Overview
--------

This package contains a library and testing suite for motion detection, 
extrapolation and morphing algorithms. 

The motion detection algorithms included in this package are:
  * Lucas&Kanade local least-squares algorithm
  * global anisotropic diffusion algorithm by Proesmans et. al.
  * OpenCV implementation of the Lucas&Kanade algorithm

The first two algorithms are implemented by the author of this package. 
They generate dense vector fields, i.e. motion vectors for 
each pixel in the source image. On the other hand, the OpenCV 
implementation generates sparse motion fields. It extracts 
the motion of a given set of "feature points". All algorithms 
in this package use image pyramids.

This package also contains algorithms for image warping by using 
a computed motion field. The following methods are implemented:
  * Inverse pixel-based extrapolation: For each pixel in the second source 
    image, find the corresponding pixel in the first source image. Requires 
    the inverse motion field (image 2->image 1).
  * Forward mesh extrapolation: construct a textured triangle mesh from the image 
    and transform the mesh by using the forward motion field (image 1->image 2).

The third feature implemented in this package is morphing between two 
images. In addition to cross-fading between images, the algorithm 
uses the forward and inverse motion fields to warp the images 
closer to each other before interpolating pixel values linearly. 
This technique yields much better results than simple cross-fading. 

Dependencies and installation
-----------------------------

Required dependencies (included in the package):
  CImg                    http://cimg.sourceforge.net

Optional dependencies:
  Boost.Program_options   http://www.boost.org
  CGAL                    http://www.cgal.org
  OpenCV                  http://sourceforge.net/projects/opencv

This package uses CMake as its build system. To build and 
install the package, create a build directory, and type the 
following commands in it:

  cmake <src_dic> [optional_flags]
  make
  make install

The last step is optional. You can also use the compiled packages 
in the build directory. The src_dir argument points to the source directory. 
Enabling and disabling external dependencies is done via the optional_flags 
argument. The following flags are currently supported:

  -DWITH_BOOST_PROGRAM_OPTIONS=ON/OFF  command-line interface via Boost.Program_options
  -DWITH_CGAL=ON/OFF                   support for sparse motion fields via CGAL
  -DWITH_OPENCV=ON/OFF                 support for OpenCV algorithms

The test programs for the above features are "extractmotion", 
"extrapolate" and "morph". To print their command-line syntax, 
run them without arguments.

Experimental support for MATLAB is also implemented. A MEX-file 
and a MATLAB script for testing it are located in the matlab-directory.

References
----------

The motion detection algorithms in this package are based on the 
following articles:

[1] J. Bouguet, Pyramidal Implementation of the Lucas Kanade 
    Feature Tracker: Description of the Algorithm, Technical 
    report, OpenCV documents, Intel Corporation, Microprocessor 
    Research Labs, 2000

[2] B.D. Lucas and T. Kanade, An iterative image registration 
    technique with an application to stereo vision, in Proc. 
    Seventh International Joint Conference on Artificial 
    Intelligence, Vancouver, 1981, pp. 674-679

[3] M. Proesmans, L. Van Gool, E. Pauwels, and A. Oosterlinck, 
    Determination of optical flow and its discontinuities using 
    non-linear diffusion, in 3rd European Conference on Computer 
    Vision, ECCV’94, 1994, Vol. 2, pp. 295–304.

and the C-based implementations found at:

ftp://ftp.csd.uwo.ca/pub/vision
http://of-eval.sourceforge.net

Author: Seppo Pulkkinen <seppo.pulkkinen@utu.fi>