https://svn/svn/oln/prototypes/proto-1.0/olena Index: ChangeLog from Nicolas Widynski &lt;nicolas.widynski(a)lrde.epita.fr&gt; Update anisotropic diffusion filter. * oln/level/anisotropic_diffusion.hh: Fix bug. anisotropic_diffusion.hh | 41 ++++++++++++++++++++++++++++------------- 1 file changed, 28 insertions(+), 13 deletions(-) Index: oln/level/anisotropic_diffusion.hh --- oln/level/anisotropic_diffusion.hh (revision 394) +++ oln/level/anisotropic_diffusion.hh (working copy) @@ -31,8 +31,9 @@ # include <oln/basics.hh> # include <oln/utils/clone.hh> # include <oln/morpho/gradient_morpho.hh> + # define SQR(u) u*u -# define ABS(u) u > 0 ? u : -u +# define ABS(u) u >= 0 ? u : -u namespace oln { @@ -94,7 +95,6 @@ double compute_capa(const abstract::image<I>& input) { double K = 0; - // on admet un 255 dans l'image double seuil = 0.9 * 255; oln_type_of(I, concrete) grad = morpho::gradient_morpho(input); @@ -108,25 +108,27 @@ return K; } + double diffusion_coefficient(double grad, double K) { - std::cout << "----------------" << std::endl; - std::cout << grad << std::endl; - std::cout << K << std::endl; return 1 / (1 + SQR(grad / K)); } template <typename I> oln_type_of(I, concrete) - anisotropic_diffusion_(const abstract::image<I>& input) + anisotropic_diffusion_(const abstract::image<I>& input, unsigned nb_it) { oln_type_of(I, concrete) a_d(input.size(), "gradient"); oln_ch_concrete_type(I, std::vector<double>) grad(input.size(), "gradient"); oln_type_of(I, fwd_piter) p(input.size()); bool conv = false; + bool change = false; double K = 0; double modify = 0; + double error = 0; + double tmp = 0; + unsigned it = 1; a_d = clone(input.exact()); @@ -135,6 +137,8 @@ clean_ima(a_d.exact()); grad = gradient_on_each_side_(a_d); K = compute_capa(a_d); + change = false; + error = 0; for_all_p(p) { @@ -142,17 +146,28 @@ typename std::vector<double>::const_iterator it; - for (it = grad[p].value().begin(); it != grad[p].value().end(); it++) - modify += *it * diffusion_coefficient(ABS((double)*it), K); + for (it = grad.at(p).begin(); it != grad.at(p).end(); it++) + modify += (double)(*it) * diffusion_coefficient((double)*it, K); - if (a_d[p] + 0.14 * modify >= 0 and a_d[p] + 0.14 * modify <= 255) + if (modify != 0 and a_d[p] + 0.14 * modify >= 0 and a_d[p] + 0.14 * modify <= 255) + { + tmp = a_d[p]; a_d[p] = a_d[p] + 0.14 * modify; + if ((double)a_d[p] != tmp) + change = true; + } + + error += ABS(modify); } - std::cout << modify << std::endl; - if (not modify) + std::cout << "error " << error << std::endl; + + if (it >= nb_it or error < 0.1 or change == false) conv = true; + + it++; }; + return a_d; } @@ -163,9 +178,9 @@ template <typename I> oln_type_of(I, concrete) - anisotropic_diffusion(const abstract::image<I>& input) + anisotropic_diffusion(const abstract::image<I>& input, unsigned nb_it) { - return impl::anisotropic_diffusion_(input.exact()); + return impl::anisotropic_diffusion_(input.exact(), nb_it); } } // end of namespace oln::level