Hi Lubomir,
>Luminosity is a single channel, and single-channel images in GIL are specified using a gray_layout_t layout. It doesn't make sense to keep the layout of the destination view; in fact you should require that the destination view is a single >channel view (i.e. has a gray_layout_t layout)
You mentioned requiring inputs to be a specific layout and it got me thinking about what I currently do. I have generic functions that use the ColorSpacesCompatibleConcept but will compile and execute if the inputs are of different colorspaces. For example, a rgb8_view_t and a gray8_view_t can be passed in and pass the ColorSpacesCompatibleConcept check. What makes 2 colorspaces compatible? The documentation leads me to believe that they have to be identical for them to be compatible. How does the example I gave (rgb8 and gray8) not cause an issue?
I can try to build an example if this is not expected behavior. If it is expected behavior, can you explain how the static_transform operation pairs the channels between those 2 images?
Thanks,
Mark