Hi Olivier, sorry for the late reply. Interesting problem you have and I started looking into it.

...

What is the best way to compute the difference between 2 any_image? Of course, the input any_image may not be of the same type... So, if pixels are unsigned int for the 1st one and double for the 2nd one, the result must be stored as double. I (accidently) found pixel_minus_t in pixel_numeric_operations.hpp but cannot figure how to use it.

Where did you find pixel_numeric_operations.hpp? Regards, Christian

Hi Olivier,

Great, thank you! What about my first solution?

Yes. I think it's the right way. Here is what I have which uses transform_pixels. #include #include namespace boost { namespace gil { typedef double bits64F; GIL_DEFINE_BASE_TYPEDEFS(64F,gray) typedef float bits32F; GIL_DEFINE_BASE_TYPEDEFS(32F,gray) } } using namespace boost; using namespace gil; template< typename DST, typename SRC1, typename SRC2 > DST generate_diff( const SRC1& src1, const SRC2& src2 ) { return pixel_minus_t()(src1, src1); } int main(int argc, char** argv) { gray16s_image_t mns( 10, 10 ); gray32F_image_t mnt( 10, 10 ), mne( 10, 10 ); fill_pixels( view( mns ), gray16s_image_t::value_type( 12 )); fill_pixels( view( mnt ), gray32F_image_t::value_type( 5 )); fill_pixels( view( mne ), gray32F_image_t::value_type( 0 )); auto fun = generate_diff< gray16s_image_t::value_type , gray32F_image_t::value_type , gray32F_image_t::value_type >; transform_pixels( view(mne), view(mns), view(mnt), fun ); return 0; } Next I'll try dynamic_image. Regards, Christian

Hi Olivier, here is a slightly better version: #include #include #include namespace boost { namespace gil { template struct channel_abs_t : public std::binary_function { ChannelR operator()(typename channel_traits<Channel1>::const_reference ch1, typename channel_traits<Channel2>::const_reference ch2) const { return std::abs( ChannelR(ch1) - ChannelR(ch2) ); } }; template // models pixel value concept struct pixel_abs_t { PixelR operator() (const PixelRef1& p1, const PixelRef2& p2) const { PixelR result; static_transform(p1,p2,result, channel_abs_t()); return result; } }; typedef double bits64F; GIL_DEFINE_BASE_TYPEDEFS(64F,gray) typedef float bits32F; GIL_DEFINE_BASE_TYPEDEFS(32F,gray) } // namespace gil } // namespace boost using namespace boost; using namespace gil; template< typename SRC1, typename SRC2, typename DST > void generate_diff( const SRC1& src1, const SRC2& src2, const DST& dst ) { transform_pixels( src1, src2, dst, pixel_abs_t< typename SRC1::value_type , typename SRC2::value_type , typename DST::value_type >() ); } int main(int argc, char** argv) { { gray16s_image_t mns( 10, 10 ); gray32F_image_t mnt( 10, 10 ), mne( 10, 10 ); fill_pixels( view( mns ), gray16s_image_t::value_type( 5 )); fill_pixels( view( mnt ), gray32F_image_t::value_type( 12 )); fill_pixels( view( mne ), gray32F_image_t::value_type( 0 )); generate_diff( const_view( mns ), const_view( mnt ), view( mne ) ); gray32F_image_t::value_type p = *view( mne ).xy_at( 0, 0 ); } return 0; } Does that somehow fit your needs? Regards, Christian