Converting an image to its DFT representation, we can compress images by dropping (setting to 0) signals of the DFT that are less than a specified tolerance. As a result, this "drop ratio" corresponds to image compression, as we don't need to store pixels that we set to zero.
Run the notebook. You can replace operahall.png with any other image, ensuring that it is a square image, dimensions of multiple 32, e.g. 512x512.
The final plot is an error plot, which illustratively shows which values were dropped in a specified compression.
The pdf also shows the result of running the ImageCompression DFT algorithm on the operahall.png example.