Optflow光流算法以及基于运动的变形和外推的C ++ / Python实现。

Overview -------- Optflow is a C++ library for motion detection and motion-based extrapolation and morphing algorithms. The following motion detection algorithms are implemented in Optflow: * the Horn&Schunck variational approach * global anisotropic diffusion algorithm by Proesmans et. al. * Lucas&Kanade least-squares -based block matching * interface for the Lucas&Kanade algorithm implemented in OpenCV The first two algorithms generate dense vector fields (i.e. motion vectors for each pixel in the source image). On the other hand, the Lucas&Kanade -based algorithms generate sparse motion fields. This is done by computing the motion for a given set of "feature points". All the above algorithms are implemented by using image pyramids. Optflow also contains algorithms for image warping by using a motion field. The following methods are implemented: * Inverse mapping: or each pixel in the second input image, find the corresponding pixel in the first one. * Forward mesh extrapolation: construct a textured triangle mesh from the image and transform the mesh by using the motion field. Dependencies and installation ----------------------------- Required dependencies: CImg >= 1.5 http://cimg.sourceforge.net Optional dependencies: Boost.Program_options http://www.boost.org CGAL >= 4.2 http://www.cgal.org OpenCV >= 2.4 http://sourceforge.net/projects/opencv Optflow uses CMake for generating the makefiles. To build and install the package, create a build directory, and type the following commands in it: cmake <src_dic> -DCMAKE_INSTALL_PREFIX=<dest_dir> make sudo make install The last step is optional. You can also use the compiled packages from the build directory. The src_dir argument points to the source directory, and dest_dir is the installation directory (e.g. /usr or /usr/local). Enabling and disabling external dependencies can be done via the following options: -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 installation is done to the following subdirectories in the destination directory: bin binaries lib libraries (liboptflow.so) include/optflow headers The compiled package includes three binaries: `extractmotion`, `extrapolate` and `morph`. To print their command-line syntax, run them without arguments. The file `liboptflow.so` contains a linkable library. Experimental support for MATLAB is also implemented. A MEX-file and a MATLAB script for running it are located in the matlab-directory. References ---------- The motion detection algorithms in this package are based on the following references: [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.K. Horn and B. Schunck, Determining Optical Flow, Artificial Intelligence, vol. 17, pp. 185-203, 1981 [3] 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 [4] 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@fmi.fi>