Add arguments for estimateRigidTransform

[take1014]

resolves #4270

[hrnr]

there is already estimateAffinePartial2D and friends https://docs.opencv.org/master/d9/d0c/group__calib3d.html#gad767faff73e9cbd8b9d92b955b50062d

I think we should really get rid of estimateRigidTransform in OpenCV 4, or reimplement using estimateAffinePartial2D to avoid duplicate code.

[terfendail]

It looks like the test checks the result doesn't depend on RANSAC parameters rather than its correctness. Wouldn't it be better to check the result against aff matrix?

[take1014]

Thanks for the follow.
Certainly, I need to fix this code as you pointed out.
However, as alalek and hrnr say, I do not modify the interface of estimateRigidTransform, because we need to unify the "posture estimation" algorithm to calib3d's function instead of video module's.

[alalek]

I agree with @hrnr to keep "pose estimation" algorithms in a single place - calib3d module.
"Video" module should reuse existed functionality instead of re-implementing similar algorithms.

[take1014]

@alalek, @hrnr, @terfendail
Thanks for the comments.
I agree with you.
I think "estimateAffinePartial2D" should be reused instead of "estimateRigidTransform".

[hrnr]

maybe we should deprecate estimateRigidTransform in OpenCV 4. The RANSAC part could be reimplemented using calib3d right now.