基于SIFT的图像建模算法研究  

Research on Algorithm of IBM Based on SIFT

作　　者：易成涛[1] 王孝通[1] 徐晓刚[2] 

机构地区：[1]海军大连舰艇学院航海系,辽宁大连116018 [2]海军大连舰艇学院装备系统与自动化系,辽宁大连116018

出　　处：《微电子学与计算机》2009年第12期1-3,8,共4页Microelectronics & Computer

基　　金：辽宁省自然科学基金项目(20082176);浙江大学CAD@CG国家重点实验室开放基金项目

摘　　要：为解决飞行或航海模拟器中建模手段单一、真实感较差等问题,研究了基于SIFT的图像建模算法.该算法针对目前图像建模算法中存在的人工参入图像匹配负担较大问题,实现了基于尺度不变特征的图像自动匹配;针对重建物体视觉外观真实感较差问题,采用了图像融合算法首先对纹理图像进行处理,然后再进行纹理映射的方法.实验结果表明,算法具有较好的鲁棒性,生成的三维模型准确真实,较之以往算法具有建模简单、操作方便,真实感强的特点.Aiming at the problem of low efficiency and low photorealistic models in the current virtual simulator, a new algorithm of image-based modeling（IBM） based on scale invariant feature transform （SIFT） was studied. To overcome the shortcoming of manual matching in IBM, a new algorithm based on SIFT was provided. In order to improve the third dimension of 3D reconstruction, texture image extraction and fusion were adopted. Experimental results show that the 3D models generated by this paper are sufficiently accurate and photorealistic. Compare with the previous algorithm of image-based modeling, it is simpler, more convenient and easier to create photorealistic models.

关 键 词：基于图像建模 尺度不变特征变换 层次建模 图像融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]