Point Cloud Filtering and processing using PCL is slow

edit

Hi All

I have code for a ROS node that subscribes to A PointCloud2 topic published by iai_kinect2's kinect bridge. My problem is that the code runs quite slowly. I have narrowed down the slowing-down to my spinOnce() statement, so I'm pretty sure it's my callback function that's being slow. However, my callback is using the reccommended way of converting the pc2 message into a PCL point cloud. Here it is for reference:

void cloudCallback(const sensor_msgs::PointCloud2ConstPtr& msg){
  // Create a container for the data.
  pcl::PCLPointCloud2 pcl_pc2;
  pcl_conversions::toPCL(*msg,pcl_pc2);
  pcl::fromPCLPointCloud2(pcl_pc2,*cloud);

  pcl::removeNaNFromPointCloud(*cloud,*outputCloud, indices);

  sor.setInputCloud(outputCloud);
  sor.setLeafSize (0.01f, 0.01f, 0.01f);
  sor.filter (*outputCloudFilt);
}

All the objects used here were pre-initialised before main(). The code runs just as slowly if I remove the filtering from the callback.

I should also add that the speed is slow independently of how many points are in the cloud.

EDIT: Using fromROSMsg() doesn't help with my problem.