On Mar 07, 2011, at 07:44 AM, Robert Lupton the Good <rhl@astro.princeton.edu> wrote:

...

Sorry for the slow response. I simply defined the following:

typedef boost::gil::pixel<float, boost::gil::gray_layout_t> gray_float_pixel_t; typedef boost::gil::image<gray_float_pixel_t, false> gray_float_image_t; typedef gray_float_image_t::view_t gray_float_view_t; typedef gray_float_image_t::const_view_t gray_float_const_view_t;

As Dr. Bourdev advised, I do not use with these typedefs any of the color conversion operations offered by GIL. Apparently those rely on the scaling properties of the float wrapper type.

So far, all I have done with these typedefs is ...

1. construct views using interleaved_view and pixel data pointed to by [const] float*, as illustrated near the beginning of the GIL tutorial 2. create x- and y- iterators for the views to apply row- and column-oriented algorithms

Sorry for _my_ slow response (I was in Japan). I don't care about colour conversions, but I think that these types will convert to integral types with scaling unless you add more boilerplate. I added the following -- but it may not all have been needed; this was quite a while ago and I was new to gil. I also have some remaining conversion problems where I have to rewrite expressions to make them compile, but this may be something else. Anyway, I'd really like to see native gil support for a complete set of NON-scaling types.

namespace boost { namespace gil { /* * Define a type that's a pure float, without scaling into [0, 1] */ typedef float bits32f_noscale;

GIL_DEFINE_BASE_TYPEDEFS(32f_noscale, gray) GIL_DEFINE_ALL_TYPEDEFS_INTERNAL(32f_noscale, dev2n, devicen_t<2>, devicen_layout_t<2>)

template<> struct channel_multiplier<bits32f_noscale> : public std::binary_function<bits32f_noscale,bits32f_noscale,bits32f_noscale> { bits32f_noscale operator()(bits32f_noscale a, bits32f_noscale b) const { return a*b; } };

template <typename DstChannelV> struct channel_converter<bits32f_noscale, DstChannelV> : public std::unary_function<bits32f_noscale,DstChannelV> { DstChannelV operator()(bits32f_noscale x) const { return DstChannelV(x + 0.5f); } };

template <typename SrcChannelV> struct channel_converter<SrcChannelV,bits32f_noscale> : public std::unary_function<SrcChannelV,bits32f_noscale> { bits32f_noscale operator()(SrcChannelV x) const { return bits32f_noscale(x); } };

// // Totally specialised templates to resolve ambiguities // #define CONVERT_NOOP(T1, T2) \ template <> \ struct channel_converter<T1, T2> : public std::unary_function<T1, T2> { \ T2 operator()(T1 x) const { return static_cast<T2>(x); } \ }; \ \ template <> \ struct channel_converter<T2, T1> : public std::unary_function<T2, T1> { \ T1 operator()(T2 x) const { return static_cast<T1>(x); } \ }

CONVERT_NOOP(unsigned char, short); CONVERT_NOOP(unsigned char, unsigned short); CONVERT_NOOP(unsigned char, int); CONVERT_NOOP(unsigned short, short); CONVERT_NOOP(unsigned short, int); CONVERT_NOOP(short, int);

#undef CONVERT_NOOP

Using only the four typedefs that I provided earlier, I don't see any sign that the GIL classes that I'm using are scaling the floats behind the curtain. I haven't been able to follow the details of the metaprogramming inside GIL, but I have observed two key facts that suggest that no scaling is occuring: 1. The end result has the scale I expect (without scaling). 2. std::less<typename std::iterator_traits<InIter>::value_type> works with InIter = gray_float_const_view_t::x_iterator (or y_iterator). Hopefully, Christian or Lubomir will provide a more authoritative response <nudge/>