Optimal Depth Estimation from a Single Image by Computational Imaging Using Chromatic Aberrations (bibtex)
by Muhammad Atif, Bernd Jähne
Abstract:
We present a computational imaging approach to estimate the depth from a single image using axial chromatic aberrations. It includes a co-design of optics and digital processing to select the optimal parameters of a lens such as focal length, f-number, and chromatic focal shift according to the performance of a depth estimation algorithm on the digital side. A simulation framework evaluates the complete systems performance in different imaging conditions including optimal axial chromatic lens aberration. A low-complexity algorithm estimates the depth map of real scenes. Experiments on real and synthetic scenes show the feasibility of the proposed system for depth estimation. In the case of relatively broadband object spectra and a lens with focal length of 4 mm, depth is estimated with an RMS error of 6–10%.
Reference:
Optimal Depth Estimation from a Single Image by Computational Imaging Using Chromatic Aberrations (Muhammad Atif, Bernd Jähne), In tm --- Technisches Messen, volume 80, 2013.
Bibtex Entry:
@string{tm="tm --- Technisches Messen"}
@Article{atif2013,
  Title                    = {Optimal Depth Estimation from a Single Image by Computational Imaging Using Chromatic Aberrations},
  Author                   = {Muhammad Atif and Bernd Jähne},
  Journal                  = tm,
  Year                     = {2013},
  Pages                    = {343--348},
  Volume                   = {80},

  Abstract                 = {We present a computational imaging approach to estimate the depth from a single image using axial chromatic aberrations. It includes a co-design of optics and digital processing to select the optimal parameters of a lens such as focal length, f-number, and chromatic focal shift according to the performance of a depth estimation algorithm on the digital side. A simulation framework evaluates the complete systems performance in different imaging conditions including optimal axial chromatic lens aberration. A low-complexity algorithm estimates the depth map of real scenes. Experiments on real and synthetic scenes show the feasibility of the proposed system for depth estimation. In the case of relatively broadband object spectra and a lens with focal length of 4 mm, depth is estimated with an RMS error of 6–10%.},
  Access                   = {1},
  Doi                      = {10.1515/teme.2013.0042},
  Groupid                  = {hci},
  Owner                    = {bjaehne},
  Timestamp                = {2014.06.14}
}