Fig. 1. Sub-region image before and after deformation

Fig. 2. Prediction results of seven models. (a) Porous material; (b) DIC; (c) FlowNet2; (d) PWC-Net; (e) RAFT; (f) GMA; (g) SeparableFlow; (h) GMFlow; (i) FlowFormer

Fig. 3. Performance comparison of seven models. (a) Number of parameters and AEE values; (b) calculation time and AEE values

Fig. 4. Feature map sampling. (a) Reference image feature map; (b) target image feature map; (c) four-dimensional cost volume map

Fig. 5. GPU video memory occupancy with different step sizes

Fig. 6. Displacement field under different sampling steps. (a) (g) Real displacement field; (b) (h) S=1; (c) (i) S=2; (d) (j) S=4; (e) (k) S=8; (f) (l) S=16

Fig. 7. Speckle dataset. (a) Speckle image; (b) displacement field u; (c) displacement field v

Fig. 8. AEE value of speckle images with different batch sizes in test set

Fig. 9. GMA-ICGN algorithm flow

Fig. 10. Calculation results of random deformation displacement field by five algorithms. (a) (h) Deformation image; (b) (i) real displacement field; (c) (j) GMA-ICGN; (d) (k) SIFT-ICGN; (e) (l) FFT-ICGN; (f) (m) Ncorr; (g) (n) DICe

Fig. 11. Initial displacement field values and error along Δ direction of the field. (a) Displacement u and error; (b) displacement v and error

Fig. 12. Displacement field values and error along Δ direction of the field. (a) Displacement u and error; (b) displacement v and error

Fig. 13. Calculation time of three algorithms. (a) Initial guess time; (b) ICGN time

Fig. 14. Deformation experiment of wood compression. (a) Experimental device; (b) image before deformation; (c) image after deformation

Fig. 15. Displacement field of wood compression. (a) Displacement u; (b) displacement v

Fig. 16. Strain field of wood compression. (a) εxx strain field; (b) εyy strain field; (c) γxy strain field