This application allows a user to remove moving objects from an "image". Given a video taken by a still camera, the program will generate an image with infrequent elements removed from the image. For instance, given a clip of a man running through the field, it will ideally remove the man from the field.
The program performs its operation by searching for the most common pixel value for a given pixel across all frames. The provided software allowed this to be done with either k-means or DBSCAN, although other clustering algorithms would also work in its place. The operation is applied across each pixel in order to generate an output image.
The program utilizes ffmpeg in order to split a given clip into frames.
Using a package manager, this can be done with one of these methods:
# Apt-Get
sudo apt-get install ffmpeg
# Homebrew
brew install ffmpeg
# Pacman
sudo pacman -S ffmpeg
The program also relies on the use of python, which can be similarly
installed. Finally, the program requires numpy, which is used in the
computation of the clusters, Pillow, which is used for image manipulation,
and scikit-learn, which is used to perform the clustering. These can be
acquired using pip like so:
pip install numpy scikit-learn Pillow
To run the program, simply run
python main.py
along with any necessary arguments.