DeltaEncoding

Delta encoding is a data encoding technique that stores the difference between sequential data elements rather than the elements themselves. It is used to exploit temporal redundancy in sequences such as time series, logs, or image and video frames, reducing storage and bandwidth when values change gradually.

How it works: a base value is stored first, and each subsequent sample is represented by its delta, the difference from the previous sample. Deltas are typically encoded with compact or variable-length representations to save space. Reconstruction adds each delta to the previous value to recover the original sequence.

Example: the sequence 100, 102, 105, 108 can be stored as 100 plus deltas 2, 3, 3. This can dramatically reduce data size when consecutive values are close.

Variants and domains: delta encoding can use fixed or adaptive bases, and deltas can be signed integers or bitwise differences. It is common in time-series databases, log and event compression, and in distributed version control systems where packs store changes between objects. In multimedia, delta frames in inter-frame compression encode changes relative to a reference frame, rather than storing complete frames.

Advantages and limitations: delta encoding is effective when data changes slowly or predictably, yielding smaller representations. Its efficiency diminishes with highly variable data and large deltas, and it may introduce overhead for storing the initial base value and metadata. Error propagation can occur if a delta is corrupted, and random access often requires decoding from the base up to the point of interest.

See also: differential encoding, delta coding.