论文部分内容阅读
在计算机立体视觉研究中，主要针对人造场景，提出了一种新的三维表面重建方法。该方法首先对景物的灰度图像进行分割，将景物图像分成各种不同的区域，利用了将立体像对进行边缘特征匹配的结果，并使用各种曲面模型对各区域进行拟合，以求得整个区域内的视差值，再对整个视差图进行自适应平滑，得到了较为满意的结果。
In the research of computer stereoscopic vision, aiming at man-made scenes, a new method of 3D surface reconstruction is proposed. Firstly, the method divides the grayscale images of the scene, divides the scene images into different regions, and uses the result of edge feature matching of the stereo pair, and uses various surface models to fit each region in order to find Get the parallax value in the whole area, and then adaptively smoothes the entire parallax map to get more satisfactory results.