基于双目视觉的三维人脸重建方法

提出了一种基于双目视觉的三维重建方法,无须昂贵设备和通用模型。从校正后的图像中提取有效人脸重建区域以降低整体耗时。改进了Realtime局部立体匹配算法和基于种子视差约束的区域生长算法,融合Realtime阈值排序和置信度排序进行区域生长,提高了种子像素提取的可靠性和降低区域生长误匹配的可能性。最后,研究了纹理映射技术,提高了重建模型的逼真度。实验结果表明,该方法能够产生逼真光滑的三维人脸模型。

3D face reconstruction using binocular vision

This paper proposed an algorithm to generate 3D face model using binocular vision.It extracted reconstruction region from the rectified images to reduce the system’s consuming.Besides,image matching accomplished based on improved realtime matching algorithm,and a region growing algorithm where realtime-threshold and reliability of seed pixels were sorted to guarantee that the sufficient growing pixels be selected and minimize the erroneous of matching possibility.Finally,a texture mapping method was used to enhance the fidelity of the reconstruction model.Experimental result shows that the proposed method can generate a smooth and vivid 3D face model.