Fig. 1. CycleGAN fringe pattern filtering model and loss model. (a) Fringe pattern filtering model; (b) forward cycle-consistency loss; (c) backward cycle-consistency loss

Fig. 2. Structure of fringe pattern filtering model based on CycleGAN

Fig. 3. Schematic of ideal noiseless phase map generation steps

Fig. 4. Loss change curve when the learning rate is 0.0001

Fig. 5. Model effects obtained under different number of iterations

Fig. 6. Loss function curves under different learning rates

Fig. 7. Filtering results of different methods for simulated noise fringe patterns

Fig. 8. Original noise fringe images and filtering results of different methods

Fig. 9. Filtering application optimization process. (a) Cropping from the original image; (b) cropping from filtered image; (c) image mosaic

Fig. 10. Filtering results for fringe pattern under different steps