Direct Self-rectification

This repo is the MATLAB implementation code for the paper "DSR: Direct Self-rectification for Uncalibrated Dual-lens Cameras" on 3DV 2018.

Acknowledgement

• The statistics reported in the paper is implemented by C++ and opencv, which may not be consistent with the one in matlab.
• Unlike in the paper, we use SIFT feature matcher implemented by Professor A. Fusiello, Some modification is made to support MATLAB 2015B. We thank Professor Fusiello for his code.
• If you have any further question regarding to the paper, you may contact me at xiaoruichao at gmail.com.

If you find our algorithm helpful in your work, please consider cite:

@article{3DV2018DSRxiao,
  title={DSR: Direct Self-rectification for Uncalibrated Dual-lens Cameras},
  author={Xiao, Ruichao and Sun, Wenxiu and Pang, Jiahao and Yan, Qiong and Ren, Jimmy},
  journal={3DV},
  year={2018}
}

Dependency

• MATLAB 2015B on or above (The code is tested on MATLAB2015B);

• VLFeat 0.9.21 released, download avaliabe at Download

Usage

1. Download the VLFeat toolkit and extract the directory vlfeat-0.9.21/ into self-rectification/ directory;
2. To run the self-rectification, execute run_rectify.m.

Example results

some image pairs are provided for test purpose.

master image(left) and right image(right)

After self-rectification, The result is

Rectified master image(left) and slave image(right), notice that our algorithm makes no transfomration on master image.