Quality map guided parallel phase unwrapping algorithm for multi-lateral shearing interferometry

Good robustness and fast processing speed are the two main aspects of phase unwrapping algorithm for two-dimensional interferogram. Multi-lateral shearing interferometry requires the use of multidirectional differential wavefronts to reconstruct the wavefront to be measured, which leads to higher requirements on the phase unwrapping speed of the interferogram. The quality map guided algorithm is a reliable algorithm in the phase unwrapping of two-dimensional interferogram. In this paper, the computer unified device architecture (CUDA) programming platform is used to realize the parallel processing of the phase derivative variance map, the maximum phase gradient map, and the pseudo-correlation quality map, in order to improve the speed of phase unwrapping. Parallel calculations are respectively performed on the variance of the wrapped phase gradient in the horizontal and vertical directions in the phase derivative variance map, the wrapped phase gradient in the horizontal and vertical directions of the maximum phase gradient map, and the sine and cosine values of the wrapped phase in the pseudo-correlation quality map. Finally, the CUDA instructions are used to distribute the calculation task and complete the parallel calculation of the quality map. The computer-generated multi-lateral shearing interferograms are used for simulation analysis. The results show that for the same interferogram of 512×512 pixels, the processing speed can be approximately increased by 3 times on the Graphics Processing Unit (GPU) by performing parallel calculation to generate the quality maps. The fastest parallel processing time is 7.34ms for generating the derivative variance map. As the number of pixels increases, GPU processing speed becomes more advantageous, far exceeding that of CPU. Using the self-made quadriwave lateral shearing interferometer, the speed and accuracy of the three quality map guided phase unwrapping algorithms are compared and analyzed. The experimental results show that the derivative variance quality map, which shows the fastest parallel processing speed and the smallest measurement error among the three kinds of quality maps, is more suitable for the parallel phase unwrapping of multi-lateral shearing interferograms.