Re: [gsoc2020] Answer to histogram design questions from Slack

From: Mateusz Loskot (mateusz_at_[hidden])
Date: 2020-05-13 17:40:41

On Wed, 13 May 2020 at 18:12, Debabrata Mandal
<mandaldebabrata123_at_[hidden]> wrote:
>
> I would also like to add some comments.
>
> 1. The idea of simplifying the implementation details to single channeled
> views seem sound to me. In this way we could support STL (compatible)
> containers for now. And easily provide specializations for other types
> (Boost.Histogram) as required. Please correct me if I am wrong.

Yes, as what Pranam suggested and I agreed to as well,
a single channel does sound reasonable to start with and focus on.

> 2. We could support Boost.Histogram for now but that might be enforcing
> something on the user side. We could leave the container choice to the user
> for histogram of single channeled images for now.

Certainly, for single channel images, such flexibility is feasible and
desirable.

> 3. I have proposed to add algorithms related to histogram manipulations so
> eventually we might need to standardize on some container choice unless we
> either decide on removing support for those algorithms or let them take the
> choice of histogram container as well.

Yes, as image and operation complexity increases, we may need to narrow
histogram container choices.

> 4. There is no usage of multi dimensional histograms in my project apart
> from the part of implementing them as part of Boost.GIL .

I did not suggest anything other than what has been proposed:
- generic 1D, 2D, 3D histogram functions
- 1D, 2D cumulative histogram function
In fact, I assumed we may only be able to provide calculations of
2-3 D histograms during this project, nothing else.

Best regards,
-- 
Mateusz Loskot, http://mateusz.loskot.net