Inconsistency in image and pixel types using floating point channels.

[awulkiew]

trafficstars

The typedefs for images and pixels using 32-bit and 64-bit floating point channel values are inconsistent. The 32-bit types defined in boost/gil/typedefs.hpp uses boost::gil::scoped_channel_value<float, ...> while 64-bit types defined in boost/gil/io/typedefs.hpp uses double.

float64_t is defined in boost/gil/typedefs.hpp as scoped_channel_value<double, ...> so my guess is that this is the correct channel type for 64-bit images and pixels.

[mloskot]

So, there is 64-bit floating-point type for channel defined using the scoped_channel_value channel adaptor:

https://github.com/boostorg/gil/blob/ffdbf4e249260e43f6438177f1b7ebdee639fa03/include/boost/gil/typedefs.hpp#L128

and this type will be used as ChannelModel in pixel<ChannelModel, ...>:

https://github.com/boostorg/gil/blob/ffdbf4e249260e43f6438177f1b7ebdee639fa03/include/boost/gil/typedefs.hpp#L33

while Core IO re-defines it using raw double type as the ChannelModel:

https://github.com/boostorg/gil/blob/ffdbf4e249260e43f6438177f1b7ebdee639fa03/include/boost/gil/io/typedefs.hpp#L53

This does look fishy! Thank you for reporting this.

Any comments on that, @chhenning & @stefanseefeld?

The only references in the Boost list to that adaptor are rather general in "scoped_channel_value type" https://lists.boost.org/Archives/boost//2010/07/169159.php and in the "image rescale algorithms", users make two interesting observations:

1. https://lists.boost.org/Archives/boost//2009/09/156246.php

   The current state of rgba32f_pixel_t is meant to be scoped in the 0..1 range 'by contract'

   • even if GIL does not (currently) implement range checking

2. https://lists.boost.org/Archives/boost//2009/09/156290.php

   I think rgba32f_pixel_t is a good model for hdr pixels. I don't see the 0..1 range in floating point channels as limits. And GIL does not enforce them in any way. They only define a mapping between integer and floating point channels.

Those, however, apply to the core definitions.

@chhenning is the main author of the new IO, so I hope he will be able to comment on the mismatches in gil/io/typedefs.hpp.