Synopsis

This library instantiates a directional vesselness filter that enhances threadlike structures. This filter is used in the conference paper "Automatic initialization and dynamic tracking of surgical suture threads" by Jackson et al. This package includes both a CPU and GPU implementation.

Code Example

The basic Robot OS call for the filters are:

rosrun vesselness_image_filter vesselness_image_filter_gpu_node     #GPU based vesselness filter
rosrun vesselness_image_filter vesselness_image_filter_cpu_node      #CPU based vesselness filter
rosrun vesselness_image_filter vesselness_image_filter_cpu_bw_node  #CPU based vesselness filter

By defaults the filters subscribe to the ros topic: "image_in" and publish to the ros topic, "image_thin".

The viewers can be called as follows:

rosrun vesselness_image_filter vesselness_image_viewer_gpu_node    #GPU based vesselness viewer
rosrun vesselness_image_filter vesselness_image_viewer_cpu_node    #CPU based vesselness viewer

By defaults the viewers subscribe to the ros topic: "image_in" and display the image using OpenCV.

The filter has several parameters that can be configured using rqt_reconfigure.

Installation

The CPU based filter and viewer will operate with ROS and OpenCV. If GPU functionality is desired, OpenCV must be recompiled with CUDA support on a CUDA capable GPU.

Follow the instructions in "Building OpenCV With CUDA.pdf"

Then make sure you have the modified compatibility version of vision_opencv: git clone https://github.com/cwru-robotics/vision_opencv.git -b kinetic

License

This package is licensed under the BSD-3-Clause. Please direct all inquiries to [email protected].