Stephen Seo
c1c271b77e
Previously, the code only checked the first OpenCL platform and failed if thet platform didn't have any devices. |
||
|---|---|---|
| doxygen | ||
| res | ||
| src | ||
| .gitignore | ||
| .lvimrc | ||
| CMakeLists.txt | ||
| Doxyfile | ||
| LICENSE | ||
| README.md | ||
Dithering with OpenCL
Note about Maintenance and Versions
The project at tag 1.0.0 refers to the state of the project upon completion
of the course. Newer versions are meant to be maintenance releases; fixes that
keep the project working. As of 2022-01-18, there may be (after now) a version
1.1.0 or 1.0.1 due to keeping up to date with the FFmpeg dependency as a new
major version of FFmpeg was released (version 5.0).
To be clear, version 1.0.0 of this project is expected to work with FFmpeg
version 4.4.
Fixes were applied for FFmpeg version 5. A macro is used to change incompatible code based on what version of FFmpeg is specified in the headers, so external checking of version is not necessary. Thus, the latest version of this repo should work with both FFmpeg 5.0 and 4.4.
Background
The proposal mentioned dithering live-input frames, but I ended up with just dithering single images or turning a video into dithered images.
Usage
A "blue-noise" image is integral to dithering images. A generated blue-noise image is provided in the "res/" directory. Run the program with "--help" to get info on how to use it.
Blue-noise images were generated using my other project that generates blue-noise with/without OpenCL.
PNG, PGM, and PPM image formats are supported.
For decoding video, any format that ffmpeg can read should work (though if things don't work, try using MP4 files).
WARNING: Video decoding is still a WIP. The video is currently decoded, but an
output video being encoded hasn't been implemented yet. The current
implementation writes each video frame to a PNG image in the current
directory.
Video decoding and encoding is finished, but there is some noticable drops in
quality when encoding to colored dithered video but the resulting video may
end up being too large. It may be better to just output each frame to individual
PNGs, then combining them later just like as mentioned here with the params you
want set: https://trac.ffmpeg.org/wiki/Slideshow .
Other Notes
I plan on adding the MIT License to this project once the course (that this
project was made for) is over.
I went ahead and applied the MIT License earlier than noted here.
C++ Compatability
The source is written to work with C++11 but not any higher. The only significant part that this affects are the functions in the "Image" class that return an optional value (non-optional on success). Ideally, C++17's std::optional would be used in that case, but since I am keeping things to work up to C++11, std::unique_ptr is used instead.
Project Presentation
Click here to view the presentation
Legal Stuff
This program uses FFmpeg which is licensed under the LGPL 2.1 license
This program uses libpng which is licensed under the PNG Reference Library License Version 2