Analysis and Research of a Method of Improving the Accuracy of the Fundamental Matrix

Abstract. The fundamental matrix contains all of a camera’s inner and outer parameters. So estimating the fundamental matrix is an important issue in 3D reconstruction and even the whole field of computer vision. As for the accuracy, stability and universality in calculating the fundamental matrix, this paper gets the coordinates of matched points using SURF algorithm at first, then compares and analyses the results of the two factors affecting the estimation results, such as wrong matching and image noise. Experimental results show that using RANSAC algorithm to remove the wrong matching and the less-noise images can achieve higher accuracy of the fundamental matrix