Batch skeleton extraction from ESPI fringe patterns using pix2pix conditional generative adversarial network

The key to measurement by electronic speckle pattern interferometry (ESPI) is to obtain accurate phase information from the ESPI fringe patterns. We propose a fast batch skeleton extraction method for ESPI fringe patterns using the pix2pix conditional generative adversarial network (pix2pix cGAN). The network is trained by ESPI fringe patterns and complete skeleton images, and the trained network can quickly extract skeletons; it took 11.7 s to extract the skeletons of 200 experimental ESPI fringe patterns. Compared to the fringe skeleton method, cycle GAN method, and U-net method, our method can obtain accurate, complete, and smooth skeletons faster. In addition, for some broken ESPI fringe patterns, the traditional fringe skeleton method will fail, whereas complete skeletons can be obtained through the trained network.