Skip to content

Latest commit

 

History

History
63 lines (45 loc) · 2.95 KB

README.md

File metadata and controls

63 lines (45 loc) · 2.95 KB
Raw

flatcamRBG-D

This repository is the implementation of our paper: Joint Image and Depth Estimation with Mask-Based Lensless Cameras.

Please cite the following paper when using this code or data:

@article{zheng2020joint,
  title={Joint image and depth estimation with mask-based lensless cameras},
  author={Zheng, Yucheng and Asif, M. Salman},
  journal={IEEE Transactions on Computational Imaging},
  volume={6},
  pages={1167--1178},
  year={2020},
  publisher={IEEE}
}

Abstract

image

In this project, we propose an alternating gradient descent algorithm that jointly estimates a continuousdepth map and light distribution of the unknown scene from itslensless measurements. We built a prototype lensless camera and present experimental results for reconstruction of intensity and depth maps of different real scenes.

Run the code

Run demo_depth.m file in Matlab.

Expected results

The algorithm has 3 main steps, here we list the expected results in each step.

Step 1: Initialization

image

In this step, we try to reconstruct the scene at 10 predefined depth planes separately, then we pick the one that has smallest measurement error as our initialization.

Step 2: Optimization

image

In this step, we jointly estimate image and depth. At every iteration, we first estimate the depth value of each pixel by applying gradient descent with the current estimated image, and then estimate the image intensity based on the estimated depth.

Step 3: Final results

image

In this step, we remove the depth values that belong to pixels with low image intensities, because low-intensity pixels mostly provide inaccurate depth estimates.

Parameters

Four example data files are included in this repository: a single card scene, two cards scene, a cup scene, and a hand sculpture scene. The algorithm iteratively solves for image and depth map. Here we list some parameters that you can play with in the code.

  • dataset: data files to be evaluated.
  • niter: number of iterations in the loop.
  • damp: l2 regularization parameter in the image solver.
  • itnlim: number of iterations in the image solver.
  • maxIter: number of iterations in the depth solver.
  • lambda: TVL2 regularization parameter in the depth solver.

Packages

We include the packages we used in the /code/utils/ folder. Here we list these packages and their sources.

RBG: May her memory be a blessing