PixiCUDA Motion Blur Filter for images with CUDA acceleration

        — by Mykyta Kyselov (TheMegistone4Ever).

Table of Contents

1. Getting Started
   1. Deploying the Software
   2. Purpose of Development
2. Preparing to Work with the Software
   1. System Requirements
   2. Software Requirements
3. Running the Program
   1. The main page of the PixiCUDA
      1. An example of CPU usage
      2. An example of GPU (CUDA) usage with 64 threads per block
      3. Graph of acceleration of the parallel algorithm
      4. An example of benchmark usage
4. License

1. Getting Started

1.1 Deploying the Software

To deploy the motion blur filter application, follow these steps:

• Install CUDA Toolkit 12.2 or later from the official NVIDIA website.

• Install the latest version of CMake.

• Install the latest version of Visual Studio.

• Build the OpenCV library from source with CUDA 12.2 or later support for Visual Studio you have installed. The instructions for building OpenCV with CUDA support can be found here.

• Create a new CUDA 12.2 (or or later) Runtime project in Visual Studio or another C++ environment and clone the PixiCUDA repository:

  git clone https://github.com/TheMegistone4Ever/PixiCUDA.git

• Open the project in Visual Studio.

• Edit path to images in main.cpp file.

• Build the project in Visual Studio into an executable (.exe) file.

• Run the application.

• Select needed parameters. Application will create a new image with motion blur effect.

• Enjoy the result!

1.2 Purpose of Development

The development aims to provide users with a simple and fast way to apply motion blur effect to images using CUDA acceleration. The application is designed to be user-friendly and easy to use. The application is intended for users who want to apply motion blur effect to images with CUDA acceleration. It also providef usual CPU version of motion blur filter for time comparison.

2 Preparing to Work with the Software

2.1 System Requirements

Minimum Hardware Configuration:

• Processor type: Intel Core i3 or equivalent;
• RAM: 2 GB;
• Graphics card: NVIDIA GeForce 600 series or later;
• Required Driver Version for CUDA Compatibility: >=525.60.13 for Linux, >=528.33 for Windows;
• Internet connection: Not required;
• Disk space: 1 GB;
• C++ compiler: Visual Studio 2019 or later.

Recommended Hardware Configuration:

• Processor type: AMD Ryzen 9 6900HS with Radeon Graphics or equivalent;
• RAM: 16 GB;
• Graphics card: NVIDIA GeForce RTX 3050 or later;
• Required Driver Version for CUDA Compatibility: >=525.60.13 for Linux, >=528.33 for Windows;
• Internet connection: Not required;
• Disk space: 2 GB;
• C++ compiler: Visual Studio 2022 or later.

2.2 Software Requirements

Minimum Software Configuration:

• Operating system: Windows 10 or later;

• C++ compiler: Visual Studio 2019 or later;

• CUDA Toolkit 12.2 or later;

• CMake 3.21 or later;

• OpenCV 4.5 or later.

• Git 2.33 or later.

• Recommended Software Configuration:

• Operating system: Windows 11 or later;

• C++ compiler: Visual Studio 2022 or later;

• CUDA Toolkit 12.2 or later;

• CMake 3.28 or later;

• OpenCV 4.9 or later.

• Git 2.43 or later.

3 Running the Program

Launch the motion blur filter application PixiCUDA by running the PixiCUDA.exe file, located in the PixiCUDA.zip in release, and you will be presented with the main window of the application. You can also select the needed image and paste it in the images folder in the project directory. Rename it to to_launch.png and run the application.

3.1 The main page of the PixiCUDA

3.1.1 An example of CPU usage

3.1.2 An example of GPU (CUDA) usage with 64 threads per block

3.1.3 Graph of acceleration of the parallel algorithm

3.1.4 An example of benchmark usage

4 License

The project is licensed under the CC BY-NC 4.0 License.