Skip to the content.

Poisson Disk Sampling with Randomized Satellite Points for Projected Texture Stereo

Project page of the paper “Poisson Disk Sampling with Randomized Satellite Points for Projected Texture Stereo”.

top

Paper

Citation

J. Takeda and N. Fukushima, “Poisson Disk Sampling with Randomized Satellite Points for Projected Texture Stereo,” Optics Continuum, vol.1, no.5, pp.974-988, May 2022.

BibTeX

@article{takeda2022poisson,
author = {Jun Takeda and Norishige Fukushima},
title = {Poisson Disk Sampling with Randomized Satellite Points for Projected Texture Stereo},
journal = {Opt. Continuum},
volume = {1},
number = {5},
pages = {974--988},
month = {May},
year = {2022},
publisher = {OSA},
doi = {10.1364/OPTCON.451197},
}

Compile

The codes are written in C++. The library is only tested in Windows with Visual Studio.

The dependency is OpenCV and our library of OpenCP. For compiling the OpenCP library, Qt maybe required. Pre-compiled OpenCP libraries are attached in OpenCP dir.

The project use the following system environment variable as Macro; please set your environment variables according to your environment or write the path directly in the sln project file. The examples are as follows;

OPENCV_INCLUDE_DIR = "C:\OpenCV\include"
OPENCP_INCLUDE_DIR = "C:\OpenCP\include"
OPENCV_LIB_DIR = "C:\OpenCV\lib"
OPENCP_LIB_DIR = "C:\OpenCP\lib"

Also, OpenCP.dll must be located in the active pass. For example,

PATH = "C:\OpenCP\bin"

Usage

In main.cpp, we can switch the following functions. The main simulation code is stereoTest.

stereoTest();//test for stereo matching
pointcloudTest();//test for point cloud rendering for Fig. 12
//generatePDSMask(Size(1400, 1120), 2.f, 10);// function to prepare PDS masks for stereo test.

For stereoTest, we can change the following conditions;

The processing flow is as follows;

  1. load stereo images and disparity maps,
  2. compute min/max search range
  3. generate the center disparity map for projection
  4. add noise (optional)
  5. call the method of gui in class StereoMatch, which is the main stereo matching function.

The overview of the function call is shown in the following figure. We have 7 windows; disparity map, projection, parameters, cost function, profile, console information, and point cloud (optional). window

Each window and keyboard shortcut can change the parameters.

Patent

The matching method of the skipped census transform and the projection method of Poisson disk sampling with randomized satellite points are patent pending.

Note

Link