Image processing for pattern diffracted by phase plate

Optical measurement systems and instruments with diffractive optics have been successfully applied in various fields, particularly for the tasks, such as measuring geometries of large-scale workpieces, profiling large aspheres used in synchrotron radiation facility, etc. In this paper efforts to further improve the measurement resolution and accuracy of these optical measuring techniques are presented, in which principles of the optical systems employing typical diffractive optics as π-jump phase plate are, in detail, described, and characteristics of the image patterns created within these systems are analyzed. Finally a novel diffraction pattern processing method based on multi-layer feedforward neural networks is proposed, which is proved by digital simulation to be better than the conventional method based on the Least-squares curve fitting. Preliminary experimental setups have been built up to verify the feasibility and effectiveness of this novel method, and the results coincide with simulation very well.