Fig. 1. Layered expression for edge disparity in all EPIs according disparity

Fig. 2. Foreground depth propagation image of camera view

Fig. 3. Foreground labeled image of camera view

Fig. 4. Camera view images after removing foreground occlusion

Fig. 5. Relationship between disparity and view shifted in EPI

Fig. 6. Camera view images on the specific depth plane after removing occlusion

Fig. 7. Algorithm workflow

Fig. 8. Reconstructed results of outdoor occluded objects with the traditional synthetic aperture imaging and our algorithm. (a) Image of camera view in the Disney dataset; (b) an foreground labeled image of camera view, the foreground disparity range is about [10.5,13.2]; (c) reconstructed images with the traditional synthetic aperture imaging, focusing on the red frame areas, respectively; (d) reconstructed images with our algorithm, focusing on the same red frame areas, respectively, and the disparit

Fig. 9. (a) Foreground depth propagation image of camera view; (b) foreground labeled image of camera view

Fig. 10. Images after removing foreground plants from scene

Fig. 11. Details of reconstructed CD cases behind the plant with different algorithms. (a1) The traditional synthetic aperture imaging, focusing on the left CD case; (a2) the traditional synthetic aperture imaging, focusing on the right CD case; (b1) our algorithm, focusing on the left CD case; (b2) our algorithm, focusing on the right CD case; (c1) the algorithm in ref.[12], focusing on the left CD case; (c2) the algorithm in ref.[12], focusing on the right CD case

Fig. 12. PSNR of reconstructed CD cases images with different algorithms