reprojectImageTo3D output to point cloud

edit

Hi to all,

I have a stereo camera and my goal is to obtain the 3d Point cloud fusing the two images. I have successfully created the depth image and the disparity map. Then, I used the reprojectImageTo3D function from Opencv to get the 3d points from the disparity map. I am a bit confused about how to convert such points in a Point Cloud, that I want to stream over ROS network.

I am using python and Opencv. Seems that the result of reprojectImageTo3D function is a numpy array. I have seen some example in which this data is saved into an output file, but I have not found any automatic conversion into PCL. Do you have any suggestion?

Thanks,