Skip to content

A Julia project demonstrating the fast f-k migration algorithm.

License

Notifications You must be signed in to change notification settings

analytech-solutions/fkMigration.jl

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

fkMigration.jl

A Julia project demonstrating the fast f-k migration algorithm presented in Wave-based non-line-of-sight imaging using fast f−k migration by Lindell et al. at SIGGRAPH 2019.

Setup

If you are new to non-line-of-sight (NLOS) imaging, we have written an introductory blog post to provide some background and explain how we wrote this code.

First, clone the repository and run Julia from it.

cd fkMigration.jl
julia --project

Then, enter package mode ] and instantiate the project.

(fkMigration) pkg> instantiate
Project fkMigration v0.1.0
    Status `fkMigration.jl/Project.toml`
  [7a1cc6ca] + FFTW v0.3.0
  [23992714] + MAT v0.5.0
    Status `fkMigration.jl/Manifest.toml`
  [7a1cc6ca] + FFTW v0.3.0
  [23992714] + MAT v0.5.0

You will need to download and extract NLOS datasets from the Stanford Computational Imaging Lab in order to continue. We used the "teaser" dataset, and have it extracted to ../teaser. You will need to use those directory paths in calls to the fkMigration.jl project.

vertical-temporal waves vertical-horizontal waves

Usage

There is a simple demo function which you can run, but it requires a lot of system memory (>=32GB) to run. A couple of optional arguments can be provided to downsample and crop the data to reduce the memory usage. Either way, the function returns a dense array 3D volumetric represention of the scene. Therefore, the array must be collapsed in order to form a more traditional 2D image of the scene, and we simply use maximum as a way to achieve that below.

julia> using fkMigration

julia> fullVolume = demo("../teaser")
512×512×1024 Array{Float64,3}:


julia> lowResVolume = demo("../teaser", 64, 512)
64×64×512 Array{Float64,3}:


julia> lowResImage = maximum(lowResVolume, dims=3)[:, :] / maximum(lowResVolume)
64×64 Array{Float64,2}:

You now have a normalized array which you can further manipulate, view with the ImageView.jl package, or save with the FileIO.jl and ImageMagick.jl packages.

reconstructed scene

Also, if you want more control over the whole process, the lower-level functions used by demo are available to use for yourself.

julia> tau, calib = loadDataset("../teaser", "meas_10min.mat") ;

julia> calibrate!(tau, calib) ;

julia> tau = downsampleAndCrop(tau, 64, 512) ;

julia> tau = reconstruct(tau) ;

Releases

No releases published

Sponsor this project

 

Packages

No packages published

Languages