[Olena-patches] milena r1070: Fast gaussian works on 2d images

URL: https://svn.lrde.epita.fr/svn/oln/trunk/milena ChangeLog: 2007-09-05 Simon Nivault &lt;simon.nivault(a)lrde.epita.fr&gt; Fast gaussian works on 2d images * sandbox/nivault/fast_gaussian.hh: . * sandbox/nivault/tests/test.cc: . fast_gaussian.hh | 25 +++++++++++++++++-------- tests/test.cc | 7 ++++--- 2 files changed, 21 insertions(+), 11 deletions(-) Index: trunk/milena/sandbox/nivault/tests/test.cc =================================================================== --- trunk/milena/sandbox/nivault/tests/test.cc (revision 1069) +++ trunk/milena/sandbox/nivault/tests/test.cc (revision 1070) @@ -34,9 +34,10 @@ int main() { mln::image2d_b<mln::value::int_u<8u> > im1 = mln::io::load_pgm("../../../img/lena.pgm"); - mln::image2d_b<float> im2; + mln_precondition(im1.has_data()); + mln::image2d_b<mln::value::int_u<8u> > im2(im1.domain()); - mln::linear::gaussian(im1, 0.2, im2); + mln::linear::gaussian(im1, 2, im2); - // mln::io::save_pgm(im2, "gausslena.pgm"); + mln::io::save_pgm(im2, "gausslena.pgm"); } Index: trunk/milena/sandbox/nivault/fast_gaussian.hh =================================================================== --- trunk/milena/sandbox/nivault/fast_gaussian.hh (revision 1069) +++ trunk/milena/sandbox/nivault/fast_gaussian.hh (revision 1070) @@ -30,7 +30,7 @@ #include <mln/core/concept/image.hh> #include <mln/level/paste.hh> -#include <mln/geom/ncols.hh> + namespace mln @@ -178,7 +178,7 @@ + c.n[3]*ima(current - d - d - d) - c.d[1]*tmp1[i - 1] - c.d[2]*tmp1[i - 2] - c.d[3]*tmp1[i - 3] - c.d[4]*tmp1[i - 4]; - current += d; + current = current + d; } // Non causal part @@ -211,7 +211,7 @@ + c.nm[4]*ima(current + d + d + d) - c.dm[1]*tmp2[i+1] - c.dm[2]*tmp2[i+2] - c.dm[3]*tmp2[i+3] - c.dm[4]*tmp2[i+4]; - current -= d; + current = current - d; } // Combine results from causal and non-causal parts. @@ -220,7 +220,7 @@ for (int i = 0; i < len; ++i) { ima(current) = tmp1[i] + tmp2[i]; - current += d; + current = current + d; } } @@ -230,14 +230,23 @@ gaussian_(Image<I>& img_, const F& coef) { I& img = exact(img_); - typedef mln_point(I) P; - // Apply on columns. + // Apply on rows. + for (int j = 0; j < geom::ncols(img); ++j) + recursivefilter_<float>(img, coef, + make::point2d(-img.border(), j), + make::point2d(geom::nrows(img) - 1 + img.border(), j), + geom::nrows(img) + 2 * img.border(), + make::dpoint2d(1, 0)); + + // Apply on columns. + for (int i = 0; i < geom::nrows(img); ++i) recursivefilter_<float>(img, coef, - make::point2d(0, - img.border()), // FIXME - make::point2d(0, geom::ncols(img) - 1 + img.border()), + make::point2d(i, -img.border()), + make::point2d(i, geom::ncols(img) - 1 + img.border()), geom::ncols(img) + 2 * img.border(), make::dpoint2d(0, 1)); + }