during my work I’ve developed a node/nodelet that merges the depth images of two depth cameras (RGBD, TOF, whatever generates depth images) into a single depth image as it would be seen from an (almost) arbitrary perspective. It basically transforms and combines the point clouds from two depth cameras of a known transformation to each other into one, and uses a pinhole camera model to create the depth image from a third given perspective.
Here’s the repository on GitHub: https://github.com/HWiese1980/merge_depth_cam
Feel free to comment, criticize and contribute! But be aware that this has evolved from the mere need of enhancing the field of view of a depth cam by adding a second one. It’s not optimized in any way. Hence, I’ll certainly be grateful for any contribution especially regarding performance optimization!