Skip to content

Variants of the classic Lucas-Kanade image alignment algorithm

Notifications You must be signed in to change notification settings

cheind/image-align

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

79 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

About this library

Image Align is a C++ library providing variants of the classic image alignment algorithm by Lucas-Kanade.

Image Align under Euclidean Motion

The project emerged while working on AAM, an active appearance models library. Fitting active appearance models is similar to the classic image alignment problem:

The goal of image alignment is to find the locally 'best' transform between a template image and a target image by minimizing an energy function measuring the fitness of the alignment. -- Ian Matthews

Algorithms and Features

All image alignment algorithms implemented in this library are based on the original formulation of Lucas-Kanade:

  • Forward additive algorithm
  • Forward compositional algorithm
  • Inverse compositional algorithm

For convergence and runtime reasons all algorithms support multi-level hierarchical matching.

The alignment algorithms are independent of the chosen warp function. Currently the library provides the following warp modes:

  • 2D Translational Warp
  • 2D Euclidean Warp
  • 2D Similarity Warp
  • 2D Affine Warp

User defined warp functions can be easily added.

Usage

Image Align is quite simple to use. Start by including the necessary headers

#include <imagealign/imagealign.h>

Next, declare the type / precision of warp, and the alignment variant you wish to use

namespace ia = imagealign;

// Use a double precision warp describing a similarity motion
// (rotation, translation and uniform scale).
typedef ia::WarpSimilarityD WarpType;

// Use Inverse Compositional algorithm for image alignment.
typedef ia::AlignInverseCompositional<WarpType> AlignType;

Given a template image and a target image you can now perform alignment

cv::Mat tpl;    // The template image
cv::Mat target; // The target image

namespace ia = imagealign;

// Instance necessary objects
WarpType w;
AlignType a;

// Prepare for alignment using 3 levels of hierarchy
a.prepare(tpl, target, w, 3);

// Perform iterative alignment over all levels in hierarchy.
a.align(w, 30, 0.003);

When alignment has finished, w will hold the warp that best aligns the template image with the target image.

Please note, Lucas-Kanade methods are locally operating methods that require a good guess of true warp parameters to converge. To provide a guess, simple adjust the parameters of w using methods such as w.setParameters() and similar before calling a.align().

Image Align comes with a couple of examples that illustrate further usage. you can find these in the examples directory. Additionally these unit tests might provide in-depth information.

Building from source

Image Alignment requires the following pre-requisites

  • CMake - for generating cross platform build files
  • OpenCV 2.x / 3.x - for image processing related functions

To build from source

  1. Point CMake to the cloned git repository
  2. Click CMake Configure
  3. Point OpenCV_DIR to the directory containing the file OpenCVConfig.cmake
  4. Activate / Deactivate IMAGEALIGN_USE_OPENMP
  5. Click CMake Generate

Although Image Alignment should build across multiple platforms and architectures, tests are carried out on these systems

  • Windows 8/10 MSVC10 / MSVC12 x64
  • OS X 10.10 XCode 7.x x64

If the build should fail for a specific platform, don't hesitate to create an issue.

References

  1. Lucas, Bruce D., and Takeo Kanade. "An iterative image registration technique with an application to stereo vision." IJCAI. Vol. 81. 1981.
  2. Baker, Simon, and Iain Matthews. "Equivalence and efficiency of image alignment algorithms." Computer Vision and Pattern Recognition, 2001. CVPR 2001. Proceedings of the 2001 IEEE Computer Society Conference on. Vol. 1. IEEE, 2001.
  3. Baker, Simon, and Iain Matthews. Lucas-Kanade 20 years on: A unifying framework: Part 1. Technical Report CMU-RI-TR-02-16, Carnegie Mellon University Robotics Institute, 2002.
  4. Baker, Simon, and Iain Matthews. Lucas-Kanade 20 years on: A unifying framework: Part 2. Technical Report CMU-RI-TR-03-01, Carnegie Mellon University Robotics Institute, 2003.
  5. Baker, Simon, et al. "Lucas-Kanade 20 years on: A unifying framework: Part 3." The Robotics Institute, Carnegie Mellon University (2003).

License

This file is part of Image Alignment.

Copyright Christoph Heindl 2015

Image Alignment is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.

Image Alignment is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with Image Alignment.  If not, see <http://www.gnu.org/licenses/>.

About

Variants of the classic Lucas-Kanade image alignment algorithm

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages