AutoStitch:Panoramic Image Stitching

Introduction

This repository demonstrates the implementation of AutoStitch in Python, based on the traditional image stitching pipeline using SIFT + RANSAC for robust matching. We provide a variety of configurable parameters for key stages including feature extraction , matching , seam estimation , and image blending .

Our method achieves state-of-the-art (SOTA) performance when compared to traditional / deep learning-based methods. users can input all images captured from a single scene, and automatically return seamless panoramic image.

On an AMD 9950X3D CPU , stitching 12 frames of 2K resolution images takes less than 3 seconds . If you have access to a GPU, enable CUDA acceleration by setting the parameter try_cuda=True for improved performance.

File Structure

AutoStitch
├── images               # Directory containing input images for stitching
├── Stitcher             # Panorama stitching implemented in Python
│   ├── __init__.py      # Marks the directory as a Python package
│   ├── Constant.py      # Constants used throughout the project
│   └── main.py          # Main script for stitching images
├── Stitcher++           # Panorama stitching implemented in C++
│   ├── README.md        # Documentation specific to the C++ implementation
│   └── Stitcher.cpp     # C++ source file for stitching logic
├── README.md            # Project documentation
├── LICENSE              # License information for the project
└── requirements.txt      # Lists Python dependencies required for the project

Installation

To run the AutoStitch project, you need to have Python installed on your machine. Additionally, you will need to install the required dependencies listed in requirements.txt.

1. Clone the repository:

   git clone https://github.com/KiMomota/AutoStitch.git
   cd AutoStitch
   

2. Install the required packages:

   pip install -r requirements.txt
   

Usage

To use the AutoStitch functionality, modify the path to the images you want to stitch. Then run the main.py script with the path to the directory containing the images you want to stitch:

python Stitcher/main.py

WebUI

We have provided a GUI for users to stitch images. Users can automatically adjust the parameters, and after selecting the input images, the image will display in panorama.

FurtherMore

we will release end-to-end model for image stitching. The code will be available soon.
You are welcome to contact me via email for any questions or suggestions or issues.
If you find this project helpful, please consider giving it a star ⭐ on GitHub.