Extend cudaimgproc::demosaicing for f32

[apbr]

Extend cudaimgproc::demosaicing to also allow images with f32 values.

Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

• I agree to contribute to the project under Apache 2 License.
• To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
• The PR is proposed to the proper branch
• There is a reference to the original bug report and related work -> not applicable
• There is accuracy test, performance test and test data in opencv_extra repository, if applicable
  Patch to opencv_extra has the same branch name.
• The feature is well documented and sample code can be built with the project CMake

[llschloesser]

@apbr please see #3777

It seems like adding 32f to that PR would be easier than deconflicting with this one and we might get all functionality merged sooner. Can I ask what sensor you are using where you have 32f raw pixels?

[apbr]

@llschloesser #3777 was merged 1 hour after your comment. So I'll just rebase my PR.

Can I ask what sensor you are using where you have 32f raw pixels?

We do some image processing steps before the debayering. Their resulting datatype is f32. So it's not directly from the sensor.

[llschloesser]

If you follow the template pattern of the earlier PR you'll get 32f to gray and 4 channel, there is no need to restrict to 3 channel.

[apbr]

If you follow the template pattern of the earlier PR you'll get 32f to gray and 4 channel, there is no need to restrict to 3 channel.

Good point, I'll rework and test it.

[apbr]

Reworked the changes.

[asmorkalov]

@apbr Could you add simple test for the new feature?

[apbr]

@apbr Could you add simple test for the new feature?

@asmorkalov I could give it a try. Should they be in here?

[asmorkalov]

Yes, as shortcut you can use existing demosaic test and reference data for u8, but scale it for larger range. I just want to be sure that the template implementation does not lead to overflows or other arithmetical issues.