Lossy Compression

Your eyes and ears are not equally sensitive to everything. You notice big changes in brightness much more than tiny shifts in color, and you hear mid-range tones much better than extremely high or low frequencies. Lossy compression exploits this gap. It does not treat every bit of data as equally important. Instead it uses a model of human perception to find the data you would never miss, and throws that away first, keeping file size down while keeping the result looking or sounding right to you.

For images, a lossy algorithm often groups nearby pixels and simplifies small color variations within each group, since your eye blends subtle color differences together anyway, while it preserves sharp brightness changes, since your eye is very sensitive to edges and contrast. For sound, the algorithm analyzes which frequencies are present and removes ones outside typical human hearing range, or ones masked by louder nearby sounds you would not notice anyway. In both cases the deleted data is gone permanently. You cannot recover the original file from the compressed one.

This connects directly to how data gets represented as bits in the first place: compression only works because images and sound are already stored as numbers that can be approximated or rounded. It also sets up a tradeoff that runs through the rest of the unit, between file size and fidelity. Compress once lightly and the loss is invisible. Compress repeatedly or too aggressively and the damage compounds, visible as blurry blocks in images or tinny, hollow audio.