A OpenCV-based image rectification tool for a camera based ambilight on a raspberry pi.
I wrote this small piece of code because I wanted an ambilight on my TV that worked for all contents displayed on the TV (not only certain HDMI-sources etc.). That's why I chose a camera based approach: A camera films the TV, a tiny bit of image processing rectifies the image and maybe does some color correction, and then hands it off to an ambilight software.
...
This application takes images from a Video4Linux-Device (the camera), rectifies them using the supplied configuration files (config.yaml and camera.yaml) and outputs the rectified images on a different V4L-Device, which has to be set up by v4l2loopback.
How this works is not that complicated - the code should suffice for now.
To build the software, the following dependency needs to be satisfied.
- OpenCV 3
Then building should be fairly straightforward. In the main directory:
mkdir build
cd build
cmake ..
cmake --build .
This should build a pibilight-executable in the build directory.
...
There are a lot of ways this project can be improved. Here are some I can think of.
- Automatically detect the screen and adapt the rectification (useful if the camera is accidentally moved).
- Get rid of dependency to libyuv
Feel free to let me know about issues or anything. Also fork and implement stuff and feel free to make PRs. I would be very happy to know if you are using this for a project of your own! :)