RGBD tests refactoring

[DumDereDum]

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 other license that is incompatible with OpenCV
• The PR is proposed to proper branch
• There is reference to original bug report and related work
• 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

[tomoaki0705]

If you are intended of refactoring RGBD module, I really appreciate if you could merge those 2 implementations.
i.e. merge TSDFVolumeGPU class and TSDFVolumeCPU class and just keep TSDFVolume
TSDFVolumeGPU class triggers several kernels under the hood, but single kernel build failure leads directly to runtime error.
The difference can be only decided at the beginning.

It will be way more convenient for user if any kernel build/run failure fall back to CPU version during the function call.

More precisely, use following cascaded function call everywhere

CV_OCL_RUN(ocl_integrate(....))

// followed by CPU version
cpu_integrate();

Putting HAVE_OPENCL in the test code is really the opposite way to go.

That's why it's called "T-API" (Transparent API) 1 2

[savuor]

@tomoaki0705, the way to organize code could be more consistent but would involve constant copying data from/to host and GPU memory, which would involve huge stalls during code execution.

[savuor]

We don't use namespace indentation, please fix this.
Also when running TSDF GPU tests with display=true flag I see quite glitchy image during rendering. Using viz I could find out that rendered points are OK but normals have wrong direction. Now investigating this.

[savuor]

@DumDereDum Found out what's the reason of the bug:
https://github.com/opencv/opencv_contrib/blob/master/modules/rgbd/src/opencl/tsdf.cl#L230

Here you subtract compressed values then convert them to float. This is incorrect: the values should be converted to float, then subtracted, this solves the bug.
Please fix this in this PR and add the corresponding mention to PR description.

[savuor]

Please add explicit setUseOpenCL() calls here and in normal_test function

[DumDereDum]

If we have OpenCL we don't need to use setUseOpenCL

[savuor]

Please remove OCL tests from this file, we have them in ocl/ folder

[DumDereDum]

No, if I remove them there would be two equal GPU tests

[savuor]

👍