• Acta Optica Sinica
  • Vol. 42, Issue 16, 1612003 (2022)
Tianyi Liu, Jun Qiu, Di He*, and Chang Liu
Author Affiliations
  • Institute of Applied Mathematics, Beijing Information Science and Technology University, Beijing 100101, China
  • show less
    DOI: 10.3788/AOS202242.1612003 Cite this Article Set citation alerts
    Tianyi Liu, Jun Qiu, Di He, Chang Liu. Light Field alpha Matting Based on Spatial-Angular Consistency[J]. Acta Optica Sinica, 2022, 42(16): 1612003 Copy Citation Text show less
    Schematic diagram of mapping relationship of object points in different viewpoints
    Fig. 1. Schematic diagram of mapping relationship of object points in different viewpoints
    Sparse representation of light field. (a) Central sub-aperture image and disparity map; (b) sub-aperture images
    Fig. 2. Sparse representation of light field. (a) Central sub-aperture image and disparity map; (b) sub-aperture images
    Sub-aperture images and EPI. (a) Sub-aperture images; (b) EPI
    Fig. 3. Sub-aperture images and EPI. (a) Sub-aperture images; (b) EPI
    EPI of foreground alpha images
    Fig. 4. EPI of foreground alpha images
    Disparity map. (a) Disparity map of platonic scene; (b) disparity map of antinous scene
    Fig. 5. Disparity map. (a) Disparity map of platonic scene; (b) disparity map of antinous scene
    Comparison of alpha images in different scenes. (a) Central sub-aperture alpha image; (b) alpha image of propagation algorithm at u=1,v=1; (c) alpha image of each sub-aperture matting algorithm at u=1,v=1
    Fig. 6. Comparison of alpha images in different scenes. (a) Central sub-aperture alpha image; (b) alpha image of propagation algorithm at u=1,v=1; (c) alpha image of each sub-aperture matting algorithm at u=1,v=1
    Comparison of alpha matting results in different scenes. (a) Central sub-aperture alpha matting result; (b) alpha matting result of propagation algorithm at u=1,v=1; (c) alpha matting result of each sub-aperture matting algorithm at u=1,v=1
    Fig. 7. Comparison of alpha matting results in different scenes. (a) Central sub-aperture alpha matting result; (b) alpha matting result of propagation algorithm at u=1,v=1; (c) alpha matting result of each sub-aperture matting algorithm at u=1,v=1
    EPIs of different scenes. (a) EPI obtained by propagation algorithm in platonic scene when v=1,y=92; (b) EPI obtained by each sub-aperture matting algorithm in platonic scene when v=1,y=92; (c) EPI obtained by propagation algorithm in antinous scene when v=1,y=445; (d) EPI obtained by each sub-aperture matting algorithm in antinous scene when v=1,y=445
    Fig. 8. EPIs of different scenes. (a) EPI obtained by propagation algorithm in platonic scene when v=1,y=92; (b) EPI obtained by each sub-aperture matting algorithm in platonic scene when v=1,y=92; (c) EPI obtained by propagation algorithm in antinous scene when v=1,y=445; (d) EPI obtained by each sub-aperture matting algorithm in antinous scene when v=1,y=445
    Central sub-aperture alpha image and disparity map of real light field data. (a) Central sub-aperture alpha image; (b) disparity map
    Fig. 9. Central sub-aperture alpha image and disparity map of real light field data. (a) Central sub-aperture alpha image; (b) disparity map
    alpha images and alpha matting results of real light field data obtained by different algorithms. (a1)(a2) Central sub-aperture images; (b1)(b2) propagation algorithm; (c1)(c2) each sub-aperture matting algorithm
    Fig. 10. alpha images and alpha matting results of real light field data obtained by different algorithms. (a1)(a2) Central sub-aperture images; (b1)(b2) propagation algorithm; (c1)(c2) each sub-aperture matting algorithm
    EPIs of real light field data alpha matting. (a) EPI obtained by propagation algorithm in platonic scene when v=1,y=390; (b) EPI obtained by each sub-aperture matting algorithm in platonic scene when v=1,y=390; (c) EPI obtained by propagation algorithm in antinous scene when v=1,y=435; (d) EPI obtained by each sub-aperture matting algorithm in antinous scene when v=1,y=435
    Fig. 11. EPIs of real light field data alpha matting. (a) EPI obtained by propagation algorithm in platonic scene when v=1,y=390; (b) EPI obtained by each sub-aperture matting algorithm in platonic scene when v=1,y=390; (c) EPI obtained by propagation algorithm in antinous scene when v=1,y=435; (d) EPI obtained by each sub-aperture matting algorithm in antinous scene when v=1,y=435
    Comparison of real light field data fusion (u=1,v=1). (a) Result obtained by propagation algorithm; (b) result obtained by each sub-aperture matting algorithm; (c) partial enlarged image of Fig. 12(a); (d) partial enlarged image of real light field data; (e) partial enlarged image of Fig. 12(b)
    Fig. 12. Comparison of real light field data fusion (u=1,v=1). (a) Result obtained by propagation algorithm; (b) result obtained by each sub-aperture matting algorithm; (c) partial enlarged image of Fig. 12(a); (d) partial enlarged image of real light field data; (e) partial enlarged image of Fig. 12(b)
    EPI of fusing light field data (v=1,y=598). (a) EPI obtained by propagation algorithm; (b) EPI of real light field data; (c) EPI obtained by each sub-aperture matting algorithm
    Fig. 13. EPI of fusing light field data (v=1,y=598). (a) EPI obtained by propagation algorithm; (b) EPI of real light field data; (c) EPI obtained by each sub-aperture matting algorithm
    Comparison of new generated viewpoint of real light field data (u=1,v=1.5). (a) New viewpoint generated by propagation algorithm; (b) new viewpoint generated by each sub-aperture matting algorithm
    Fig. 14. Comparison of new generated viewpoint of real light field data (u=1,v=1.5). (a) New viewpoint generated by propagation algorithm; (b) new viewpoint generated by each sub-aperture matting algorithm
    EPIs of new viewpoint images of real light field data. (a) EPI of new viewpoint image obtained by propagation algorithm when v=1,y=71; (b) EPI of new viewpoint image obtained by each sub-aperture matting algorithm when v=1,y=71; (c) EPI of new viewpoint image obtained by propagation algorithm when v=1,y=528; (d) EPI of new viewpoint image obtained by each sub-aperture matting algorithm when v=1,y=528
    Fig. 15. EPIs of new viewpoint images of real light field data. (a) EPI of new viewpoint image obtained by propagation algorithm when v=1,y=71; (b) EPI of new viewpoint image obtained by each sub-aperture matting algorithm when v=1,y=71; (c) EPI of new viewpoint image obtained by propagation algorithm when v=1,y=528; (d) EPI of new viewpoint image obtained by each sub-aperture matting algorithm when v=1,y=528
    ScenePropagation algorithmEach sub-aperture matting algorithm
    Platonic0.270.44
    Antinous0.380.65
    Table 1. Consistency evaluation of synthetic light field alpha images
    ScenePropagation methodEach sub-aperture matting method
    Real light field data0.380.48
    Table 2. Consistency evaluation of real light field alpha images
    Tianyi Liu, Jun Qiu, Di He, Chang Liu. Light Field alpha Matting Based on Spatial-Angular Consistency[J]. Acta Optica Sinica, 2022, 42(16): 1612003
    Download Citation