eduzhai > Applied Sciences > Engineering >

Deep Spatial-angular Regularization for Compressive Light Field Reconstruction over Coded Apertures

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 17 pages

Abstract: Coded aperture is a promising approach for capturing the 4-D light field(LF), in which the 4-D data are compressively modulated into 2-D codedmeasurements that are further decoded by reconstruction algorithms. Thebottleneck lies in the reconstruction algorithms, resulting in rather limitedreconstruction quality. To tackle this challenge, we propose a novellearning-based framework for the reconstruction of high-quality LFs fromacquisitions via learned coded apertures. The proposed method incorporates themeasurement observation into the deep learning framework elegantly to avoidrelying entirely on data-driven priors for LF reconstruction. Specifically, wefirst formulate the compressive LF reconstruction as an inverse problem with animplicit regularization term. Then, we construct the regularization term withan efficient deep spatial-angular convolutional sub-network to comprehensivelyexplore the signal distribution free from the limited representation abilityand inefficiency of deterministic mathematical modeling. Experimental resultsshow that the reconstructed LFs not only achieve much higher PSNR SSIM but alsopreserve the LF parallax structure better, compared with state-of-the-artmethods on both real and synthetic LF benchmarks. In addition, experiments showthat our method is efficient and robust to noise, which is an essentialadvantage for a real camera system. The code is publicly available at url{this https URL}

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×