GuidedFilter is an implementation of the Guided Image Filtering algorithm in OpenCL. The Guided Filter is an image filter with many applications, one of which is edge-preserving smoothing. It has a non-approximate algorithm which is O(1) in the filter window size. It has excellent performance characteristics which make it a great alternative to the popular Bilateral Filter.
You can watch how the algorithm performs as a smoothing operator in the following two videos:
- Guided Image Filtering on Kinect RGB stream with OpenCL
- Guided Image Filtering on Kinect RGB and Depth streams with OpenCL
In the latter video, a point cloud is built from the Kinect streams. The Guided Image Filtering algorithm is applied to the Depth frame and each of the 3 channels of the RGB frame, separately. The frames are transfered to the GPU, processed with OpenCL, and then delivered directly to OpenGL. On my machine, I was able to get a mean running time of 5.2 ms.
The code was developed and tested on Ubuntu 14.04.2, on a system with an AMD R9 270X GPU.
The complete documentation is available here.
For more details on the implemented algorithms, take a look at the project's wiki.
The project has a dependency on CLUtils (which is automatically downloaded by cmake). If you'd like to remove this dependency, you should be able to modify the kernel interface classes with minimal effort.
Currently, there are 3 example applications. For guided_filter_image, you'll need OpenCV. For guided_filter_kinect_rgb and guided_filter_kinect_point_cloud, you'll need a Kinect and libfreenect.
git clone https://github.com/nlamprian/GuidedFilter.git
cd GuidedFilter
mkdir build
cd build
cmake -DBUILD_EXAMPLES=ON ..
# or to build the tests too
cmake -DBUILD_EXAMPLES=ON -DBUILD_TESTS=ON ..
make
# to run the examples (from the build directory!)
./bin/guided_filter_image
./bin/guided_filter_kinect_rgb
./bin/guided_filter_kinect_point_cloud
# to run the tests (e.g.)
./bin/guided_filter_tests_box
# or with profiling information
./bin/guided_filter_tests_box --profiling
# to install the libraries
sudo make install
# you'll need to copy manually the kernel
# files into your own projects
# to build the docs
make doxygen
firefox docs/html/index.html