Image Stitching

Although there are multiple implementations of image stitching using different libraries (OpenCV, skimage, ...) this repository contains image stitching using only numpy for computations and cv2 for displaying images. Both projective transformation fitting and RANSAC are implemented using numpy, but the undistortion is done using cv2 library.

Repository Contents

There are 2 example photos already present in distorted and undistorted form: examples/distorted000.png, examples/undistorted000.png, examples/distorted001.png, examples/undistorted001.png:

Example 000:

Example 001:

If you wish to take your own photos using ESP32-CAM you can capture them using image_capture.py script. You can take a screenshot with the space bar. The repository also contains examples/matching_points_000_001.txt - 5 pairs of matching keypoints between images examples/undistorted000.png and examples/undistorted001.png picked by hand.

Contents of examples/matching_points_000_001.txt:

295 264 25 272
295 368 27 385
335 266 72 271
336 369 76 381
507 265 251 266

How To Run The Code

To setup the environment it is recommended to create a virtual environment in this directory:

python -m venv venv

Activate the virtual environment and install required dependencies:

. venv/bin/activate
pip install -r requirements.txt

There are 2 main scripts that can be executed:

1. Run python stitch_using_points.py <path_to_img1> <path_to_img2> <path_to_points_file> to generate a stitched image from 2 given images using hand-picked matching keypoints. You can simply run python stitch_using_points.py to use example images. Sample panorama generated by this script examples/panorama_task_5.png:

2. Run python stitch.py <path_to_img1> <path_to_img2> to generate a stitched image from 2 given images using automatically picked matching keypoints. You can simply run python stitch.py to use example images. Sample panorama generated by this script examples/panorama_task_7.png:

Running Tests

Execute:

pytest .