The answer is no. The image server can only give you the original image.
How should I map bounding boxes then? Is there a guide for sending bounding boxes on fisheye cameras so that they map correctly on the dewarped smart client view?
I just talked to my colleague regarding this question, but the answer is still no, there is no such a guide unfortunately. The only thing that you could do is to develop your own dewarping but it might be not so easy, we assume.