Can Exp-Golomb coding encode zero?

Yes. The order-zero code for zero is 1.

Is it the same as Elias gamma coding?

Order-zero Exp-Golomb corresponds to leading-zero gamma coding of n + 1, while the naming and input domain differ.

Does a signed value include a separate sign bit?

Not necessarily. A specified mapping can interleave positive and negative values into nonnegative code numbers.

AI Glossary / MLOps and deployment / Intermediate

Exponential-Golomb coding

Definition

Exponential-Golomb coding, often shortened to Exp-Golomb coding, is a family of self-delimiting binary codes for nonnegative integers. It combines a prefix indicating magnitude with additional binary information. The order-zero version gives very short codewords to zero and other small values and is used for selected syntax elements in video-coding standards.

What it means

A variable-length code can save space when small values occur frequently but occasional larger values must still be represented. Exp-Golomb coding does this without assigning every value the same fixed width. The family includes orders that alter the balance between prefix and suffix. It is related to Elias coding, but should not be treated as identical to ordinary Golomb quotient-remainder coding with an arbitrary divisor. The exact variant and any signed-value mapping are part of a format's specification.

How it works

For order zero, add one to the nonnegative integer n, write n + 1 in binary, and prefix that representation with one fewer zero than its binary length. A decoder counts leading zeros, reads the corresponding binary value beginning at the first one, and subtracts one. This is the leading-zero Elias gamma representation of n + 1. Higher orders modify the construction and should be read from the applicable specification. A signed syntax element first maps signed values to nonnegative code numbers according to a defined rule.

Three practical examples

A video-bitstream parser reads an unsigned parameter coded with the specification's Exp-Golomb rule.
A developer stores nonnegative counts that are usually small and compares the code's bit cost with fixed-width fields.
An educational codec maps small signed prediction differences to code numbers, then demonstrates why the mapping affects which values receive the shortest representations.

Think of it this way

Think of a label whose short opening announces how many digits are needed for the number that follows. Zero receives a particularly short label because the code adds one before representing the value. The analogy explains variable size, but the actual bit sequence must follow a precise convention; it is not an informal abbreviation scheme chosen independently by each encoder.

Worked example

For order zero, n = 0 becomes binary 1 and is coded as 1. For n = 1, n + 1 = 2 is binary 10, so the code is 010. For n = 5, n + 1 = 6 is binary 110, giving 00110. To decode 00110, count two leading zeros and read the next three-bit binary value, 110 = 6; subtract one to recover 5. Under the common signed mapping used for H.264 signed Exp-Golomb syntax, values 0, +1, −1, +2 and −2 map to code numbers 0, 1, 2, 3 and 4. Other formats must explicitly state their own mapping.

Illustrative case study

Fictional case: Walnut Video builds a small metadata parser for an instructional stream. Its initial implementation mistakenly applies the unsigned result directly to a signed field. Positive values appear plausible, but negative values become incorrect positive numbers. The team separates bit parsing from signed mapping and adds examples for zero and both signs. This fictional case illustrates a parsing error and its correction, not the results of a real codec implementation.

Common misconceptions and mistakes

Do not omit the addition of one in the order-zero construction, because that is what allows zero to be represented. Do not assume all values in a video stream use Exp-Golomb coding; standards employ several coding processes. Long runs of leading zeros in malformed input can trigger excessive reads or integer overflow unless the parser enforces the syntax's bounds. A valid variable-length code identifies a value's endpoint, but it does not itself verify that the decoded value is permitted in that field.

Business use

Businesses usually encounter this term when integrating codecs, analyzing media files or developing specialist binary protocols. The practical requirements are interoperability, robust parsing and accurate size estimates. Use the exact standard version and syntax definition, including permitted ranges and signed mappings. Compare coding methods on the actual distribution of values. A method designed to favor small integers can be inefficient if large values dominate the workload.

Check your understanding

Decode the order-zero bitstring 011.

Answer: There is one leading zero, so read the next two bits as binary 11 = 3. Subtract one to obtain code number 2. If a field uses the stated H.264 signed mapping, that code number represents −1 instead.

Related terms

Golomb coding
Elias gamma coding
Rice coding
syntax element
signed mapping
variable-length code

Learn next

Build an order-zero table for values 0 through 7 and decode the concatenated codes. Then apply a signed mapping to a few positive and negative values. Keep parser, mapping and allowed-range checks as separate steps so that each can be verified clearly.

Sources

Primary specification: ITU-T H.264, including the Exp-Golomb parsing and signed-mapping clauses, https://www.itu.int/rec/T-REC-H.264 ; 2024 specification PDF, https://www.itu.int/rec/dologin_pub.asp?id=T-REC-H.264-202408-I%21%21PDF-E&lang=e&type=items . Primary specification search excerpts consulted; full clause-level and editorial review remain pending.

Previous: Exponential Function Next: Exponential growth