This is the CPU and GPU(based on Opencl) implementation of canny edge detection for High Performance Computing with Graphic Cards praktikum.

Canny edge detection consists of 5 steps.

1. Gaussian Filter
2. Sobel Filter
3. Non-Max Suppression
4. Double Threshold
5. Hysteresis edge tracking

Here is how you can use this code to determine edges in the images and the intermediate results.

Instructions to run this code in Visual Studio

1. Download NVIDIA CUDA Toolkit from https://developer.nvidia.com/cuda-downloads
2. Download Boost library from https://sourceforge.net/projects/boost/files/boost-binaries/1.76.0_b1/boost_1_76_0_b1-msvc-14.2-64.exe

1. Download the Canny_edge_detection_opencl project from the github.
2. In the CMakeLists.txt change the boost include directory and boost library directory to system include folder.
   • set(BOOST_INC "C:/local/boost_1_76_0_b1_rc2")
   • set(BOOST_LIB "C:/local/boost_1_76_0_b1_rc2/lib64-msvc-14.2")

1. Open Visual Studio-> Choose "Open a local folder" -> and select "Canny_edge_detection_opencl" folder.
2. CMake generation will automatically start.
3. Build the project (Ctrl + Shift + B)
4. In the solution explorer Choose "Switch between solutions and the available views" Choose "CMake Targets View" Expand "CannyEdgeDetection Project" Configure the "CannyEdgeDetection (executable)" as a startup item.

1. Execution is started by clicking "CannyEdgeDetection.exe" play button.
2. For each of these steps, a CPU and GPU output image will be generated and saved in out/build/x64-Debug.

1. Different sample examples are placed in the images directory (src/InputImages).
2. To run different example set we have to change the image name in the CannyEdgeDetection.cpp source file.

   • Canny_edge_detection_opencl/src/CannyEdgeDetection.cpp

     Line 211 in e03c6ea

     Core::readImagePGM("../../../src/InputImages/Lizard.pgm", inputData, inputWidth, inputHeight);

3. Rebuid the project and execute.