Below GIF made with library PygameShader (GPU shader ==> wave_gpu method). Wave_gpu method is the cython method developped in this github page ported to CUDA for higher resolutions and better perforamces. For a full GPU demo install PygameShader(*) and run the demo PygameShader/Demo/GPU_demo_ripple.py
(*) refers to the dependencies section (CUDA must be installed).
This github page host the full CPU demo, please click on the new release version 1.0.2 as shown below
Click on the executable setup_RippleEffect.exe to download the installer and run it onto your machine, The installation process might be flagged as unsafe by your anti-virus.
The most recent versions of the below algorithms are now integrated within the project PygameShader, and will be updated from that project only.
- Water ripple effect and seabed distortion (GPU)
- Surface water ripple effect (CPU method)
This section describes two techniques for performing a blur effect using pure Python. These implementations are educational and prototypical in nature, and not optimized for high-performance real-time rendering.
Overview
This approach loops through each pixel of an image (of dimensions width × height) and applies a blur effect by averaging the values of neighboring pixels.
How It Works
- Each pixel is visited individually.
- Neighboring pixels are accessed directly using nested loops.
- The result is stored in a new image or buffer.
Performance
- Very slow, especially for large images.
- Not suitable for real-time rendering above 100×100 pixels.
- Slight speed improvements may occur in Python 3.11+ due to interpreter optimizations.
Overview
This method uses NumPy's roll() function to simulate directional blurring, eliminating the need for nested loops. This results in a highly vectorized and much faster computation.
Concept
Each directional blur is calculated using a roll operation along a specific axis, and the resulting arrays are summed to simulate a convolution-like effect:
| Convolution Direction | Pixel Offset | NumPy Operation |
|---|---|---|
| Down | (x, y + 1) | np.roll(array, +1, axis=0) |
| Up | (x, y - 1) | np.roll(array, -1, axis=0) |
| Left | (x + 1, y) | np.roll(array, -1, axis=1) |
| Right | (x - 1, y) | np.roll(array, +1, axis=1) |
You can sum these shifted arrays along with the original to produce a blur effect. Optionally, divide by the number of terms to average.
Performance
- Approximately 300× faster than the pure Python loop.
- Still not ideal for real-time use with large resolutions.
The Cython version into the folder Cython and CythonOpenMP (multiprocessing) provide performance of 100 FPS for a screen 600 x 600 with water ripple effect with seabed distortion.
- Pygame 3 (pip install pygame)
- Numpy (pip install numpy)
- Cython (refer to a tutorial to see how to install cython)
- A compiler such visual studio, MSVC, CGYWIN setup correctly on your system.
- a C compiler for windows (Visual Studio, MinGW etc) install on your system and linked to your windows environment. Note that some adjustment might be needed once a compiler is install on your system, refer to external documentation or tutorial in order to setup this process. e.g https://devblogs.microsoft.com/python/unable-to-find-vcvarsall-bat/
C:\>python setup_RippleEffect.py build_ext --inplace
When python has been compiled, run the program main_RippleEffect.py from your favorite python IDE.
Or run the demo executable where the fodler Asset has been copied: C:>demo.exe