#image #opencv #multi-threading #rayon #alignment #stack #image-stacking

libstacker

Image alignment and stacking functions based on OpenCV and Rayon

7 releases

0.0.7 Oct 24, 2023
0.0.6 Nov 21, 2021
0.0.3 Oct 31, 2021

#303 in Images

MIT/Apache

26KB
369 lines

crates.io Documentation Workflow dependency status license

libstacker

A multithreaded port of the python code found here: github.com/maitek/image_stacking

This crate contains multithreaded functions that aligns and stacks images using OpenCV and Rayon.

Read more about image alignment with OpenCV here.

Usage:

Download the test images:

git clone https://github.com/maitek/image_stacking image_stacking_py

Build the code:

Opencv-rust can be little tricky to install. Follow the instructions from rust opencv

You will need the "clang-runtime" feature if you experience problems with your clang environment .

cargo build --release

or

cargo build --release --features "clang-runtime"

Run the example:

cargo run --example main --release

or

cargo run --example main --release --features "clang-runtime"

and then wait a few seconds. The program should sort the images by quality, drop the least sharp image, and align and stack the rest. The result should be two windows showing the stacked images using two different alignment methods.

API

let keypoint_match_img:opencv::core::Mat = keypoint_match(
   // a Vec<PathBuf> containing paths to image files
   collect_image_files(&PathBuf::from("image_stacking_py/images"))?,
   KeyPointMatchParameters {
      method: opencv::calib3d::RANSAC,
      ransac_reproj_threshold: 5.0,
   },
)?;

Depending on the parameters the ecc_match() is much slower, but also more accurate.

let ecc_match_img:opencv::core::Mat = ecc_match(
   // a Vec<PathBuf> containing paths to image files
   collect_image_files(&PathBuf::from("image_stacking_py/images"))?,
   EccMatchParameters {
      motion_type: MotionType::Homography,
      max_count: Some(5000),
      epsilon: Some(1e-5),
      gauss_filt_size: 5,
   },
)?;

Todo

  • Figure out the docs.rs problem
  • Figure out what to do with .reshape() in keypoint_match()
  • Figure out some opencv parameters responsibility sneakily shifted to end user.
  • Complete the sharpness_tenengrad() function. Mat not square?
  • Command line options in the example

License

Licensed under either of

at your option.

Dependencies

~3–5.5MB
~76K SLoC