JPEG 2000 in digital cinema

Consumer video codecs save space by describing how each frame differs from the ones around it. JPEG 2000 does not: every frame is encoded as an independent image using wavelet compression. That costs a great deal of storage, but it buys properties cinema cares about — no motion artefacts between frames, graceful degradation rather than blocking when data is tight, and the ability to decode any frame without its neighbours.

The same codestream can also be decoded at 2K from a 4K file, which is how a 4K DCP plays on a 2K projector without a separate version.

It explains the file sizes. A feature at typical cinema bitrates runs to well over a hundred gigabytes, which is why DCPs arrive on drives or by satellite and broadband delivery rather than as downloads, and why ingest takes as long as it does.

It also explains why a DCP made by a free tool such as DCP-o-matic can take hours to encode: it is compressing every one of roughly 150,000 frames of a feature as a separate image.