Recovering 3-D motion parameters from image sequences with gross errors

A robust algorithm to estimate 3-D motion parameters from a sequence of extremely noisy images is developed. The noise model includes correspondence mismatch errors, outliers, uniform noise, and Gaussian noise. More than 100000 controlled experiments were performed. The experimental results show that the error in the estimated 3-D parameters of the linear algorithm almost increases linearly with fraction of outliers. However, the increase for the robust algorithm is much slower, indicating its better performance and stability with data having blunders. The robust algorithm can detect the outliers, mismatching errors and blunders up to 30% of observed data. Therefore, it can be an effective tool in estimating 3-D motion parameters from multiframe time sequence imagery