Fig. 1. Overall framework of unsupervised large field of view light field content generation method based on depth perception

Fig. 2. Proposed center alignment module

Fig. 3. Schematic diagram of the structure of correlation layer and regression network module

Fig. 4. Schematic diagram of image fusion module

Fig. 5. Visual comparison results of different methods on real scene datasets. (a) Input images; (b) APAP^[8]; (c) ELA^[9]; (d) SPW^[10]; (e) UDIS^[18]; (f) UDIS2^[19]; (g) HOSVD^[24]; (h) proposed method

Fig. 6. Visual comparison results of different methods in the distant and near scenes. (a) Input images; (b) APAP^[8]; (c) ELA^[9]; (d) SPW^[10]; (e) UDIS^[18]; (f) UDIS2^[19]; (g) HOSVD^[24]; (h) proposed method

Fig. 7. Visual comparison results of EPI in large field of view light field images stitched by different methods

Fig. 8. Spatial quality comparison of ablation for the center alignment module. (a) Input images; (b) proposed method; (c) without center alignment

Fig. 9. Comparison of angular quality of ablation for center alignment module under different scenes. (a) Scene 1; (b) scene 2; (c) scene 3

Fig. 10. Visual results of ablation experiment on depth information. (a) Input images; (b) no depth information is introduced; (c) proposed method