基于柱面反投影算法的三维物体表面纹理重建

从图像中提取纹理用作纹理映射是三维重建的常用方法之一，然而通过相机获取的照片纹理经常存在失真变形．不能直接用于纹理映射，故需要研究失真纹理的校正针对圆柱面物体提出一种柱面反投影算法，将物体在圆柱面上的影像纹理投影到二维视平面上显示，并用线性插值算法对缺损的像素值进行插值校正，然后利用SIFT（Scale Invariant Feature Transform）尺度不变特征点匹配算法对校正后的纹理图像进行特征点提取与匹配，最后采用加权平滑算法实现图像的拼接，从而得到宽视角的纹理图像．实验表明，利用上述方法能够有效地生成三维重建所需的纹理．

Surface Texture Reconstruction for 3D Object Based on Cylindrical Back-Projection Algorithm

Extracted texture from the image used for texture mapping is a commonly used method of 3D reconstruction. However,the texture obtained by camera will usually distortion,can not be directly used for texture mapping,so study the texture correction is necessary. An cylindrical back-projection algorithm was proposed for cylindrical objects,image texture on the cylindrical surface was projected onto 2D plane, and the defect pixels were corrected with linear interpolation algorithm,then SIFT algorithm was used for feature extraction and matching on corrected texture image. Finally, the seamless automatic image mosaic was completed with smoothing algorithm, wide viewing angle image was got. Experiments show that these methods can effectively generate the texture for 3D reconstruction.