Boost logo

Boost :

From: Olzhas Zhumabek (anonymous.from.applecity_at_[hidden])
Date: 2021-08-01 20:17:49


As mentioned in the issue (https://github.com/boostorg/gil/issues/622),
current image processing and computer vision algorithms could use the
matrix types for simplicity and ease of use. Although not strictly
required, it would remove most of the problems associated with conversions
and pixel incompatibilities. As @sdebionne
<https://github.com/sdebionne> mentioned,
the conversion can be handled by writing conversion functions like the ones
defined in color_convert.hpp.

At the moment there is a need only for single channel matrix types, as
multi channel matrices with pixels not properly packed might confuse future
users.

What do you think? Is it a good idea to bring those into extension/numeric,
and then into GIL? Or shall we find another solution? Although it is easy
to have user defined images, I believe having one default is a good idea.
Those who would like to add some extra context could use type alias or
something like that.

Best,
Olzhas


Boost list run by Boost-Gil-Owners