OpenCV Core Stitching Pipeline

MATLAB example based on OpenCV's stitching module, inspired by the source code of stitching_detailed.cpp.

Overview

Welcome to this MATLAB-based project that replicates the core functionalities of the OpenCV Stitching module, specifically tailored for rotation-only homography scenarios. This repository is designed to help you understand and visualize the intricate details behind image stitching, from feature detection to the final panorama projection.

Features

• Core Stitching Functionality:

  • Feature Point Detection and Matching: Identify and match key points across images to prepare for stitching.
  • Homography Estimation and Decomposition: Estimate the homography matrix and decompose it to obtain camera parameters (K, R).
  • Bundle Adjustment (BA) Optimization: Optimize camera parameters for better alignment and stitching accuracy.

• Image Correction:

  • Wave Correction: Correct any waveform distortions in the images, supporting both horizontal and vertical corrections to ensure accurate stitching.

• Panorama Projection:
  Supports multiple projection types to cater to different stitching needs:

  • Plane
  • Spherical
  • Cylindrical
  • Fisheye
  • Stereo

• Multi-Band-Blender Fusion:

  • Multi-band Blending: Seamlessly blend images using multi-band blending techniques to reduce visible seams.

• Visualization Tools:

  • Intermediate Step Visualization: Visualize each step of the stitching process to gain deeper insights into the underlying techniques.

Usage

Clone the repository:

git clone https://github.com/cuixing158/OpenCV-Core-Stitching.git
cd OpenCV-Core-Stitching

Open MATLAB and navigate to the project directory.Run the provided example script main_opencv_stitching_demo.mlx to see the stitching process in action.

Explore the code and modify parameters to see how different settings affect the stitching outcome.

Requirements

MathWorks Products(<www.mathworks.com>)

• MATLAB R2024b or later
• Computer Vision Toolbox™
• Image Processing Toolbox™
• Optimization Toolbox™

Contributing

If you have ideas for improvements or new features, please feel free to open an issue or submit a pull request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Inspired by the OpenCV Stitching module.
• Special thanks to the MATLAB community for their support and resources.

Contact & Feedback

If you have any suggestions about this project, feel free to contact me.

[e-mail: cuixingxing150[at]gmail.com]

Happy stitching! 😄😄😄

References

[1] https://pages.cs.wisc.edu/~dyer/cs534/papers/szeliski-alignment-tutorial.pdf

[2] Matthew Brown and David G Lowe. Automatic panoramic image stitching using invariant features. International journal of computer vision, 74(1):59–73, 2007.