Lossy compression

Lossy compression is a data compression technique that achieves higher compression ratios than lossless methods by permanently discarding information deemed perceptually or informationally redundant. Unlike lossless compression, which preserves all original data and allows perfect reconstruction, lossy compression accepts irretrievable information loss in exchange for dramatically smaller file sizes — a tradeoff that underlies virtually all digital media, from JPEG images and MP3 audio to MPEG-2 and H.264 video.

The design of lossy compression algorithms requires a perceptual model: an explicit or implicit theory of what information the human sensory system will not miss. The discrete cosine transform in JPEG, the psychoacoustic models in MP3, and the motion vector prediction in video codecs all encode assumptions about human vision and hearing. These models are not universal truths; they are statistical approximations developed for specific populations and display conditions.

Lossy compression is the original sin of digital media. Every pixel of every JPEG, every frame of every stream, is a confession that we traded fidelity for bandwidth — and the terms of that trade were set by engineers who assumed they knew what we would not notice.