投影三维重建中基础矩阵的鲁棒性估计方法

A New Robust Method for Computing the Fundamental Matrix

在未定标系统中 ,基础矩阵给出了图像间对极几何关系的代数描述 ,解决了许多视觉问题的关键环节 ,本文提出了一种新的鲁棒性基础矩阵估计方法 ,它引入了能够较好表达代数余数标准方差的Sampson加权算子 ,并且用迭代法来满足图像点噪声为非高斯白噪声的情况 .实验结果表明 ,在较大噪声干扰的条件 ,仍能较为准确地估计基础矩阵 ,具有良好的鲁棒性和较快的运算速度 .

Two perspective images of single scene taken by uncalibrated perspective cameras are related by the fundamental matrix, which is the key to many problems of computer vision. This paper addresses the problem of computing the fundamental matrix using a new method. Sampson weighting, which represents the sum of square of the algebraic residuals divided by their standard deviations, is adapted to improve the accuracy of the method, and the iterative method is called for the condition that the noise is not Gaussianly distributed. Experiments show that the method is robust and precise even if there is great noise.