文件列表:

lk20\baker_simon_2002_3.pdf (521908, 2009-05-14)
lk20\lk20-common\hessian.m (461, 2003-08-20)
lk20\lk20-common\init_a.m (710, 2003-08-20)
lk20\lk20-common\init_h.m (716, 2003-08-20)
lk20\lk20-common\jacobian_a.m (535, 2003-08-20)
lk20\lk20-common\jacobian_h.m (1056, 2003-08-20)
lk20\lk20-common\quadtobox.m (1312, 2003-08-20)
lk20\lk20-common\quadtobox_h.m (1433, 2003-08-20)
lk20\lk20-common\sd_images.m (541, 2003-08-20)
lk20\lk20-common\sd_update.m (491, 2003-08-20)
lk20\lk20-common\verb_init_a.m (900, 2003-08-20)
lk20\lk20-common\verb_init_h.m (943, 2003-08-20)
lk20\lk20-common\verb_plot_a.m (648, 2003-08-20)
lk20\lk20-common\verb_plot_h.m (691, 2003-08-20)
lk20\lk20-common\warp_a.m (771, 2003-08-20)
lk20\lk20-common\warp_h.m (712, 2003-08-20)
lk20\lk20-common (0, 2009-05-15)
lk20\lk20-p1\affine_fa.m (2577, 2003-08-20)
lk20\lk20-p1\affine_fc.m (2781, 2003-08-20)
lk20\lk20-p1\affine_ia.m (2610, 2003-08-20)
lk20\lk20-p1\affine_ic.m (2757, 2003-08-20)
lk20\lk20-p1\affine_ic_d.m (3246, 2003-08-20)
lk20\lk20-p1\affine_ic_lm.m (3712, 2003-08-20)
lk20\lk20-p1\affine_ic_nt.m (3519, 2003-08-20)
lk20\lk20-p1\affine_ic_nt_d.m (3906, 2003-08-20)
lk20\lk20-p1\affine_ic_sd.m (2809, 2003-08-20)
lk20\lk20-p1\data\affine_pt_offset.mat (240200, 2003-08-20)
lk20\lk20-p1\data\denso.mat (345984, 2003-08-20)
lk20\lk20-p1\data\homo_pt_offset.mat (320200, 2003-08-20)
lk20\lk20-p1\data\images\denso.png (553763, 2003-08-20)
lk20\lk20-p1\data\images\knee.jpg (24236, 2003-08-20)
lk20\lk20-p1\data\images\simon.jpg (10049, 2003-08-20)
lk20\lk20-p1\data\images\takeo.pgm (33962, 2003-08-20)
lk20\lk20-p1\data\images (0, 2009-12-17)
lk20\lk20-p1\data\knee.mat (262016, 2003-08-20)
lk20\lk20-p1\data\simon.mat (34128, 2003-08-20)
lk20\lk20-p1\data\takeo.mat (34128, 2003-08-20)
lk20\lk20-p1\data (0, 2009-05-15)
lk20\lk20-p1\figures\cnvg_plot_affine.m (1791, 2003-08-20)
lk20\lk20-p1\figures\freq_plot_affine.m (2006, 2003-08-20)
... ...

Lucas-Kanade 20 Years On: A Unifying Framework Part 1: The Quantity Approximated, the Warp Update Rule and the Gradient Descent Approximation Simon Baker and Iain Matthews The Robotics Institute Carnegie Mellon University Pittsburgh, PA 15213 USA Contact: simonb@cs.cmu.edu, iainm@cs.cmu.edu --------------------------------------------------------------------------- URL: http://www.ri.cmu.edu/projects/project_515.html Bibtex reference: @article{Baker_2004_4293, author = "Simon Baker and Iain Matthews", title = "Lucas-Kanade 20 Years On: A Unifying Framework Part 1: The Quantity Approximated, the Warp Update Rule, and the Gradient Descent Approximation", journal = "International Journal of Computer Vision", year = "2004" } --------------------------------------------------------------------------- MATLAB SOURCE CODE --------------------------------------------------------------------------- The source code is split into two directories, both need to be added to your Matlab search path: lk20-p1 The algorithms described in the paper and test scripts lk20-common Common code for all algorithms (and for later papers) lk20-p1/ -------- The Gauss-Newton affine warp algorithms in Section 3 of the paper are: affine_fa.m Forwards-Additive affine_fc.m Forwards-Compositional affine_ia.m Inverse-Additive affine_ic.m Inverse-Compositional The Gauss-Newton projecive warp algorithms in Section 3 of the paper are: homo_fa.m Forwards-Additive homo_fc.m Forwards-Compositional homo_ic.m Inverse-Compositional Note: the inverse-additive algorithm cannot be applied with a projective warp. The additional algorithms from Section 4 of the paper are: affine_ic_nt.m Inverse-Compositional Newton (4.2) affine_ic_sd.m Inverse-Compositional Steepest Descent (4.3) affine_ic_d.m Inverse-Compositional with diagonal Hessian (4.4) affine_ic.nt_d.m Inverse-Compositional Newton with diagonal Hessian (4.4) affine_ic_lm.m Inverse-Compositional Levenburg-Marquardt (4.5) To run the perturbation experiments described in the paper examine: run_affine.m Run an affine warp experiment run_homo.m Run a projective warp experiment These files should be edited for the experimental parameters you would like to test. By default they run everything for a few tests. To replicate the experiments in the paper run each algorithm for 100 convergence tests and 1000 frequency of convergence tests. This will take some time! lk20p1/data/ ------------ The data directory contains precomputed random perturbation data (for experimental consistency) and the images and template definitions. lk20p1/figures/ --------------- The figures directory contains matlab scripts for plotting your results and creating graphs similar to those shown in the paper.

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：