oln::convol::fast::internal Namespace Reference

Internal purpose only. More...

Classes
struct	gaussian_
	Compute the gaussian filter. More...
struct	gaussian_< 1 >
struct	gaussian_< 2 >
struct	gaussian_< 3 >
struct	recursivefilter_coef_
	Data structure for coefficients used for a recursive filter call. More...
Functions
template<class WorkType, class FloatType, class I> void	recursivefilter_ (I &image, const recursivefilter_coef_< FloatType > &c, const oln_point_type(I)&start, const oln_point_type(I)&finish, coord len, const oln_dpoint_type(I)&d)
	Recursive filter.
Variables
mute< I, typename convoutput< C, B, oln_value_type(I)>::ret >::re	gaussian_common_ )(const convert::abstract::conversion< C, B > &c, const abstract::image< I > &in, const F &coef, ntg::float_s sigma, const abstract::behavior< BE > &behavior)
	Common code for filter (derivative, second derivative, etc.).

---

Detailed Description

Internal purpose only.

---

Function Documentation

template<class WorkType, class FloatType, class I> void recursivefilter_ (  I &  image, const recursivefilter_coef_< FloatType > &  c, const oln_point_type(I)&  start, const oln_point_type(I)&  finish, coord  len, const oln_dpoint_type(I)&  d )

Recursive filter. Recursive filter, works on a line. Todo: FIXME: Until something clever is designed, the line is defined by two points (START and FINISH) and a displacement dpoint (D). Parameters: Worktype  Type the algorithm work on. FloatType  Type of coefficients. I  Type of image. image Image to process. c Coefficients. start Start of the line. finish End point of the line. len Length of data to process. d Displacement dpoint. Definition at line 67 of file fast_gaussian.hxx. References oln::coord, oln::convol::fast::internal::recursivefilter_coef_< FloatT >::d, oln::convol::fast::internal::recursivefilter_coef_< FloatT >::dm, oln::convol::fast::internal::recursivefilter_coef_< FloatT >::n, and oln::convol::fast::internal::recursivefilter_coef_< FloatT >::nm. { std::vector<WorkType> tmp1(len); std::vector<WorkType> tmp2(len); // The fourth degree approximation implies to have a special // look on the four first points we consider that there is // no signal before 0 (to be discussed) // -- // Causal part tmp1[0] = c.n[0]*image[start]; tmp1[1] = c.n[0]*image[start + d] + c.n[1]*image[start] - c.d[1]*tmp1[0]; tmp1[2] = c.n[0]*image[start + d + d] + c.n[1]*image[start + d] + c.n[2]*image[start] - c.d[1]*tmp1[1] - c.d[2]*tmp1[0]; tmp1[3] = c.n[0]*image[start + d + d + d] + c.n[1]*image[start + d + d] + c.n[2]*image[start + d] + c.n[3]*image[start] - c.d[1]*tmp1[2] - c.d[2]*tmp1[1] - c.d[3]*tmp1[0]; oln_point_type(I) current(start + d + d + d + d); for (coord i = 4; i < len; ++i) { tmp1[i] = c.n[0]*image[current] + c.n[1]*image[current - d] + c.n[2]*image[current - d - d] + c.n[3]*image[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; } // Non causal part tmp2[len - 1] = 0; tmp2[len - 2] = c.nm[1]*image[finish]; tmp2[len - 3] = c.nm[1]*image[finish - d] + c.nm[2]*image[finish] - c.dm[1]*tmp2[len-2]; tmp2[len - 4] = c.nm[1]*image[finish - d - d] + c.nm[2]*image[finish - d] + c.nm[3]*image[finish] - c.dm[1]*tmp2[len-3] - c.dm[2]*tmp2[len-2]; current = finish - d - d - d ; for (coord i = len - 5; i >= 0; --i) { tmp2[i] = c.nm[1]*image[current] + c.nm[2]*image[current + d] + c.nm[3]*image[current + d + d] + c.nm[4]*image[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; } // Combine results from causal and non-causal parts. current = start; for (coord i = 0; i < len; ++i) { image[current] = ntg::cast::force<oln_value_type(I)>(tmp1[i] + tmp2[i]); current += d; } }

---

Variable Documentation

mute<I, typename convoutput<C,B,oln_value_type(I)>::ret>::re oln::convol::fast::internal::gaussian_common_)(const convert::abstract::conversion<C,B>& c, const abstract::image<I>& in, const F& coef, ntg::float_s sigma, const abstract::behavior<BE> &behavior)

Common code for filter (derivative, second derivative, etc.). Parameters: C  Exact type of the conversion object. B  Base type of the conversion object. I  Exact type of the image. BE  Exact type of the behavior. input_conv Converter object. in Input image. sigma Value of sigma when computing the gaussian. behavior Object to know how to work on borders. Definition at line 297 of file fast_gaussian.hxx. { typename mute<I, ntg::float_s>::ret work_img(in.size()); oln_iter_type(I) it(in); for_all(it) work_img[it] = ntg::cast::force<ntg::float_s>(in[it]); // On tiny sigma, Derich algorithm doesn't work. // It is the same thing that to convolve with a Dirac. if (sigma > 0.006) { /* FIXME: relation between sigma and the border shouldn't be linear, so when sigma is big enougth, the signal may be parasitized by the non signal values. */ behavior.adapt_border(work_img, ntg::cast::round<coord>(5 * sigma)); gaussian_<I::dim>::doit(work_img, coef); } /* Convert the result image to the user-requested datatype. FIXME: We are making an unnecessary copy in case the user expects a ntg::float_s image. */ typename mute<I, typename convoutput<C,B,oln_value_type(I)>::ret>::ret out_img(in.size()); for_all(it) out_img[it] = c(work_img[it]); return out_img; }

---