Retrieval of Mueller matrices from intensity measurements is a noise-sensitive process. In addition, the choice of the method used for extracting Mueller matrix elements greatly influences the precision of the final results. Among available procedures, three have been tested and their robustness analyzed by adding Gaussian noise to computer synthesized data and have been verified by comparing experimental data. As expected, the methods based on classical matrix inversion reach their noise reduction limit even if more data are considered. In contrast, the discrete Fourier transform approach features a remarkable stability. The identified reason is that the retrieval process corresponds to filtering with a periodic kernel.
© 2006 Optical Society of America
Equations on this page are rendered with MathJax. Learn more.