Abstract

Epipolar geometry is a key point in computer vision and the fundamental matrix estimation. The relational view can be obtained from the fundamental matrix. In this way, we interest to calculate an exact matrix based from characteristics unequally distributed in complex scene images. This paper presents a method based on the detection of points by the Harris detector, after we develop a new modification of the multi-level function related to the M-estimator algorithm. The experimental comparisons were conducted by a simulation between RANSAC, LMeds, and M-Estimator in order to estimate the projection error. As a result, the proposed method gives a significant improvement and performance with a low projection error compared to other methods.