Remote handling systems are commonly used for decommissioning and maintenance of hazardous environments, specially in the nuclear sector. The necessity for a more realistic and accurate user interaction with the remote environment has led research towards the usage of immersive technologies such as augmented and virtual reality. In order for this to succeed, the state of the remote environment needs to be known accurately at all times. Information gathered using RGB-D cameras can serve this purpose. The high accuracy and density of data retrieved by these devices provide an extraordinary insight of the remote environment but can represent a burden on the communication channels. This paper addresses two point cloud compression techniques based on kd-trees and octrees for point cloud data transmission within a Robot Operative System (ROS) communications middleware.