耿振伟；

【导师】 蔡宣平；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2002， 硕士

【摘要】 基于图像的绘制技术目前已在图形学领域引起了极大的关注，它在图像合成方面的强大优势给传统基于几何绘制技术带来了另一种选择。基于深度图像的几何模型绘制从另一个角度弥补了传统基于几何绘制技术的不足，它绘制速度更快、与模型复杂度无关而且对资源要求更少，因而对于大数据量几何模型的简化和实时三维仿真具有重要的研究价值。 本文深入研究了基于深度图像的几何模型建模绘制技术，着重分析研究了这一技术领域的3D Warp算法，在深入理解其算法原理的基础上对这一重要算法进行了理想的数学分解，使图像的3D Warp变换在实现中可分为预Warp和纹理映射两步完成，充分发挥了纹理映射硬件的优势，极大地加快了图像Warp变换的速度。分解后，图像预Warp变换等式具有非常简单的一维结构，可以沿图像的行、列独立地一维操作实现，每次只须在两个相邻的像素之间线性插值处理空洞填充。最后本文构建了一个软件系统，能够采样几何模型6个面的深度图像，按包围盒方式对整个模型重新进行基于图像的建模绘制，在微机上实现了大数据量几何模型的快速实时显示。 本文在理论上详细推导了3D Warp等式的数学分解；实践上系统论证了图像一维Warp变换的正确性和健壮性；实现了采样几何模型6个面的深度图像后按包围盒方式重新绘制模型的具体过程。更多还原

【Abstract】 Image-based rendering techniques have gained considerable attention in the graphics community because of their potential advantage in image synthesizing which can bring another alternative for traditional geometry-based rendering techniques. Depth image-based rendering for geometry models gives another compensation for traditional geometry-based rendering techniques with its faster rendering,independent of geometry complexity and needing less computing resource,So it is very valuable for simplifying very large data geometry models and displaying them in real time.This dissertation deeply investigates depth image-based modeling and rendering techniques,further on 3D image-warping,presents a division of 3D image-warping in mathematics with fully understanding it,Then 3D image-warping can he implemented by pre-warp and texture mapping. This can efficiently take advantage of the texture mapping hardware. So the warping operation is much faster. After factoring the 3D image-warping equation,the pre-warping equation has a very simple structure. It can be divided into two-pass process using 1-D operations along rows and columns and only interpolating between two adjacent pixels each time. At last,an image-based modeling and rendering system is cast,which can display a geometry model with images after sampling six depth images around a model’s bounding box,displaying geometry models in real-time is realized under ordinary computer.This dissertation presents a concrete deduce for factoring the 3d image-warping equation;confirmed the correctness and robustness of ID warp and reconstruction;realizes the whole modeling procedure after sampling six depth images around the model’s bounding box.更多还原